_id stringlengths 4 10 | text stringlengths 0 18.4k | title stringlengths 0 8.56k |
|---|---|---|
d220047793 | We present an efficient annotation framework for argument quality, a feature difficult to be measured reliably as per previous work. A stochastic transitivity model is combined with an effective sampling strategy to infer highquality labels with low effort from crowdsourced pairwise judgments. The model's capabilities are showcased by compiling Webis-ArgQuality-20, an argument quality corpus that comprises scores for rhetorical, logical, dialectical, and overall quality inferred from a total of 41,859 pairwise judgments among 1,271 arguments. With up to 93% cost savings, our approach significantly outperforms existing annotation procedures. Furthermore, novel insight into argument quality is provided through statistical analysis, and a new aggregation method to infer overall quality from individual quality dimensions is proposed. | Efficient Pairwise Annotation of Argument Quality |
d5921877 | ||
d59659044 | This paper describes the parsing scheme in the <$Dm Dia l OG speech-to-speech dialog translation system, with special emphasis on the integration of speech and natural language processing. We propose an integrated architec ture for parsing speech inputs based on a parallel marker-passing scheme and attaining dynamic participation of knowledge from the phonological-level to the discourse-level. At the phonological level, we employ a stochastic model using a transition matrix and a confusion matrix and markers which carry a probability measure. At a higher level, syntactic/semantic and discourse processing, we integrate a case-based and constraint-based scheme in a consistent manner so that a priori probability and constraints, which reflect linguistic and discourse factors, are provided to the phonological level of processing. A probability/cost-based scheme in our model enables ambiguity resolution at various levels using one uniform principle. | Massively Parallel Parsing in ^D m D ia lo g : Integrated Architecture for Parsing Speech Inputs |
d15109999 | Nowadays, most of the statistical translation systems are based on phrases (i.e. groups of words). We describe a phrase-based system using a modified method for the phrase extraction which deals with larger phrases while keeping a reasonable number of phrases. Also, different alignments to extract phrases are allowed and additional features are used which lead to a clear improvement in the performance of translation. Finally, the system manages to do reordering. We report results in terms of translation accuracy by using the BTEC corpus in the tasks of Chinese to English and Arabic to English, in the framework of IWSLT'05 evaluation. | Tuning a phrase-based statistical translation system for the IWSLT 2005 Chinese to English and Arabic to English tasks |
d11653683 | Semantic Role Labeling (SRL) plays an important role in different text mining tasks. The development of SRL systems for the biomedical area is frustrated by the lack of large-scale domain specific corpora that are annotated with semantic roles. In our previous work, we proposed a method for building FramenNet-like corpus for the area using domain knowledge provided by ontologies. In this paper, we present a framework for supporting the method and the system which we developed based on the framework. In the system we have developed the algorithms for selecting appropriate concepts to be translated into semantic frames, for capturing the information that describes frames from ontology terms, and for collecting example sentence using ontological knowledge. | A System for Building FrameNet-like Corpus for the Biomedical Domain |
d1786876 | This paper addressesthe problem Of | CASE ROLE FILLING AS A SIDE EFFECT OF VISUAL SEARCH |
d259360414 | Understanding interpersonal communication requires, in part, understanding the social context and norms in which a message is said. However, current methods for identifying offensive content in such communication largely operate independent of context, with only a few approaches considering community norms or prior conversation as context. Here, we introduce a new approach to identifying inappropriate communication by explicitly modeling the social relationship between the individuals. We introduce a new dataset of contextually-situated judgments of appropriateness and show that large language models can readily incorporate relationship information to accurately identify appropriateness in a given context. Using data from online conversations and movie dialogues, we provide insight into how the relationships themselves function as implicit norms and quantify the degree to which context-sensitivity is needed in different conversation settings. Further, we also demonstrate that contextualappropriateness judgments are predictive of other social factors expressed in language such as condescension and politeness. | Your spouse needs professional help: Determining the Contextual Appropriateness of Messages through Modeling Social Relationships |
d1197999 | The aim of this paper is to present a computational model of the dynamic composition and update of verb argument expectations using Distributional Memory, a state-of-the-art framework for distributional semantics. The experimental results conducted on psycholinguistic data sets show that the model is able to successfully predict the changes on the patient argument thematic fit produced by different types of verb agents. | Composing and Updating Verb Argument Expectations: A Distributional Semantic Model |
d38000118 | ||
