Base_JSON/prefixW/json/W97/W97-0118.json

{
    "paper_id": "W97-0118",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T04:36:20.672395Z"
    },
    "title": "The Effects of Corpus Size and Homogeneity on Language Model Quality",
    "authors": [
        {
            "first": "Tony",
            "middle": [
                "G"
            ],
            "last": "Rose",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "Nicholas",
            "middle": [
                "J"
            ],
            "last": "Haddock",
            "suffix": "",
            "affiliation": {},
            "email": ""
        },
        {
            "first": "Roger",
            "middle": [
                "C F"
            ],
            "last": "Tucker",
            "suffix": "",
            "affiliation": {},
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "Generic speech recognition systems typically use language models that are trained to cope with a broad variety of input. However, many recognition applications are more constrained, often to a specific topic or domain. In cases such as these, a knowledge of the particular topic can be used to advantage. This report describes the development of a number of techniques for augmenting domain-specific language models with data from a more general source. Two investigations are discussed. The first concerns the problem of acquiring a suitable sample of the domain-specific language data from which to train the models. The issue here is essentially one of quality, since it is shown that not all domain-specific corpora are equal. Moreover, they can display significantly different characteristics that affect the quality of any language models built therefrom. These characteristics are defined using a number of statistical measures, and their significance for language modelling is discussed. The second investigation concerns the empirical development and evaluation of a set of language models for the task of email speech-u>-text dictation. The issue here is essentially one of quantity, since it is shown that effective language models can be built from very modestly sized corpora, providing the training data matches the target appfication. Evaluations show that a language model trained on only 2 million words can perform better than one trained on a corpus of over 100 times that size.",
    "pdf_parse": {
        "paper_id": "W97-0118",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "Generic speech recognition systems typically use language models that are trained to cope with a broad variety of input. However, many recognition applications are more constrained, often to a specific topic or domain. In cases such as these, a knowledge of the particular topic can be used to advantage. This report describes the development of a number of techniques for augmenting domain-specific language models with data from a more general source. Two investigations are discussed. The first concerns the problem of acquiring a suitable sample of the domain-specific language data from which to train the models. The issue here is essentially one of quality, since it is shown that not all domain-specific corpora are equal. Moreover, they can display significantly different characteristics that affect the quality of any language models built therefrom. These characteristics are defined using a number of statistical measures, and their significance for language modelling is discussed. The second investigation concerns the empirical development and evaluation of a set of language models for the task of email speech-u>-text dictation. The issue here is essentially one of quantity, since it is shown that effective language models can be built from very modestly sized corpora, providing the training data matches the target appfication. Evaluations show that a language model trained on only 2 million words can perform better than one trained on a corpus of over 100 times that size.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "The development of robust speech recognition technology offers great potential for the design of improved interfaces to a wide range of applications. The current project concerns the development of one such application: the speech-to-text dictation of email messages. The work makes use of the Abbot recogniser, which is a eonnectionist/HMM continuous speech recognition system developed by the Connectionist Speech Group at Cambridge University. It is designed to recognise British English and American English, clearly spoken in a quiet acoustic environment (I-Ioehberg et al., 1994 ).",
                "cite_spans": [
                    {
                        "start": 560,
                        "end": 584,
                        "text": "(I-Ioehberg et al., 1994",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1."
            },
            {
                "text": "The Abbot system is available with a vocabulary of 20,000 words, which means that anything spoken outside this vocabulary cannot be recognised (and therefore will be recognised as another word or string of words). The vocabulary and grammar 0-aM) were optimised for the task of reading from a North American Business newspaper, in this case the Wall Street Journal. Some 227 million words of training text were used in building this LM and it is widely used throughout the speech community. However, despite the size of the original training corpus, this LM was dearly not designed for the specific task of email dictation, so its performance is likely to be sub-optimal. However, a new vocabulary and LM can .I I easily be created and then substituted for the one supplied. The LMs described in this paper were all 'back-off trigram' LMs (Katz; 1987) , built using the CMU SLM toolkit (Rosenfeld, 1994) .",
                "cite_spans": [
                    {
                        "start": 839,
                        "end": 845,
                        "text": "(Katz;",
                        "ref_id": "BIBREF11"
                    },
                    {
                        "start": 846,
                        "end": 851,
                        "text": "1987)",
                        "ref_id": "BIBREF11"
                    },
                    {
                        "start": 886,
                        "end": 903,
                        "text": "(Rosenfeld, 1994)",
                        "ref_id": "BIBREF20"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1."
            },
            {
                "text": "In order to build a LM for the task of email dictation, it is necessary to acquire a corpus of suitable email training data. However, behind this ostensibly simple objective lie several subtle challenges. \"Email\", as a general term, describes a great variety of types of communication. These types are perhaps best illustrated by considering the range of functions that email messages typically provide. For example, email can be used as a medium for:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "\u2022 a formal face-to-face meeting;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "\u2022 a casual face-to-face chat;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "\u2022 a broadcast (e.g. \"Tannoy\") message;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "\u2022 requesting information;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "\u2022 replacing an office memo;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "\u2022 replacing a phone call, etc.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "Clearly, the purpose of each communication can be very different, and the language used will reflect this. Furthermore, apart from the issue of domain (i.e. subject matter), the type of language used will also vary according to the role of the participants. For example, when requesting advice from a mailing list one would tend to be more formal and polite than when requesting the same advice from a friend or colleague. Consequently, it would appear that email messages vary almost as much as spontaneous, spoken dialogue. If this is indeed so, then the prospects for building effective language models for email may appear somewhat limited. Clearly, in order to move forward, it is necessary to define some limits.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "The first of these concerns the quantity. What is a reasonable size for an email corpus? There are few precedents for this so the question was answered empirically, by finding a compromise between the need to acquire sufficient training data and the need to complete the acquisition phase within a reasonable space of time. However, the time taken to reach a certain quantity depends very much on the source of the data, which forms the second limit: from where should the email be acquired? A range of possibilities exists, e.g. the Intemet (i.e. bulletin boards, mailing lists, email archives) or specific individuals (i.e. previously saved messages, day-by-day output).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "Evidently, emaiI acquired from the Intemet exhibits a wide range of authorship, function and subject matter. In addition, downloading large quantities of text from such sources without the authors' consent may involve certain copyright issues. Clearly, the limits of the source are more easily defined if the email is restricted to the output of a group of specific individuals. However, unless the group is very large, acquiring just 1 million words from their day-by-day output would be too slow to enable the acquisition to be completed within a reasonable space of time. Therefore, individuals with a large collection of previously saved messages were identified as more suitable candidates. Furthermore, a restriction that all members of this group must be employees of HPLB (Hewlett-Packard Labs, Bristol) placed a further constraint on the source. To ensure controlled authorship, only the outgoing messages of these individuals were collected.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The form of email messages",