d17468996 | We describe our participation in the MTPIL Hindi Parsing Shared Task-2012. Our system achieved the following results: 82.44% LAS/90.91% UAS (auto) and 85.31% LAS/92.88% UAS (gold). Our parser is based on the linear classification, which is suboptimal as far as the accuracy is concerned. The strong point of our approach is its speed. For parsing development the system requires 0.935 seconds, which corresponds to a parsing speed of 1318 sentences per second. The Hindi Treebank contains much less different part of speech tags than many other treebanks and therefore it was absolutely necessary to use the additional morphosyntactic features available in the treebank. We were able to build classifiers predicting those, using only the standard word form and part of speech features, with a high accuracy. | Parsing Hindi with MDParser |
d8400806 | This paper describes several novel hybrid semantic similarity measures. We study various combinations of 16 baseline measures based on WordNet, Web as a corpus, corpora, dictionaries, and encyclopedia. The hybrid measures rely on 8 combination methods and 3 measure selection techniques and are evaluated on (a) the task of predicting semantic similarity scores and (b) the task of predicting semantic relation between two terms. Our results show that hybrid measures outperform single measures by a wide margin, achieving a correlation up to 0.890 and MAP(20) up to 0.995. | A Study of Hybrid Similarity Measures for Semantic Relation Extraction |
d10183660 | Recently, users who search on the web are targeting to more complex tasks due to the explosive growth of web usage. To accomplish a complex task, users may need to obtain information of various entities. For example, a user who wants to travel to Beijing, should book a flight, reserve a hotel room, and survey a Beijing map. A complex task thus needs to submit several queries in order to seeking each of entities. Understanding complex tasks can allow a search engine to suggest related entities and help users explicitly assign their ongoing tasks. | Identifying Real-Life Complex Task Names with Task-Intrinsic Enti- ties from Microblogs |
d202707711 | ||
d219303928 | ||
d215824540 | Interpersonal relations are fickle, with close friendships often dissolving into enmity. In this work, we explore linguistic cues that presage such transitions by studying dyadic interactions in an online strategy game where players form alliances and break those alliances through betrayal. We characterize friendships that are unlikely to last and examine temporal patterns that foretell betrayal.We reveal that subtle signs of imminent betrayal are encoded in the conversational patterns of the dyad, even if the victim is not aware of the relationship's fate. In particular, we find that lasting friendships exhibit a form of balance that manifests itself through language. In contrast, sudden changes in the balance of certain conversational attributes-such as positive sentiment, politeness, or focus on future planning-signal impending betrayal. | Linguistic Harbingers of Betrayal: A Case Study on an Online Strategy Game |
d21709466 | In several areas of NLP evaluation, test suites have been used to analyze the strengths and weaknesses of systems. Today, Machine Translation (MT) quality is usually assessed by shallow automatic comparisons of MT outputs with reference corpora resulting in a number. Especially the trend towards neural MT has renewed peoples' interest in better and more analytical diagnostic methods for MT quality. In this paper we present TQ-AutoTest, a novel framework that supports a linguistic evaluation of (machine) translations using test suites. Our current test suites comprise about 5000 handcrafted test items for the language pair German-English. The framework supports the creation of tests and the semi-automatic evaluation of the MT results using regular expressions. The expressions help to classify the results as correct, incorrect or as requiring a manual check. The approach can easily be extended to other NLP tasks where test suites can be used such as evaluating (one-shot) dialogue systems. | TQ-AutoTest -A Semi-Automatic Test Suite for (Machine) Translation Quality |
d12269372 | We extendZhao and Ng's (2007)Chinese anaphoric zero pronoun resolver by (1) using a richer set of features and (2) exploiting the coreference links between zero pronouns during resolution. Results on OntoNotes show that our approach significantly outperforms two state-of-the-art anaphoric zero pronoun resolvers. To our knowledge, this is the first work to report results obtained by an end-toend Chinese zero pronoun resolver. | Chinese Zero Pronoun Resolution: Some Recent Advances |
d14984736 | We introduce TRAAM, or Transduction RAAM, a fully bilingual generalization ofPollack's (1990)monolingual Recursive Auto-Associative Memory neural network model, in which each distributed vector represents a bilingual constituent-i.e., an instance of a transduction rule, which specifies a relation between two monolingual constituents and how their subconstituents should be permuted. Bilingual terminals are special cases of bilingual constituents, where a vector represents either (1) a bilingual token -a token-totoken or "word-to-word" translation rule -or (2) a bilingual segment-a segmentto-segment or "phrase-to-phrase" translation rule. TRAAMs have properties that appear attractive for bilingual grammar induction and statistical machine translation applications. Training of TRAAM drives both the autoencoder weights and the vector representations to evolve, such that similar bilingual constituents tend to have more similar vectors. | Transduction Recursive Auto-Associative Memory: Learning Bilingual Compositional Distributed Vector Representations of Inversion Transduction Grammars |
d258865270 | In this paper, we introduce Ranger -a toolkit to simplify the utilization of effect-size-based meta-analysis for multi-task evaluation in NLP and IR. We observed that our communities often face the challenge of aggregating results over incomparable metrics and scenarios, which makes conclusions and take-away messages less reliable. With Ranger, we aim to address this issue by providing a task-agnostic toolkit that combines the effect of a treatment on multiple tasks into one statistical evaluation, allowing for comparison of metrics and computation of an overall summary effect. Our toolkit produces publication-ready forest plots that enable clear communication of evaluation results over multiple tasks. Our goal with the ready-to-use Ranger toolkit is to promote robust, effect-size based evaluation and improve evaluation standards in the community. We provide two case studies for common IR and NLP settings to highlight Ranger's benefits. | Ranger: A Toolkit for Effect-Size Based Multi-Task Evaluation |
d238744341 | Dominant sentence ordering models can be classified into pairwise ordering models and set-to-sequence models. However, there is little attempt to combine these two types of models, which inituitively possess complementary advantages. In this paper, we propose a novel sentence ordering framework which introduces two classifiers to make better use of pairwise orderings for graph-based sentence ordering(Yin et al., 2019(Yin et al., , 2021. Specially, given an initial sentence-entity graph, we first introduce a graph-based classifier to predict pairwise orderings between linked sentences. Then, in an iterative manner, based on the graph updated by previously predicted highconfident pairwise orderings, another classifier is used to predict the remaining uncertain pairwise orderings. At last, we adapt a GRN-based sentence ordering model(Yin et al., 2019(Yin et al., , 2021on the basis of final graph. Experiments on five commonly-used datasets demonstrate the effectiveness and generality of our model. Particularly, when equipped with BERT(Devlin et al., 2019) andFHDecoder (Yin et al., 2020), our model achieves state-of-the-art performance. Our code is available at https:// github.com/DeepLearnXMU/IRSEG. | Improving Graph-based Sentence Ordering with Iteratively Predicted Pairwise Orderings |
d232021676 | ||
d10937448 | The paper shows how the verbal lexicon can be formalised in a way that captures and exploits generalisations about the alternation behaviour of verb classes. An alternation is a pattern in which a number of words share the same relationship between • a pair of senses. The alternations captured are ones where the different senses specify different relationships between syntactic complements and semantic arguments, as between bake in "John is baking the cake"and "The cake is baking". The formal language used is DATR. The lexical entries it builds are as specified in HPSG. The complex alternation behaviour shared between families of verbs is elegantly represented in a way that makes generalisations explicit, avoids redundancy, and offers practical benefits to computational lexicographers. | Inheriting Verb Alternations |
d14247149 | We introduce PolArt, a robust tool for sentiment analysis. PolArt is a pattern-based approach designed to cope with polarity composition. In order to determine the polarity of larger text units, a cascade of rewrite operations is carried out: word polarities are combined to NP, VP and sentence polarities. Moreover, PolArt is able to cope with the target-specific polarity of phrases, where two neutral words combine to a non-neutral phrase. Target detection is done with the Wikipedia category system, but also user defined target hierarchies are allowed. PolArt is based on the TreeTagger chunker output, and is customised for English and German. In this paper we evaluate PolArt's compositional capacity. | PolArt: A Robust Tool for Sentiment Analysis |
d235650866 | The famous "laurel/yanny" phenomenon references an audio clip that elicits dramatically different responses from different listeners. For the original clip, roughly half the population hears the word "laurel," while the other half hears "yanny." How common are such "polyperceivable" audio clips? In this paper we apply ML techniques to study the prevalence of polyperceivability in spoken language. We devise a metric that correlates with polyperceivability of audio clips, use it to efficiently find new "laurel/yanny"-type examples, and validate these results with human experiments. Our results suggest that polyperceivable examples are surprisingly prevalent, existing for >2% of English words. 1 | Beyond Laurel/Yanny: An Autoencoder-Enabled Search for Polyperceivable Audio |
d11485057 | In this study, we examine the effects of using a game for encouraging the use of a spoken dialogue system. As a case study, we developed a word-chain game, called Shiritori in Japanese, and released the game as a module in a Japanese Android/iOS app, Onsei-Assist, which is a Siri-like personal assistant based on a spoken dialogue technology. We analyzed the log after the release and confirmed that the game can increase the number of user utterances. Furthermore, we discovered a positive side effect, in which users who have played the game tend to begin using non-game modules. This suggests that just adding a game module to the system can improve user engagement with an assistant agent. | Effects of Game on User Engagement with Spoken Dialogue System |
d227231791 | ||
d13509553 | In recent years, there has been an increasing amount of literature on query classification. Click-through information has been shown to be a useful source for improving this task. However, far too little attention has been paid to queries in very specific domains such as art, culture and history. We propose an approach that exploits topic models built from a domain specific corpus as a mean to enrich both the query and the categories against which the query need to be classified. We take an Art Library as the case study and show that topic model enrichment improves over the enrichment via click-through considerably. | Query classification via Topic Models for an art image archive |