                "sec_num": "2.1"
            },
            {
                "text": "The \"content\" of an email message is not an easy concept to define. Evidently, the body contains much important data, but what about the other elements, e.g. headers, signatures, quoted sections, etc. -what roles do they play? In the case of headers, a cursory analysis revealed that they could safely be discarded since few contained any useful information. However, other email components are not quite so easily categorised, e.g.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The content of email messages",
                "sec_num": "2.2"
            },
            {
                "text": "\u2022 quoted (included) messages: these are usually referred to by the message content, but are often the product of a different author; \u2022 emafl 'signatures': these are often quite verbose, but rarely contribute anything to the message content;",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The content of email messages",
                "sec_num": "2.2"
            },
            {
                "text": "\u2022 samples of Postscript/Latex: these were problematic, since people would often quote verbatim large passages to illustrate a point that did indeed contribute to the content of the message. However, to build models from data that included such heavily marked-up text is questionable.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The content of email messages",
                "sec_num": "2.2"
            },
            {
                "text": "Since the above items were all rendered as ASCII strings, their surface form could fairly reliably be predicted and they were therefore removed from the corpus using suitably designed \"filters\". However, there is a further number of email attachments that are not composed of predictable ASCII strings. These include items such as shafted or uuencoded files and word processor/DTP output. Evidently, such items need to be removed, but finding small fragments of such diverse data in a corpus of several million words is a non-trivial problem.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The content of email messages",
                "sec_num": "2.2"
            },
            {
                "text": "Alternatively, rather than trying to filter out \"noise\" from the email \"signal\", it is possible to adopt the converse approach, and try to identify those lines that constitute genuine English within the overall email data, which may then be retained as the true training data. It is possible to achieve this using various heuristics, e.g. \"retain those lines that contain at least 90% English words\". However, this approach assumes there exists some predefmed vocabulary, which is somewhat tautological since the vocabulary is one of the things we seek to define in the first place.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The content of email messages",
                "sec_num": "2.2"
            },
            {
                "text": "A programme of email data collection took place over a period of 2-3 weeks, following the principles described above. This resulted in the acquisition of some 4 million words of email data. The \"donors\" were asked to provide both previously saved messages and intermittent day-by-day output. This was necessary since (by their very nature) saved messages tend to possess some sort of significant content, and were therefore often of above average length. In contrast, much day-by-day email correspondence uses an informal dialogue that is heavily context dependent, and therefore may be no more than a single brief phrase or sentence. This was then filtered in the manner described above. The final output was a corpus of emall data of 1,962,280 words (49% of the original size). This was then partitioned (95% : 5%) into training and test data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The emall collection process",
                "sec_num": "2.3"
            },
            {
                "text": "It is theoretically possible to build a LM using the tiniest of corpora. On balance, however, the 2 million words of email training data look somewhat inadequate compared to the 227 million words used for the WSJ LM. The problem is that the coverage of the n-grams is likely to be sparse and any LM so built will be degenerate since it does not reliably predict the characteristics of the source. To illustrate, consider the distribution of unigram frequencies: a mere 14,137 word types (19%) in the email corpus have frequencies of 6 or greater. Therefore, to acquire a vocabulary of just 20k words without using frequencies of 5 or less clearly requires a training corpus larger than 2 million words.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Corpus adaptation",
                "sec_num": "3."
            },
            {
                "text": "It would be highly desirable therefore if a method could be devised whereby information from a large corpus could be combined with a smaller sample of the domain-specific training data to create an optimal language model. One such approach involves augmenting a base model built from a larger, more general corpus with information from a small sample of the domain-specific language. There is evidence to suggest that this method can improve recognition performance (e.g. Ruduicky, 1995, Vergyri, 1.995 ). An I alternative approach is to use a suitable similarity metric to acquire further \"email-like\" training data from the larger corpus (henceforth referred to as the \"background corpus\"), and then build a new language model from the combined text. This approach offers interesting possibilities regarding the development of a general methodology for corpus adaptation, by attempting to \"grow\" a suitable corpus of training data for any domain using only a small sample as a \"seed\". A number of ways to implement this technique have been developed. Broadly speaking, they fall into two categories: \"top-down\" methods and \"bottomup\" methods.",
                "cite_spans": [
                    {
                        "start": 472,
                        "end": 502,
                        "text": "Ruduicky, 1995, Vergyri, 1.995",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Corpus adaptation",
                "sec_num": "3."
            },
            {
                "text": "At its simplest, this approach involves a combination of manual inspection and regular expression searching to identify those parts of the background corpus that contain suitable material. It relies on a good classification scheme and reliable organisation of the background corpus. The British National Corpus (BNC) is a suitable example, since it contains 100 million words of modem English, both spoken and written, sampled from the widest range of materials. It is annotated with part-of-speech codes, and SGML-encoded according to the Text Encoding Initiative's Guidelines (Bumard, 1995) . It is therefore possible to use the SGML tags to identify suitable texts. For example, extracting 10 million words of text for a domain such as World Affairs is trivially easy, since domain information is encoded in the header of each individual file (of which there are over 4,000).",
                "cite_spans": [
                    {
                        "start": 578,
                        "end": 592,
                        "text": "(Bumard, 1995)",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "Since much HP email concerns the computing business, and the BNC classifies computing as a branch of Applied Science, it would appear that the 10 million words from Applied Science section of the BNC may prove sufficiently similar. Likewise, the 10 million words classified as Commerce and Finance may also prove suitable. The effect of such an addition would be to increase the size of the training corpus from 2 million words to 22 million, which constitutes an increase of 1100%.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "However, methods such as this cannot be justified by subjective judgement and anecdotal evidence. What is required is an objective measure that reliably identifies which of the domains in the BNC is most similar to HP email. There may be many standard statistical techniques for measuring the degree of similarity of two data sets, but not all are suitable for the task of comparing corpora (Clmreh et al., 1991) . For example, some assume a normal distribution, which is clearly inappropriate for textual data. What is needed therefore is a test that makes few assumptions about the distributions of the underlying data, but provides a directly usable measure of similarity. One such test is the rank correlation, using Spearman's S.",
                "cite_spans": [
                    {
                        "start": 391,
                        "end": 412,
                        "text": "(Clmreh et al., 1991)",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "The assumptions behind rank correlation are few. It measures the degree of monotonic association between two rankable variables. The distribution of r as normal (mean 0, variance I/(N-1), assuming independence) is asymptotic for large enough samples, and does not make any assumptions about normality. This test was therefore applied to the word frequency lists of each of the domains in the BNC and the email corpus, to identify which corpora were most similar. The correlation with the BNC as a whole was also measured. All calculations were based on Spearman's S, where D 2 denotes the sum of the squares of the differences between the ranks of each pair of word types, and N the number of ranked pairs:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "6D 2 r= 1 N 3 -N",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "It is known that a sublanguage corpus can have very different characteristics to a general corpus (Biber, 1993 ), yet it is not obvious how the position on this scale of a given corpus can be assessed.",