d44494239 | PROJECT GOALSThe Learning Systems Department at Siemens Corporate Research is investigating the use of concept spaces to increase retrieval effectiveness. Similar to a semantic net, a concept space is a construct that defines the semantic relationships among ideas. The current focus of our research is to exploit the information in such a structure to ameliorate known shortcomings of statistical retrieval methods while maintaining the statistical methods' robustness. Our initial concept space is extracted from WordNet, a manually-constructed lexical database developed at Princeton University. | EXPLOITING CONCEPT SPACES FOR TEXT RETRIEVAL |
d261342046 | ||
d14928811 | In this paper we present a new biologically inspired approach to the part-of-speech tagging problem, based on particle swarm optimization. As far as we know this is the first attempt of solving this problem using swarm intelligence. We divided the part-of-speech problem into two subproblems. The first concerns the way of automatically extracting disambiguation rules from an annotated corpus. The second is related with how to apply these rules to perform the automatic tagging. We tackled both problems with particle swarm optimization. We tested our approach using two different corpora of English language and also a Portuguese corpus. The accuracy obtained on both languages is comparable to the best results previously published, including other evolutionary approaches. | BioPOS: Biologically Inspired Algorithms for POS Tagging Ana Paul a Sil va 1 Ar l ind o Sil va 1 I r eneRod r i gues |
d10474066 | Collocational deficiency is a pervasive phenomenon in learner English. Language learners often fail to choose the correct combination of two or more words due to their unawareness of collocational properties in vocabulary. They are apt to adopt lexical simplification strategies such as using a synonymous or Ll-influenced expression. This paper presents a corpus-based study on the collocational deficiency of Taiwanese learners of English. The work utilizes two pre-tagged corpora, Taiwanese Learner Corpus of English and British National Corpus, to examine the learner's use of collocations over a set of synonymous words: big, large, great.The experimental findings indicate that among the three words the collocations with big are significantly overused by the learners when it is used to refer to abstract concepts. This overuse phenomenon is further investigated and it is found that the collocations of high frequency in the learner English tend to be used to express vague ideas when more specific meanings should be conveyed. It is also found that the learners are apt to apply those collocations to the cases where more concise expressions are preferred. Another finding shows that problematic collocations, pertaining to big, large and great, are produced as the result of learner's application of the Ll-transfer and synonym strategies, which the Taiwanese learners commonly adopt for lexical simplification. | Collocation Deficiency in a Learner Corpus of English: from an overuse perspective |
d6980946 | FBK participated in the WMT 2010 Machine Translation shared task with phrase-based Statistical Machine Translation systems based on the Moses decoder for English-German and German-English translation. Our work concentrates on exploiting the available language modelling resources by using linear mixtures of large 6-gram language models and on addressing linguistic differences between English and German with methods based on word lattices. In particular, we use lattices to integrate a morphological analyser for German into our system, and we present some initial work on rule-based word reordering. | FBK at WMT 2010: Word Lattices for Morphological Reduction and Chunk-based Reordering |
d7830082 | It is widely accepted that tagging text with semantic information would improve the quality of lexical learning in corpus-based NLP methods. However available on-line taxonomies are rather entangled and introduce an unnecessary level of ambiguity. The noise produced by the redundant number of tags often overrides the advantage of semantic tagging. In this paper we propose an automatic method to select from WordNet a subset of domain-appropriate categories that effectively reduce the overambiguity of WordNet, and help at identifying and categorise relevant language patterns in a more compact way. The method is evaluated against a manually tagged corpus, SEMCOR. software used in this experiment, as well as all our partners in the ECRAN project. | Automatic Selection of Class Labels from a Thesaurus for an Effective Semantic Tagging of Corpora |
d245856034 | This paper describes Naver Papago's submission to the WMT21 shared triangular MT task to enhance the non-English MT system with tri-language parallel data. The provided parallel data are Russian-Chinese (direct), Russian-English (indirect), and English-Chinese (indirect) data. This task aims to improve the quality of the Russian-to-Chinese MT system by exploiting the direct and indirect parallel resources. The direct parallel data is noisy data crawled from the web. To alleviate the issue, we conduct extensive experiments to find effective data filtering methods. With the empirical knowledge that the performance of bilingual MT is better than multi-lingual MT and related experiment results, we approach this task as bilingual MT, where the two indirect data are transformed to direct data. In addition, we use the Transformer, a robust translation model, as our baseline and integrate several techniques, averaging checkpoints, model ensemble, and re-ranking. Our final system provides a 12.7 BLEU points improvement over a baseline system on the WMT21 triangular MT development set. In the official evaluation of the test set, ours is ranked 2nd in terms of BLEU scores. | Papago's Submissions to the WMT21 triangular Translation Task |
d14184076 | Distinguishing between antonyms and synonyms is a key task to achieve high performance in NLP systems. While they are notoriously difficult to distinguish by distributional co-occurrence models, pattern-based methods have proven effective to differentiate between the relations. In this paper, we present a novel neural network model AntSynNET that exploits lexico-syntactic patterns from syntactic parse trees. In addition to the lexical and syntactic information, we successfully integrate the distance between the related words along the syntactic path as a new pattern feature. The results from classification experiments show that AntSyn-NET improves the performance over prior pattern-based methods. | Distinguishing Antonyms and Synonyms in a Pattern-based Neural Network |
d241583451 | Automatic readability assessment (ARA) is the task of automatically assessing readability with little or no human supervision. ARA is essential for many second language acquisition applications to reduce the workload of annotators, who are usually language teachers. Previous unsupervised approaches manually searched textual features that correlated well with readability labels, such as perplexity scores of large language models. This paper argues that, to evaluate an assessors' performance, rank-correlation coefficients should be used instead of Pearson's correlation coefficient (ρ). In the experiments, we show that its performance can be easily underestimated using Pearson's ρ, which is significantly affected by the linearity of the output readability scores. We also propose a lightweight unsupervised readability assessor that achieved the best performance in both the rank correlations and Pearson's ρ among all unsupervised assessors compared. | Evaluation of Unsupervised Automatic Readability Assessors Using Rank Correlations |
d8233374 | We present the first parser for UCCA, a cross-linguistically applicable framework for semantic representation, which builds on extensive typological work and supports rapid annotation. UCCA poses a challenge for existing parsing techniques, as it exhibits reentrancy (resulting in DAG structures), discontinuous structures and non-terminal nodes corresponding to complex semantic units. To our knowledge, the conjunction of these formal properties is not supported by any existing parser. Our transition-based parser, which uses a novel transition set and features based on bidirectional LSTMs, has value not just for UCCA parsing: its ability to handle more general graph structures can inform the development of parsers for other semantic DAG structures, and in languages that frequently use discontinuous structures. | A Transition-Based Directed Acyclic Graph Parser for UCCA |
d13976138 | The extraction of protein-protein interactions (PPIs) reported in scientific publications is one of the most studied topics in Text Mining in the Life Sciences, as such algorithms can substantially decrease the effort for databases curators. The currently best methods for this task are based on analyzing the dependency tree (DT) representation of sentences. Many approaches exploit only topological features and thus do not yet fully exploit the information contained in DTs. We show that incorporating the grammatical information encoded in the types of the dependencies in DTs noticeably improves extraction performance by using a pattern matching approach. We automatically infer a large set of linguistic patterns using only information about interacting proteins. Patterns are then refined based on shallow linguistic features and the semantics of dependency types. Together, these lead to a total improvement of 17.2 percent points in F 1 , as evaluated on five publicly available PPI corpora. More than half of that improvement is gained by properly handling dependency types. Our method provides a general framework for building task-specific relationship extraction methods that do not require annotated training data. Furthermore, our observations offer methods to improve upon relation extraction approaches. | Not all links are equal: Exploiting Dependency Types for the Extraction of Protein-Protein Interactions from Text |
d259370517 | The task of web information extraction is to extract target fields of an object from web pages, such as extracting the name, genre and actor from a movie page. Recent sequential modeling approaches have achieved state-of-the-art results on web information extraction. However, most of these methods only focus on extracting information from textual sources while ignoring the rich information from other modalities such as image and web layout. In this work, we propose a novel MUltimodal Structural Transformer (MUST) that incorporates multiple modalities for web information extraction. Concretely, we develop a structural encoder that jointly encodes the multimodal information based on the HTML structure of the web layout, where high-level DOM nodes, low-level text, and image tokens are introduced to represent the entire page. Structural attention patterns are designed to learn effective cross-modal embeddings for all DOM nodes and low-level tokens. An extensive set of experiments has been conducted on WebSRC and Common Crawl benchmarks. Experimental results demonstrate the superior performance of MUST over several state-of-the-art baselines. | MUSTIE: Multimodal Structural Transformer for Web Information Extraction |