                "cite_spans": [
                    {
                        "start": 98,
                        "end": 110,
                        "text": "(Biber, 1993",
                        "ref_id": "BIBREF0"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "Consequently, it is necessary to determine the homogeneity of a corpus prior to performing any similarity measures, since it is not clear what a measure of similarity would mean if a homogeneous corpus was being compared with a heterogeneous one (Kilgarriff, 1996) . A homogeneity test was therefore performed on the corpus of each domain. This was calculated using the following algorithm:",
                "cite_spans": [
                    {
                        "start": 246,
                        "end": 264,
                        "text": "(Kilgarriff, 1996)",
                        "ref_id": "BIBREF12"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "\u2022 j \u2022 ! t 1.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "For each domain corpus, do (times 10) II 2.1 Divide the corpus into two halves, by randomly placing 5k-word chunks in one of two subco~o~ l 2.2 Produce a word frequency list (wfl) for each subcorpus; i 2.3 Calculate the rank correlation between the two subcorpora; Ill 3. Calculate the mean and standard deviation of r. World affairs. BNC: the whole of the BNC Email: the 2 million word email corpus For large samples such as these the rank correlation coefficient has a normal distribution with mean 0 and variance 1/(n-l) where n is the number of common words. Although the significance of the correlation is not in doubt, the differences are highly significant too. The difference between two rank correlation coefficients will be normally distributed with mean 0. The maximum possible value for the standard deviation is (l/~(nl-1))+(l/~(n2-1)) where nl,n2 are the two common vocabulary sizes. Any difference greater than about 0.03 is therefore significant, and there are many pairs for which this is true. It is therefore possible to rank the rank correlations, and hence the BNC domains.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "Evidently, the strongest correlation with the email corpus is from As (Applied Science). Interestingly, this figure is higher than that between email and the whole BNC. The second highest domain correlation is with Cf (Commerce & Finance). This agrees with intuitions based on a manual inspection of the contents of the email corpus. The table also shows a polarity of the BNC -the \"arts\" domains at one pole, attracting each other (e.g. Ar:Bt = 0.408) but repelling the sciences (e.g. Ira:As = 0.159). Similarly, the sciences attract each other (e.g. Np:As = 0.315). In the middle are domains such as World Affairs, Social Sciences & Commerce & Finance that correlate with both poles to varying degrees. Moreover, the Ernail co~us really stands out on its own, having a very poor correlation with the others (in many eases it is negative). This suggests that even if the most strongly correlated domains are chosen, it is difficult to justify augmenting the email corpus with texts selected from the BNC using this method. Table 1 also shows the results of the homogeneity tests. Email is by far the most heterogeneous, more so even than the \"Unclassified\" section of the BNC (!) This brings into question the results of the similarity calculations in which the email corpus was involved, and mitigates further against the strategy of augmenting the email corpus with texts selected using the top.down method.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 1024,
                        "end": 1031,
                        "text": "Table 1",
                        "ref_id": "TABREF0"
                    }
                ],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "These results also provide insight into the relationship between homogeneity and language model quality.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "A common measure of LM quality is perplexity (PP), which can be thought of as a measure of the \"branching factor\" (i.e. the average size of the set of words between which a speech recogniser must choose) when transcribing a single word of the spoken text. PP thus measures the recognition difficulty of the text relative to the given LM, and is measured by applying the model to a sample of test data.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "Consequently, a LM derived from a heterogeneous corpus should have a higher perplexity than an equivalent one derived from a more homogenous corpus. However, homogeneity is defined here as a measure of unigram distributions, whereas perplexity is usually calculated using n-grams (where n is usually <=3), so it is not clear to what extent the two measures would be related.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The top-down approach",
                "sec_num": "3.1"
            },
            {
                "text": "The top-down approach assumes that the BNC classification system is perfect, in that each text classified as belonging to a certain domain really belongs in that domain. However, this is ultimately a subjective judgement, and frequently more than one classification is possible or even preferable (Lewis, 1992) . Moreover, it is often the case that texts from the same medium are more similar to each other than texts from the same domain (e.g. a journal paper on computing may be more similar to a journal paper on geology than an item from a popular computing magazine, because the \"content\" features are lost among the much more salient \"'genre\" features). Besides, no classification system is 100% reliable, so techniques that are based on them will inherit this uncertainty. Furthermore, domains such as Applied Science are very coarse-grained: they contain many more types of material than just those of computing. Even if such corpora are subdivided to a further level of classification they still suffer the same problem, albeit at a finer level of detail.",
                "cite_spans": [
                    {
                        "start": 297,
                        "end": 310,
                        "text": "(Lewis, 1992)",
                        "ref_id": "BIBREF15"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The bottom-up approach",
                "sec_num": "3.2"
            },
            {
                "text": "An alternative strategy is to work in a \"bottom-up\" direction. In this approach, a similarity metric is used to fmd and extract related material from the background corpus, regardless of the top-down classification. This method may not be as structured as the previous approach, but it is more robust in that it involves no manual intervention and does not rely on correct organisation or SGML tagging of the background corpus. Moreover, it will not \"miss\" material that is classified under an unexpected domain or medium but is otherwise suitable. The success of this approach depends on the use of a reliable similarity metric (even more so than the top-down approach, since it is now being applied to each of the 4,000+ files in the BNC rather than the 10 domain-based collections). Using this statistic to find texts that are similar to email in the BNC could be achieved using the following algorithm: One way of evaluating this result is to go through the list and calculate the mean rank of the 61 \"Computergram International\" texts, which are typical of the sort of texts this technique should identify as being similar to the email corpus. If the technique is working perfectly, the mean rank should be 31. If it is completely random, the mean rank would be 2062. It transpires that the mean rank is 959.85 (std dev = 524.44). Clearly, this result is better than chance, but far from significant. One of the main reasons for this was a tendency to sometimes give high scores to texts that were actually too short to constitute reliable sainples (the BNC attempts to maintain a standard sample size but this is not always possible). A logical modification was therefore to ignore those texts for which the number of common words was below a certain threshold. A number of threshoIds were investigated, and the optimum value (determined empirically) was around 1,370 words. However, even with this modification, the mean rank remained as high as 818.41 (std dev = 407.81). It is possible to reduce this value still further, but only by compromising the overall recall value (i.e. genuine texts are eliminated along with the \"noise\").",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The bottom-up approach",
                "sec_num": "3.2"
            },
            {
                "text": "/BNCIi. O/HIHAIHAC /BNC/I.O/J/JO/JOV IBNC/i. OICICTICTX IBNCII.01FIFTIFT8 /BNC/I.0/G/G01G00 /BNC/I.0/ClCBICBU /BNC/i.0/CICBICBX /BNC/i.0/K/KRIKRG /BNCIi.0/H/HRIHRD /BNC/i.0/EIEEIEEB /BNC/i.O/F/FUIFUS /BNC/i. OIJIJ51J5B /BNC/i.O/HIH41H4P /BNC/i.O/GIGYIGY5 /BNC/i.0/KIKPIKPS /BNC/i.O/G/G51G54 /BNC/I.OIH/HKIHKM IBNCIi.0/F/FDIFDE /BNC/i.0/G/G5/G5A IBNC/I.0/J/",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The bottom-up approach",
                "sec_num": "3.2"
            },
            {
                "text": "However, there is a more fundamental limitation to the above methodology. The rank correlation statistic compares differences in rank, ignoring absolute value (which can be significant). To illustrate, consider a case where the word \"of' is ranked 3 in one corpus and 6 is another. This is a very important difference. Conversely, ff \"banana\" is ranked 10,000 in one corpus and 100,000 in another, this is a very insignificant difference. But the difference of ranks for \"of\" = 3, for \"banana\" = 90,000. Clearly this technique is missing something important. Consequently, it was decided to investigate an alternative measure: the Loglikellhood Ratio Statistic.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The bottom-up approach",
                "sec_num": "3.2"
            },
            {
                "text": "The Logllkelihood Ratio, G 2, is a mathematically well-grounded and accurate method for calculating how \"surprising\" an event is (Dunning, 1993) . This is true even when the event has only occurred once (as is often the case with linguistic phenomena). It is an effective measure for the determination of domainspecific terms (e.g. Daille, 1995) and can be also used as a measure of corpus similarity. In the case where two corpora are being compared, it is possible to calculate the G 2 statistic either for single words (using a ~ contingency table) or for a vocabulary of N words (an N><2 table). The analysis of the 4,000+ BNC fries was therefore repeated using the Loglikelihood (instead of rank correlation) as the similarity measure. This produced the following top and bottom 10 texts: Each fine shows the filename, the title of the text, the length of the contingency table and the value for G 2. These are sorted in ascending order since comparing two identical documents would produce a G 2 of zero.",
                "cite_spans": [
                    {
                        "start": 129,
                        "end": 144,
                        "text": "(Dunning, 1993)",
                        "ref_id": "BIBREF7"
                    },
                    {
                        "start": 332,
                        "end": 345,
                        "text": "Daille, 1995)",
                        "ref_id": "BIBREF6"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The bottom-up approach",
                "sec_num": "3.2"
            },
            {
                "text": "/",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "The bottom-up approach",
                "sec_num": "3.2"
            },
            {
                "text": "A brief inspection of the titles of the documents at the top of the list would indicate that the metric has not produced an improvemenL Moreover, it transpires that the mean rank of the CI texts is now 1171.98, with std dev = 178.54. However, as before, the number of common words is very small for some of the texts. Therefore, the filter was applied to ignore eases where there were fewer than 1,370 words in eornmon. This produced a mean rank of I25.15 (std. dev. = 75.62), which is significantly lower than that produced by the rank correlation (mean rank = 818.41, std. dev. = 75.621).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "!",
                "sec_num": null
            },
            {
                "text": "So despite the absence of apparently suitable candidates in the top 10, the overall accuracy of this technique (measured by the mean rank of the 61 CI texts) is higher. The G 2 statistic appears to be more suitable for this type of data since it uses the actual frequency values for the words in the wfls, rather than just their ranks. Other independent sources indicate that the G 2 produces results that appear to correspond reasonably well with human judgement (Dallle, 1995) .",
                "cite_spans": [
                    {
                        "start": 464,
                        "end": 478,
                        "text": "(Dallle, 1995)",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "!",
                "sec_num": null
            },
            {
                "text": "However, both the rank correlation and Loglikelihood Ratio both make use only of unigrarn information. Clearly, much of the information that humans use to measure textual similarity is found not (solely) in the individual word frequencies (unigrarns), but rather in the way they combine (n-grams). The logical next step is therefore to compare word bigrams (or trigrarns) instead of just unigrarn data. A variation on this would be to compare texts using the Loglikelihood applied to bigrams that are not necessarily adjacent, i.e. counting occurrences of wordl and word2 within a limiting distance of each other. Indeed, such methods have been previously used for actually building the LMs themselves, and have been successfully applied to both speech (Rose & Lee, 1994) and handwriting data (Rose & Evett, 1995) . Counting words within a limited window would be smoother than using strict bigrarns and eousequently less affected by the problems caused by sparse data (which are inevitable when small, individual text files are compared).",
                "cite_spans": [
                    {
                        "start": 753,
                        "end": 771,
                        "text": "(Rose & Lee, 1994)",
                        "ref_id": "BIBREF18"
                    },
                    {
                        "start": 793,
                        "end": 813,
                        "text": "(Rose & Evett, 1995)",
                        "ref_id": "BIBREF17"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "!",
                "sec_num": null
            },
            {
                "text": "Another interesting possibility is to use the LM itself as the similarity metric. From an information theoretic point of view, entropy is a measure of a eorpus's homogeneity, and the cross-entropy between two corpora is a measure of their similarity (Charniak, 1993) . After all, when a LM is applied to a test text to produce a perplexity score, this value is a measure of the cross-entropy which reflects how well the LM predicts the words in the text. So if a LM is trained on text that is very similar to the test text, then it should predict the test data well and the perplexity should be low. Conversely, ff the test text is very different from the training text, then the perplexity will be high. The perplexity score can therefore be used to measure textual similarity. Moreover, it has the advantage doing so by considering (typically) uuigram, bigrarn and trigram data. Indeed, this method has already been successfully used within the development of a similarity-based Interact search agent, and preliminary findings indicate that perplexity is indeed an effective corpus similarity measure (Rose & Wyard, 1997) .",
                "cite_spans": [
                    {
                        "start": 250,
                        "end": 266,
                        "text": "(Charniak, 1993)",
                        "ref_id": null
                    },
                    {
                        "start": 1103,
                        "end": 1123,
                        "text": "(Rose & Wyard, 1997)",
                        "ref_id": "BIBREF19"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "!",
                "sec_num": null
            },
            {
                "text": "However, the use of such an approach is not entirely beyond question. Firstly, the LM is being used as the representation of a training text against which similarity is to be judged, and yet it is, by definition, undertrained and therefore degenerate. Secondly, the method by which similarity is measured should ideally be independent to the method by which success is evaluated. To use perplexity both as a similarity metric and an evaluation metric implies a certain amount of circular reasoning. However, the use of such iterafive techniques is not totally without precedent within the LM eornmunity. Several research groups have reported the successful improvement of LMs using techniques that iteratively tune the LM parameters using new samples of training data (e.g. Jelinek, 1990) . So, this approach may transpire to be I 1 I I sufficient!y well principled to merit further investigation, i ",
                "cite_spans": [
                    {
                        "start": 774,
                        "end": 788,
                        "text": "Jelinek, 1990)",
                        "ref_id": "BIBREF9"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "!",
                "sec_num": null
            },
            {
                "text": "A LM is built by collecting trigram, bigram & uuigram data from a training corpus. However, it is not always desirable to store all of this data. Thresholds can be set such that some of the lower frequency ngrams are discarded. For example, a trigram cut-off of 5 implies that all the trigrams with frequencies of 5 or fewer in the training data are not used in building the model. Setting lower thresholds allows the model to focus on more frequent events, and produces a proportionately smaller model. The LMs described in this paper were built using the CMU SLM toolkit (Rosenfeld, 1994) which facilitated the construction of a variety of LMs representing a range of different settings for each of the pertinent parameters.",
                "cite_spans": [
                    {
                        "start": 573,
                        "end": 590,
                        "text": "(Rosenfeld, 1994)",