d16427181 | 摘要 本研究採用機器學習法對語音情緒辨識進行探討。除一般常被採用之語音特徵， 如音高、共振峰、能量以及梅爾倒頻譜係數之外，研究中加入了夏農熵和曲率指 標(curvature index)[9]兩項非線性特徵，再利用費雪鑑別比與基因演算法搭配的方 式進行特徵挑選。最後使用支持向量機分類器，對柏林語音情緒資料庫進行情緒 分類分析。在加入非線性特徵後，男性及女性之情緒辨識率分別為 88.89%及 86.21%。AbstractThis study is focus on speech emotion recognition through machine learning method.We add two nonlinear dynamical features: Shannon entropy and curvature index, of each frame other than the traditional features such as pitch, formant, energy, MFCCs.After feature extraction, Fisher discriminant ratio and Genetic algorithm were applied in order to reduce the number of features. We use SVM classifier and cross validation method to discriminate seven emotions in Berlin emotion database. The analyzed results after adding of the nonlinear features show that the emotion recognition rates were 88.89% and 86.21% for male and female, respectively.關鍵詞：情緒辨識、非線性特徵、支持向量機 | Speech Emotion Recognition via Nonlinear Dynamical Features |
d51873630 | Humans rely on multiple sensory modalities when examining and reasoning over images. In this paper, we describe a new multimodal dataset that consists of gaze measurements and spoken descriptions collected in parallel during an image inspection task. The task was performed by multiple participants on 100 general-domain images showing everyday objects and activities. We demonstrate the usefulness of the dataset by applying an existing visual-linguistic data fusion framework in order to label important image regions with appropriate linguistic labels. | SNAG: Spoken Narratives and Gaze Dataset |
d7443422 | Social relations like power and influence are difficult concepts to define, but are easily recognizable when expressed. In this paper, we describe a multi-layer annotation scheme for social power relations that are recognizable from online written interactions. We introduce a typology of four types of power relations between dialog participants: hierarchical power, situational power, influence and control of communication. We also present a corpus of Enron emails comprising of 122 threaded conversations, manually annotated with instances of these power relations between participants. Our annotations also capture attempts at exercise of power or influence and whether those attempts were successful or not. In addition, we also capture utterance level annotations for overt display of power. We describe the annotation definitions using two example email threads from our corpus illustrating each type of power relation. We also present detailed instructions given to the annotators and provide various statistics on annotations in the corpus. | Annotations for Power Relations on Email Threads |
d252819014 | Long document summarisation, a challenging summarisation scenario, is the focus of the recently proposed LongSumm shared task. One of the limitations of this shared task has been its use of a single family of metrics for evaluation (the ROUGE metrics). In contrast, other fields, like text generation, employ multiple metrics. We replicated the LongSumm evaluation using multiple test set samples (vs. the single test set of the official shared task) and investigated how different metrics might complement each other in this evaluation framework. We show that under this more rigorous evaluation, (1) some of the key learnings from Longsumm 2020 and 2021 still hold, but the relative ranking of systems changes, and (2) the use of additional metrics reveals additional highquality summaries missed by ROUGE, and (3) we show that SPICE is a candidate metric for summarisation evaluation for LongSumm 1 . | Investigating Metric Diversity for Evaluating Long Document Summarisation |
d226262254 | Prior research has recognized the need to associate affective polarities with events and has produced several techniques and lexical resources for identifying affective events. Our research introduces new classification models to assign affective polarity to event phrases. First, we present a BERT-based model for affective event classification and show that the classifier achieves substantially better performance than a large affective event knowledge base. Second, we present a discourse-enhanced selftraining method that iteratively improves the classifier with unlabeled data. The key idea is to exploit event phrases that occur with a coreferent sentiment expression. The discourseenhanced self-training algorithm iteratively labels new event phrases based on both the classifier's predictions and the polarities of the event's coreferent sentiment expressions. Our results show that discourse-enhanced selftraining further improves both recall and precision for affective event classification. | Affective Event Classification with Discourse-enhanced Self-training |
d13833183 | The paper reports on the evaluation of a rule-based technique to model prototypical non-native pronunciation variants on the symbolic transcription level. This technique was developed to explore the possibility of an automatic generation of adapted pronunciation lexicons for different non-native speaker groups. The rule sets, which are currently available for nine language directions, are based on non-native speech data compiled specifically for this purpose. Since manual phonetic annotations are available for the speech data, the evaluation was performed on the transcription level by measuring the phonetic distance of the automatically generated pronunciations variants and actual pronunciations of non-native speakers. One of the central questions to be addressed by the evaluation is whether the rules have any predictive value: It has to be determined if and to what degree the rules are capable of generating realistic pronunciation variants for previously unseen speakers. Secondly, the rules should not only represent the pronunciations of individual speakers adequately; instead, they should be representative of speaker groups. The paper outlines the evaluation methodology and presents results for selected language directions. | Evaluation of Automatically Generated Transcriptions of Non-Native Pronunciations using a Phonetic Distance Measure |