                        "ref_id": "BIBREF20"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "The first of these was the Email LM. This was constructed using a vocabulary of 20,000 words that was derived directly from the ernail training data. The bigram and trigram cutoffs were both set to zero. The second LM was buik from the whole of the BNC, using the same vocabulary as the Email LM (in order to ensure consistency). So although their n-grams had been based on general English rather than Email, their vocabulary was derived from the Email data. For comparison therefore, a third BNC LM was built, using a vocabulary derived directly from the BNC (rather than email). This allowed the comparative evaluation of the contribution of vocabulary vs. n-grams to the LM effectiveness (measured using both perplexity and word error rate). Due to memory constraints it was not possible to build the BNC models with cutoffs lower than 2-2. The fourth LM investigated was the 20k WSJ LM that is available from the Abbot ftp site at Cambridge University.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "The standard measure by which LMs are assessed is by calculating their perplexity using a sample of test data. This process is usually performed off-line, i.e. independently of the speech reeogniser for which the models are intended. For the models described above, testing was performed using the CMU toolkit, by applying each LM to a sample of 10,000 words from the transcriptions of a database of video mail messages, developed by Cambridge University as part of their \"Video Mail Retrieval using Voice\" project (Jones et al., 1994) . Evidently, this data is not actually spoken email, but its domain and genre are nevertheless closely related to email. Unfortunately, it was not possible to calculate the PP of the WSJ LM due to the absence of a readily available version in the correct format.",
                "cite_spans": [
                    {
                        "start": 515,
                        "end": 535,
                        "text": "(Jones et al., 1994)",
                        "ref_id": "BIBREF10"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "A second evaluation method is to integrate the I,M with the speech reeogniser and test the combined system using recorded speech data. The models can be interchanged between trials, allowing comparative evaluation by measuring the word error rate (WER) produced by each model. More precisely, the error rates are measured using two standard metrics, percentage correct and accuracy:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": ". ~r % Correct = ~ x 100% N (H-X) 2. Accuracy = x 100% N",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "where: H is the number of correct transcriptions (words in the utterance that are found in the transcription), D is the number of deletions (words in the utterance that are missing from the transcription), S is the number of substitutions (words in the utterance that are replaced by an incorrect word in the transcription), and I is the number of insertions (extra words in the transcription). Accuracy is more critical than %correct in that it directly penalises insertions. Deletions & substitutions reduce the value of H, since H = N -(D+S).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "As mentioned above, the VMR database is a collection of speech data with transcriptions (of which the latter were used in the above evaluation). The speech part contains audio files for 15 speakers, of which 10 were used in the current investigation. The Abbot recogniser was run using each combination of the 10 speakers' data files (as input) and each of the four LMs: email, BNC with email vocabulary, BNC and the WSI LM. The output transcriptions were assessed for %correct and accuracy using the HResults program, which is part of HTK -the Hidden Markov Model Toolkit (Young & Woodland, 1993) . Table 2 shows the results of this investigation. The results for %correct and accuracy show the combined effect of the recogniser and LM. The contribution of the LM depends on its vocabulary and perplexity. As the LM changes, it produces different behaviour in the combined system and therefore different types of errors (e.g. insertions, deletions & substitutions). The net effect is that the email LM produces the highest %correct and also the highest accuracy. It is around 5% better (on both measures) than the WSJ LM. This is significant, considering the tiny corpus from which it was derived (2 million vs. 227 million in the case of WSJ). In between these two extremes are the two BNC LMs -the one with the email vocabulary performs slightly better (-0.5%) than the one with the BNC vocabulary. Table 2 . %Correct, accuracy, and perplexity of the language models",
                "cite_spans": [
                    {
                        "start": 573,
                        "end": 597,
                        "text": "(Young & Woodland, 1993)",
                        "ref_id": "BIBREF25"
                    }
                ],
                "ref_spans": [
                    {
                        "start": 600,
                        "end": 607,
                        "text": "Table 2",
                        "ref_id": null
                    },
                    {
                        "start": 1402,
                        "end": 1409,
                        "text": "Table 2",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "The result for the PP testing is highly revealing. As described earlier, a corpus of low homogeneity should produce a LM of higher PP than a corpus of high homogeneity. This is indeed shown to be the case, since the PP for email is 261.58 (homogeneity = 0.362), whereas the PP for the BNC is 227.54 (homogeneity = 0.687). These PP values are calculated using the 10K test data sample from the transcriptions of the VMR project. The higher PP value for email would tend to indicate that this is the poorer LM. However, it is clear that when used on the real spoken data, the email LM provides the lowest error rates. Initial explanations for this centred on the vocabulary, since a higher incidence of out-of-vocabulary (OOV) words can produce a lower PP but a higher WER. However, the email LM performs better (by 0.88% ,",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "I I I I k ~",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "correct) than the BNC/email LM even though both share the same vocabulary. Two explanations for this are possible. Firstly, there may be n-grams in the email corpus that are simply not found in the BNC (even though the BNC is 50 times larger). Secondly, the email LM may be better because it \"wastes\" less probability mass on n-grams that never actually occur in the test data. This implies that quality, not quantity, is a major factor in training effective LMs. Further PP testing, possibly using the complete transcriptions of the VMR data is necessary to clarify this issue.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "Evidently, the choice of vocabulary also makes an important contribution. The BNC LM with the email vocabulary performs better (by 0.64% correct) than the BNC LM with the BNC vocabulary, so clearly the email vocabulary provides better coverage of the test data. In fact, it is possible to directly compare the OOV rates with the performances shown above: the BNC LM with the ernail vocabulary has an OOV rate of 1.16% on the VMR data, and a %correct of 54.04. By contrast, the BNC LM with the BNC vocabulary has an OOV rate of 1.69% and a %correct of 53.40. These figures suggest that an increase in OOV rate of 0.56% leads to a reduction in %correct of 0.64%, or, in other words, a 1% increase in OOV rate produces a reduction in %correct of around 1.14%. Interestingly, this figure correlates extremely well with the results of a similar experiment performed by , who found that a 1% increase in the OOV rate can lead to a 1.2% increase in the word error rate.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Language model quality",
                "sec_num": "4."
            },
            {
                "text": "The analysis of the corpora has provided several revealing insights. Firstly, it is necessary to determine the homogeneity of a corpus prior to performing any similarity measures, since it is not clear what a measure of similarity would mean if a homogeneous corpus was being compared with a heterogeneous one. A methodology for calculating homogeneity has been described and the accuracy and usefulness of this is further described in Kilgarriff (1997) .",
                "cite_spans": [
                    {
                        "start": 436,
                        "end": 453,
                        "text": "Kilgarriff (1997)",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5."
            },
            {
                "text": "Clearly, the ernail corpus is highly heterogeneous. This means it is particularly prone to \"burstiness\" and unpredictability, which affects all levels of n-grams (including unlgrams). This may be due in part to the particular training corpus used, but it is more likely to be inherent to the medium, since email can fulfil so many communicative functions. It therefore exhibits a level of diversity surpassed perhaps only by spontaneous speech. Investigation of the spoken part of the BNC is therefore suggested as an area for further work.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5."
            },
            {
                "text": "To a certain extent, the apparent heterogeneity of the emall undermines the results of any similarity measures applied to this corpus. Nevertheless, the extent to which the email is unlike all the other BNC domains is quite apparent and therefore mitigates any unprincipled approaches to corpus augmentation using crude, top-down techniques that involve complete domains taken from the BNC. Consequently, the best way to acquire more ernail data appears to be either: (a) to instigate a further collection initiative, (b) to use more sophisticated bottom.up methods, or (c) to use self-organising adaptation techniques (e.g. Clarkson and Robinson, 1997) . The similarity metric used in (b) must be chosen carefully. Although the Loglikelihood and rank correlation metrics both produce results that can look intuitively plausible, this merely underlines the need for an objective, thorough evaluation method. Loglikefihood appears to be the more principled of the two measures, and it is suggested that this offers the greater potential.",
                "cite_spans": [
                    {
                        "start": 625,
                        "end": 653,
                        "text": "Clarkson and Robinson, 1997)",
                        "ref_id": "BIBREF5"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5."
            },
            {
                "text": "The results of the language modelling exercise provide clear evidence that it is possible to build effective LMs from small corpora. The email LM outperformed the other L_Ms on real spoken data (albeit taken from a technical, \"ernaiMike\" domain) for eight of the ten speakers. This is significant, considering the other LMs were trained on corpora that were several times larger. This effect can be mainly attributed to the source of the n-grams and the extent.to which the larger LMs \"waste\" probability mass on n-grams that never actually occur in the test data. Other researchers also have investigated methods for adapting large, general LMs using data from a small domain corpus and have found merit in simply building a smaller LM directly from the domain corpus. For example, Ueberla (1997) observed that the improvements gained by using adaptation techniques compared to simply \"starting from scratch\" on the domain data become quite small when several tens of thousands of words of domain data are available. (Since the email corpus is almost 2 million words it clearly meets this criterion.) It is interesting to note also that this threshold is seen to vary according to the level of similarity between the adaptation (domainspecific) corpus and the background (general) corpus.",
                "cite_spans": [
                    {
                        "start": 783,
                        "end": 797,
                        "text": "Ueberla (1997)",
                        "ref_id": "BIBREF23"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5."
            },
            {
                "text": "It is also possible to adapt LMs dynamically, using cache-based methods (e.g. Kuhn & de Mori 1990) and evidence suggests that this may prove the more effective approach (Matsunaga et al., 1992) . It is clear that ernail is highly heterogeneous and therefore inherently unpredictable. Attempting to model this by static means can thus produce only limited success. By contrast, a dynamic LM would adapt to the current input and update its probabilities accordingly. However, dynamic LMs still need a set of static, baseline probabifities, so the email LM may present the best starting point for this.",
                "cite_spans": [
                    {
                        "start": 78,
                        "end": 98,
                        "text": "Kuhn & de Mori 1990)",
                        "ref_id": "BIBREF14"
                    },
                    {
                        "start": 169,
                        "end": 193,
                        "text": "(Matsunaga et al., 1992)",
                        "ref_id": "BIBREF16"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusions",
                "sec_num": "5."
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF0": {
                "ref_id": "b0",
                "title": "Using register-diversified corpora for general language studies",
                "authors": [
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Biber",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "Computational Linguistics",
                "volume": "19",
                "issue": "2",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Biber, D. (1993) \"Using register-diversified corpora for general language studies\", Computational Linguistics, 19, No. 2.",
                "links": null
            },
            "BIBREF1": {
                "ref_id": "b1",
                "title": "A statistical approach to machine translation",
                "authors": [
                    {
                        "first": "P",
                        "middle": [],
                        "last": "Brown",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Cocke",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Della Pietra",
                        "suffix": ""
                    },
                    {
                        "first": "V",
                        "middle": [],
                        "last": "Della Pietra",
                        "suffix": ""
                    },
                    {
                        "first": "F",
                        "middle": [],
                        "last": "Jelinek",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Lafferty",
                        "suffix": ""
                    },
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Mercer",
                        "suffix": ""
                    },
                    {
                        "first": "P",
                        "middle": [],
                        "last": "",
                        "suffix": ""
                    }
                ],
                "year": 1989,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Brown, P., Cocke, J., Della Pietra, S., Della Pietra, V., Jelinek, F., Lafferty, J., Mercer, R., P. (1989) \"A statistical approach to machine translation\", Technical Report, IBM Research Division.",
                "links": null
            },
            "BIBREF2": {
                "ref_id": "b2",
                "title": "Users Reference Guide for the British National Corpus",
                "authors": [
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Burnard",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Burnard, L. (1995) Users Reference Guide for the British National Corpus, Oxford University Computing Services.",
                "links": null
            },
            "BIBREF3": {
                "ref_id": "b3",
                "title": "Statistical Language Learning",
                "authors": [
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Chamiak",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Chamiak, E. (1993) \"Statistical Language Learning\", M1T Press, Cambridge, Mass.",
                "links": null
            },
            "BIBREF4": {
                "ref_id": "b4",
                "title": "Using statistics in lexical analysis",
                "authors": [
                    {
                        "first": "K",
                        "middle": [],
                        "last": "Church",
                        "suffix": ""
                    },
                    {
                        "first": "P",
                        "middle": [],
                        "last": "Hanks",
                        "suffix": ""
                    },
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Hindle",
                        "suffix": ""
                    },
                    {
                        "first": "W",
                        "middle": [],
                        "last": "Gale",
                        "suffix": ""
                    }
                ],
                "year": 1991,
                "venue": "Lexical Acquisition: Using On-Line Resources to Build a Lexicon",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Church, K., Hanks, P., Hindle, D. and Gale, W. (1991) \"Using statistics in lexical analysis\", in Zernik, U. (Ed.) \"Lexical Acquisition: Using On-Line Resources to Build a Lexicon\", LEA, NJ.",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Language model adaptation using mixtures and an exponentially decaying cache",
                "authors": [
                    {
                        "first": "P",
                        "middle": [
                            "R"
                        ],
                        "last": "Clarkson",
                        "suffix": ""
                    },
                    {
                        "first": "A",
                        "middle": [
                            "J"
                        ],
                        "last": "Robinson",
                        "suffix": ""
                    }
                ],
                "year": 1997,
                "venue": "Proceedings of ICASSP",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Clarkson, P.R. & Robinson, AJ. (1997) \"Language model adaptation using mixtures and an exponentially decaying cache\", Proceedings of ICASSP, Munich, Germany.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "Combined approach for terminology extraction",
                "authors": [
                    {
                        "first": "B",
                        "middle": [],
                        "last": "Daille",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Daille, B. (1995) \"Combined approach for terminology extraction\", Technical Report 5, UCREL, Lancaster University.",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Accurate methods for the statistics of surprise and coincidence",
                "authors": [
                    {
                        "first": "E",
                        "middle": [],
                        "last": "Dunning",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "Computational Linguistics",
                "volume": "19",
                "issue": "1",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Dunning, E. (1993) \"Accurate methods for the statistics of surprise and coincidence\", Computational Linguistics, 19, No. 1.",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Large vocabulary continuous speech recognition using a hybrid connectionist HMM system",
                "authors": [
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Hochberg",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Robinson",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Renals",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "1499--1502",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hochberg, M., Robinson T. & Renals S. (1994) \"Large vocabulary continuous speech recognition using a hybrid connectionist HMM system\", Prec. of ICSLP, pp. 1499-1502.",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "Self-organized language modeling for speech recognition",