d34277164 | Udapi is an open-source framework providing an application programming interface (API) for processing Universal Dependencies data. Udapi is available in Python, Perl and Java. It is suitable both for full-fledged applications and fast prototyping: visualization of dependency trees, format conversions, querying, editing and transformations, validity tests, dependency parsing, evaluation etc. | Udapi: Universal API for Universal Dependencies |
d231741125 | Fine-grained sentiment analysis attempts to extract sentiment holders, targets and polar expressions and resolve the relationship between them, but progress has been hampered by the difficulty of annotation. Targeted sentiment analysis, on the other hand, is a more narrow task, focusing on extracting sentiment targets and classifying their polarity.In this paper, we explore whether incorporating holder and expression information can improve target extraction and classification and perform experiments on eight English datasets. We conclude that jointly predicting target and polarity BIO labels improves target extraction, and that augmenting the input text with gold expressions generally improves targeted polarity classification. This highlights the potential importance of annotating expressions for fine-grained sentiment datasets. At the same time, our results show that performance of current models for predicting polar expressions is poor, hampering the benefit of this information in practice. | If you've got it, flaunt it: Making the most of fine-grained sentiment annotations |
d246063436 | Sentiment analysis is one of the most widely studied applications in NLP, but most work focuses on languages with large amounts of data. We introduce the first large-scale human-annotated Twitter sentiment dataset for the four most widely spoken languages in Nigeria-Hausa, Igbo, Nigerian-Pidgin, and Yorùbá-consisting of around 30,000 annotated tweets per language, including a significant fraction of code-mixed tweets. We propose text collection, filtering, processing, and labeling methods that enable us to create datasets for these low-resource languages. We evaluate a range of pre-trained models and transfer strategies on the dataset. We find that language-specific models and language-adaptive fine-tuning generally perform best. We release the datasets, trained models, sentiment lexicons, and code to incentivize research on sentiment analysis in under-represented languages. | NaijaSenti: A Nigerian Twitter Sentiment Corpus for Multilingual Sentiment Analysis |
d17409270 | This paper proposes a two-layered model of dialogue structure for task-oriented dialogues that processes contextual information and disambiguates speech acts. The final goal is to improve translation quality in a speech-to-speech translation system. | Improving Translation through Contextual Information |
d2321969 | This paper describes an algorithm for exact decoding of phrase-based translation models, based on Lagrangian relaxation. The method recovers exact solutions, with certificates of optimality, on over 99% of test examples. The method is much more efficient than approaches based on linear programming (LP) or integer linear programming (ILP) solvers: these methods are not feasible for anything other than short sentences. We compare our method to MOSES (Koehn et al., 2007), and give precise estimates of the number and magnitude of search errors that MOSES makes. | Exact Decoding of Phrase-Based Translation Models through Lagrangian Relaxation |
d18021356 | Previous research has shown that certain discourse conditions are necessary for the felicitous use of non-canonical syntactic forms like topicalization, left-dislocation, and clefts. However, the distribution of these forms does not correlate one-to-one with the presence of these conditions, and a system that generates these statisticallyrare forms based only on these conditions will overgenerate. Instead, a generation algorithm must be based on additional communicative goals that can be achieved through the use of these forms. Based on a corpus study, I present three types of communicative goals that speakers achieve through the use of non-canonical syntax. | Syntactic form and discourse function in NLG |
d218973744 | ||
d219304239 | ||
d3911097 | Understanding Human Action | |
d11012721 | ||
d30973252 | Text comparison is an interesting though hard task, with many applications in Natural Language Processing. This work introduces a new text-similarity measure, which employs named-entities' information extracted from the texts and the ngram graphs' model for representing documents. Using OpenCalais as a namedentity recognition service and the JIN-SECT toolkit for constructing and managing n-gram graphs, the text similarity measure is embedded in a text clustering algorithm (k-Means). The evaluation of the produced clusters with various clustering validity metrics shows that the extraction of named entities at a first step can be profitable for the time-performance of similarity measures that are based on the n-gram graph representation without affecting the overall performance of the NLP task. | A Graph-based Text Similarity Measure That Employs Named Entity Information |
d195063916 | ||