                "authors": [
                    {
                        "first": "F",
                        "middle": [],
                        "last": "Jelinek",
                        "suffix": ""
                    }
                ],
                "year": 1990,
                "venue": "Readings in Speech Recognition",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jelinek, F. (1990) \"Self-organized language modeling for speech recognition\", in Waibel and Lee (Eds.), Readings in Speech Recognition, Morgan Kaufmann, San Mateo, CA.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Video mail retrieval using voice",
                "authors": [
                    {
                        "first": "G",
                        "middle": [],
                        "last": "Jones",
                        "suffix": ""
                    },
                    {
                        "first": "J",
                        "middle": [],
                        "last": "Foote",
                        "suffix": ""
                    },
                    {
                        "first": "K",
                        "middle": [],
                        "last": "Sparck Jones",
                        "suffix": ""
                    },
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Young",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Jones, G., Foote, J., Sparck Jones, K. & Young, S. (1994) \"Video mail retrieval using voice\", Technical Report 335, Cambndge University Computer Laboratory.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "Estimation of probabilities from sparse data",
                "authors": [
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Katz",
                        "suffix": ""
                    }
                ],
                "year": 1987,
                "venue": "lEE Transactions on Acoustics, Speech & Signal Processing",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Katz, S. (1987) \"Estimation of probabilities from sparse data\", lEE Transactions on Acoustics, Speech & Signal Processing, vol. ASSP-35.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "Which words are particularly characteristic of a text",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Kilgarriff",
                        "suffix": ""
                    }
                ],
                "year": 1996,
                "venue": "Language Engineering for Document Analysis & Recognition, AISB Workshop proceedings",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kilgarriff, A. (1996) \"Which words are particularly characteristic of a text?\" in Rose & Evett (Eds.) Language Engineering for Document Analysis & Recognition, AISB Workshop proceedings.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Using word frequency lists to measure corpus homogeneity and similarity between corpora",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Kilgurriff",
                        "suffix": ""
                    }
                ],
                "year": 1997,
                "venue": "Proceedings of the Fifth Workshop on Very Large Corpora",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kilgurriff, A (1997) \"Using word frequency lists to measure corpus homogeneity and similarity between corpora\", Proceedings of the Fifth Workshop on Very Large Corpora, Hong Kong.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "A cache-based natural language model for speech recognition",
                "authors": [
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Kuhn",
                        "suffix": ""
                    },
                    {
                        "first": "R",
                        "middle": [],
                        "last": "De Mori",
                        "suffix": ""
                    }
                ],
                "year": 1990,
                "venue": "IEEE Trans. on PAMI",
                "volume": "12",
                "issue": "6",
                "pages": "570--583",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Kuhn, R. & De Mori, R. (1990) \"A cache-based natural language model for speech recognition\" IEEE Trans. on PAMI, 12(6), pp. 570--583.",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Text representation for intelligent text retrieval: a classification-oriented view",
                "authors": [
                    {
                        "first": "D",
                        "middle": [],
                        "last": "Lewis",
                        "suffix": ""
                    }
                ],
                "year": 1992,
                "venue": "P. Jaeobs ''Text-based Intelligent Systems",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Lewis, D. (1992) \"Text representation for intelligent text retrieval: a classification-oriented view\", in P. Jaeobs ''Text-based Intelligent Systems\", LEA Publishers, Hfllsdale, NJ.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "Task adaptation in stochastic language models for continuous speech recognition",
                "authors": [
                    {
                        "first": "S",
                        "middle": [],
                        "last": "Matsunaga",
                        "suffix": ""
                    },
                    {
                        "first": "T",
                        "middle": [],
                        "last": "Yamada",
                        "suffix": ""
                    },
                    {
                        "first": "K",
                        "middle": [],
                        "last": "Shikano",
                        "suffix": ""
                    }
                ],
                "year": 1992,
                "venue": "Proc. ICASSP",
                "volume": "1",
                "issue": "",
                "pages": "165--168",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Matsunaga, S., Yamada, T. & Shikano, K. (1992) \"Task adaptation in stochastic language models for continuous speech recognition\", Proc. ICASSP Vol. 1, pp. 165-168.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "The use of context in cursive script recognition",
                "authors": [
                    {
                        "first": "T",
                        "middle": [
                            "G"
                        ],
                        "last": "Rose",
                        "suffix": ""
                    },
                    {
                        "first": "L",
                        "middle": [],
                        "last": "Evett",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "Machine Vision and Applications",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rose, T.G. & Evett, L. (1995) \"The use of context in cursive script recognition\", Machine Vision and Applications\", Springer International.",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "Language modelling for large vocabulary speech recognition",
                "authors": [
                    {
                        "first": "T",
                        "middle": [
                            "G"
                        ],
                        "last": "Rose",
                        "suffix": ""
                    },
                    {
                        "first": "M",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "Proc. IOA Meeting on LVSR",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rose, T.G. & Lee, M (1994) \"Language modelling for large vocabulary speech recognition\", Proc. IOA Meeting on LVSR, Cambridge, England.",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "A similarity-based agent for Intemet searching",
                "authors": [
                    {
                        "first": "T",
                        "middle": [
                            "G"
                        ],
                        "last": "Rose",
                        "suffix": ""
                    },
                    {
                        "first": "P",
                        "middle": [
                            "J"
                        ],
                        "last": "Wyard",
                        "suffix": ""
                    }
                ],
                "year": 1997,
                "venue": "Proceedings of RIAO'97 -Computer-assisted Searching on the Interact",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rose, T.G. & Wyard, PJ. (1997) \"A similarity-based agent for Intemet searching\", Proceedings of RIAO'97 -Computer-assisted Searching on the Interact, Montreal, Canada.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "The CMU Statistical Language Modeling Toolkit and its use in the 1994 ARPA CSR Evaluation",
                "authors": [
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Rosenfeld",
                        "suffix": ""
                    }
                ],
                "year": 1994,
                "venue": "Proceedings of the Spoken Language Technology Workshop",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rosenfeld, R. (1994) \"The CMU Statistical Language Modeling Toolkit and its use in the 1994 ARPA CSR Evaluation, Proceedings of the Spoken Language Technology Workshop 1995, Austin ('IX).",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "Opdrnizing lexical and n-gram coverage via judicious use of linguistic data",
                "authors": [
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Rosenfeld",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "Proc. Eurospeeeh 95",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rosenfeld, R. (1995) \"Opdrnizing lexical and n-gram coverage via judicious use of linguistic data\", Proc. Eurospeeeh 95.",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "Language modeling with limited domain data",
                "authors": [
                    {
                        "first": "A",
                        "middle": [],
                        "last": "Rudnicky",
                        "suffix": ""
                    }
                ],
                "year": 1995,
                "venue": "Proceedings of the ARPA Workshop on Spoken Language Technology",
                "volume": "",
                "issue": "",
                "pages": "66--69",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Rudnicky, A. (1995) \"Language modeling with limited domain data\", Proceedings of the ARPA Workshop on Spoken Language Technology, Morgan Kaufmann, San Mateo, pp. 66-69.",
                "links": null
            },
            "BIBREF23": {
                "ref_id": "b23",
                "title": "Domain adaptation with clustered language models",
                "authors": [
                    {
                        "first": "J",
                        "middle": [
                            "P"
                        ],
                        "last": "Ueberla",
                        "suffix": ""
                    }
                ],
                "year": 1997,
                "venue": "Proceedings of ICASSP",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Ueberla, J.P. (1997) \"Domain adaptation with clustered language models\", Proceedings of ICASSP, Munich, Germany.",
                "links": null
            },
            "BIBREF25": {
                "ref_id": "b25",
                "title": "HTK: Hidden Markov Model Toolkit V1.5 User Manual",
                "authors": [
                    {
                        "first": "Young",
                        "middle": [
                            "S"
                        ],
                        "last": "Woodland",
                        "suffix": ""
                    },
                    {
                        "first": "P",
                        "middle": [],
                        "last": "",
                        "suffix": ""
                    }
                ],
                "year": 1993,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Young S. & Woodland P. (1993) \"HTK: Hidden Markov Model Toolkit V1.5 User Manual\", Cambridge University Engineering Dept. and Entropic Research Labs Ltd.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "text": "This work was completed at HP Labs during a previous appoimmcnt funded by the Royal Academy of Engineering.",
                "uris": null,
                "type_str": "figure",
                "num": null
            },
            "FIGREF2": {
                "text": "Create a wfl for the email corpus. 2. For each individual text in the BNC, do: 2.1 Create the wfl for the BNC text 2.2 Create a contingency table from the 2 wfls (ignoring function words) 2.3 Calculate the number of common words N and the rank correlation r 3. Store the filename, fltle, N and r in RESULTS 4. Output the RESULTS sorted on the value for r. Although the rank correlation may be applied to the wfls regardless of their content, it was found empirically that performance improved if function words were excluded from the contingency table. A stop list of 241 function words was therefore applied in Step 2.2 of the above algorithm. The algorithm was run on the entire BNC (i.e. each of its 4,000+ files). The output was a list of the files sorted according to the value of r. The top and bottom 10 texts on this fist are as follows:",
                "uris": null,
                "type_str": "figure",
                "num": null
            },
            "FIGREF3": {
                "text": "Each line shows the filename, the title of the text, the number of common words and the value for r. At first glance, the results appear to be intuitively satisfying. Of the top ten texts, six have titles that are clearly related to computing, including all of the top five. The remaining four could arguably be classified as Commerce & Finance (which was identified as the second most similar domain to email). However, a suitable tide is no guarantee of suitable contents. As far as can reasonably be expected, the tides constitute a fair and accurate reflection of the contents of each text. Of course, the whole point of this \"1 approach is to develop techniques that do not rely on ambiguous manual annotations such as title or domain, so the presence of suitable floes is merely an initial indication of success.",
                "uris": null,
                "type_str": "figure",
                "num": null
            },
            "FIGREF4": {
                "text": "111",
                "uris": null,
                "type_str": "figure",
                "num": null
            },
            "TABREF0": {
                "type_str": "table",
                "num": null,
                "text": "Similarity and homogenei W of BNC domains and emailTable 1shows both sets of results. The homogeneity values are across the diagonal, with mean and standard deviation shown in each cell. The other cells show the rank correlation (r) and the value of N. \"BNC\" refers to the complete corpus. The subdomains are labelled as follows:",
                "content": "<table><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>I</td></tr><tr><td>Ar</td><td>Ar 0.789</td><td>As</td><td>Bt</td><td>Cf</td><td>Im</td><td>Le</td><td>Np</td><td>Ss</td><td colspan=\"2\">Un : Wa</td><td>BNC</td><td>Email</td><td>!</td></tr><tr><td/><td>0.001</td><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/></tr><tr><td>As</td><td colspan=\"2\">0.2.55. 0.758 72127 : 0.001</td><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>!</td></tr><tr><td>Bt Cf</td><td colspan=\"2\">0.408 69932. 73839 0.238 0.340 : 0.351</td><td>0.581 0.002 0.291</td><td>0.730</td><td/><td/><td/><td/><td/><td/><td/><td/><td>1</td></tr><tr><td/><td>70648</td><td>70452</td><td>72970</td><td>0.001</td><td/><td/><td/><td/><td/><td/><td/><td/></tr><tr><td>Im</td><td>0.404</td><td>0.159</td><td>0.340</td><td>0.215</td><td>' 0.887</td><td/><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>67604</td><td>73786</td><td>!70807</td><td>72800</td><td>0.001</td><td/><td/><td/><td/><td/><td/><td/></tr><tr><td>Le</td><td>0.459</td><td>0.284</td><td>0.310</td><td>0.337</td><td colspan=\"2\">0.407 . 0.824</td><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>67013</td><td>71281</td><td>71768</td><td>70407</td><td>67615</td><td>0.001</td><td/><td/><td/><td/><td/><td/></tr><tr><td>Np</td><td>0.122</td><td>0.315</td><td>0.150</td><td>0.161</td><td>0.061</td><td>0.145</td><td>0.662</td><td/><td/><td/><td/><td/></tr><tr><td/><td colspan=\"2\">75822. 71596</td><td>76151</td><td>75339</td><td>\u2022 76628</td><td>74952</td><td>0.002</td><td/><td/><td/><td/><td/></tr><tr><td>Ss</td><td>0.409</td><td>0.342</td><td>0.422</td><td>0.470</td><td>0.278</td><td>0.325</td><td>0.200</td><td>0.812</td><td/><td/><td/><td/></tr><tr><td/><td>68263</td><td>70405</td><td>69756</td><td>68023</td><td>70475</td><td>69796</td><td>74053</td><td>0.001</td><td/><td/><td/><td/></tr><tr><td>Un</td><td>0.273</td><td>0.161</td><td>0.174</td><td>0.229</td><td>\u2022 0.244</td><td>0.357</td><td>0.032</td><td>0.226</td><td>0.486</td><td/><td/><td/></tr><tr><td/><td colspan=\"2\">74i 10 767.54</td><td>77136</td><td>75423</td><td>. 74635</td><td>72373</td><td>79904</td><td>75161</td><td>0.002</td><td/><td/><td/></tr><tr><td>Wa</td><td>0.423</td><td>0.280</td><td>0.395</td><td>0.432</td><td>0.311</td><td>0.395</td><td>0.130</td><td>0.469</td><td>0.284</td><td>0.865</td><td/><td/></tr><tr><td/><td colspan=\"2\">68005\" 71298</td><td>70189</td><td>68404</td><td>69838</td><td>68296</td><td>75399</td><td>66921</td><td>73981</td><td>0.001</td><td/><td/></tr><tr><td>BNC</td><td>0.611</td><td>0.497</td><td>0.505</td><td>0.541</td><td>0.578</td><td>0.605</td><td>0.307</td><td>0.609</td><td>0.381</td><td>0.653</td><td>0.687 -</td><td/></tr><tr><td/><td>63522</td><td>66732</td><td>67920</td><td>66189</td><td colspan=\"2\">63755 \u2022 63439</td><td>71615</td><td>63779</td><td>72033</td><td>62450</td><td>0.001 \u2022</td><td/></tr><tr><td>Email</td><td>0.012</td><td>0.093</td><td>-0.026</td><td>\u2022 0.055</td><td>-0.062</td><td>-0.003</td><td>-0.032</td><td>0.035</td><td>-0.066</td><td>-0.012</td><td>0.073 ;</td><td>0.362</td></tr><tr><td/><td colspan=\"2\">81648 \" 79745</td><td>: 82897</td><td>80884</td><td colspan=\"2\">83083 ~ 81908</td><td>82719</td><td>81143</td><td>84338 \"</td><td>82015</td><td>80085</td><td>0.002</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>i</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>l</td></tr><tr><td>As: Ar:</td><td colspan=\"3\">Applied Science Arts</td><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>|</td></tr><tr><td>Be:</td><td colspan=\"3\">Beliefs &amp; thought</td><td/><td/><td/><td/><td/><td/><td/><td/><td/></tr><tr><td>Cf:</td><td colspan=\"3\">Commerce &amp; Finance</td><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>I</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>!</td></tr></table>",
                "html": null
            }
        }
    }
}