d5653822 | In this article we investigate logical metonymy, that is, constructions in which the argument of a word in syntax appears to be different from that argument in logical form (e.g., enjoy the book means enjoy reading the book, and easy problem means a problem that is easy to solve). The systematic variation in the interpretation of such constructions suggests a rich and complex theory of composition on the syntax/semantics interface. Linguistic accounts of logical metonymy typically fail to describe exhaustively all the possible interpretations, or they don't rank those interpretations in terms of their likelihood. In view of this, we acquire the meanings of metonymic verbs and adjectives from a large corpus and propose a probabilistic model that provides a ranking on the set of possible interpretations. We identify the interpretations automatically by exploiting the consistent correspondences between surface syntactic cues and meaning. We evaluate our results against paraphrase judgments elicited experimentally from humans and show that the model's ranking of meanings correlates reliably with human intuitions. | A Probabilistic Account of Logical Metonymy |
d241583806 | Query rewrite (QR) is an emerging component in conversational AI systems, reducing user defect. User defect is caused by various reasons, such as errors in the spoken dialogue system, users' slips of the tongue or their abridged language. Many of the user defects stem from personalized factors, such as user's speech pattern, dialect, or preferences. In this work, we propose a personalized search-based QR framework, which focuses on automatic reduction of user defect. We build a personalized index for each user, which encompasses diverse affinity layers to reflect personal preferences for each user in the conversational AI. Our personalized QR system contains retrieval and ranking layers. Supported by user feedback based learning, training our models does not require hand-annotated data. Experiments on personalized test set showed that our personalized QR system is able to correct systematic and user errors by utilizing phonetic and semantic inputs. | Personalized Search-based Query Rewrite System for Conversational AI |
d44266149 | 摘要 摘要 摘要 摘要. 近年來全球資訊網(World Wide Web，簡稱 Web)快速成長，不同來源、不同領域、不 同媒體的資訊透過網路傳遞到使用者手上。Web 除了扮演資訊傳播的角色外，也可以被視為是 一個超大的資料集，提供語料庫為基礎-統計導向方法(Corpus-Based Statistics-Oriented Approach) 所需要的統計值。本文以中文斷詞應用為例，由傳統語料庫和全球資訊網中，取得運用 word-based n-gram model 解斷詞歧義時所需要的統計值，藉以比較傳統語料庫和全球資訊網的差異。在第一 組實驗，我們假設完全沒有未知詞，運用傳統語料庫的統計值最佳，其次依序為 Google 為基礎、 AltaVista 為基礎、和 Openfind 為基礎。在第二組實驗，我們針對指定實體辨識，地名和組織名 這兩類有不錯的效能。在第三組實驗，我們整合斷詞系統與指定實體辨識模組，全球資訊網統計 值比傳統語料庫的統計值好。在最後一組實驗，我們將傳統語料庫和全球資訊網混合在一起，以 全球資訊網統計值解決未知詞問題，再以語料庫統計值解斷詞歧義性，實驗顯示具有最佳的斷詞 效能。 | 語料庫統計值與 語料庫統計值與 語料庫統計值與 語料庫統計值與全球資訊網 全球資訊網 全球資訊網 全球資訊網統計值之比較：以中文斷詞應用為例 統計值之比較：以中文斷詞應用為例 統計值之比較：以中文斷詞應用為例 統計值之比較：以中文斷詞應用為例 林筱晴 陳信希 國立台灣大學資訊工程學系 |
d8657338 | This paper describes research on automatic assessment of the pronunciation quality of spontaneous non-native adult speech. Since the speaking content is not known prior to the assessment, a two-stage method is developed to first recognize the speaking content based on non-native speech acoustic properties and then forced-align the recognition results with a reference acoustic model reflecting native and near-native speech properties. Features related to Hidden Markov Model likelihoods and vowel durations are extracted. Words with low recognition confidence can be excluded in the extraction of likelihood-related features to minimize erroneous alignments due to speech recognition errors. Our experiments on the TOEFL R Practice Online test, an English language assessment, suggest that the recognition/forced-alignment method can provide useful pronunciation features. Our new pronunciation features are more meaningful than an utterance-based normalized acoustic model score used in previous research from a construct point of view. | Improved Pronunciation Features for Construct-driven Assessment of Non-native Spontaneous Speech |
d6829190 | Ò ÒÝ ÓÒ×ØÖ ÒØ ÓÒ Ö ÙÑ ÒØ Ë Ð Ø ÓÒ £ Ã Ì Ã À ËÀÁ Ì ÍÒ Ú Ö× ØÝ Ó ÌÓ ÝÓ ¿¹ ¹½ ÃÓÑ ¸Å ÙÖÓ¹ ÙÌ Ó ÝÓ ½ ¿¹ ¼¾¸Â È AE ¹Ø Ô ÞººÙ¹ØÓ ÝÓº º Ô Ã ÝÓ× ÁËÀÁÃ Ï ÀÓ× ÍÒ Ú Ö× ØÝ ¾¹½ ¹½ Ù Ñ ¸ ÝÓ ¹ ÙÌ Ó ÝÓ ½¼¾¹ ½ ¼¸Â È AE ÝÓ× º Ó× º º Ô ×ØÖ Ø ÁÒ×Ø Ó ÔÓ× Ø Ò × Ô Ö Ø ×ÝÒØ Ø Ñ Ò ×Ñ×¸Û ÔÖÓÔÓ× × Ò Ð ÛÓÖ Ò Ñ ÓÖÝ Ñ ¹ Ò ×Ñ Ø Ø ÙÒ ÓÖÑÐÝ ÓÙÒØ× ÓÖ´ µ ÔÙÞÞÐ ÓÙØ ØÓÔ Ð Þ Ø ÓÒ ÔÓ ÒØ ÓÙØ Ò ÜÔÐ Ò Ò Ø Ä Ð Ø Ö ØÙÖ ¸´ µ ×ÝÑÑ ØÖ × Ò ÓÓÖ Ò Ø ÓÒ ×ØÖÙØÙÖ Ó × ÖÚ Ò Ò ÐÝÞ Ò Ú Ö ÓÙ× ×ÝÒØ Ø Ö Ñ ÛÓÖ ×¸ Ò ´ µ Ø « Ø× Ó Ò× ÖØ Ô Ö × × Ò Ô Ù× ×º | |
d219182207 | ||
d14974322 | Statistical-based collocation extraction approaches suffer from (1) low precision rate because high co-occurrence bi-grams may be syntactically unrelated and are thus not true collocations; (2) low recall rate because some true collocations with low occurrences cannot be identified successfully by statistical-based models. To integrate both syntactic rules as well as semantic knowledge into a statistical model for collocation extraction is one way to achieve a high precision while keeping a reasonable recall. This paper designs a cascade system which employs a hybrid model by integrating both syntactic and semantic knowledge into a statistical model for Chinese synonymous noun/verb collocations extraction. The grammatically bounded noun/verb collocations are extracted first from a syntactic-rule based module, which is then inputted to a semantic-based module for further retrieval of low frequent bi-gram collocations. | A Hybrid Extraction Model for Chinese Noun/Verb Synonym bi-gram Collocations |
d9784826 | The present status of Japanese speech database has been described. The database project in Japan started in early 1980s. The first one was a committee of Japan Electronic Industry Development Association, abbreviated as JEIDA, which aimed at creating a speech database that can commonly evaluate performance of the then existing speech input/output machines and systems. Several database projects have been undertaken since then including the one initiated by the Advanced Telecommunication Research Institute (ATR) and now it has come to the point where an enormous amount of spontaneous speech data is available. A survey has been conducted recently about the usage of the presently existing speech databases among industry and university institutions in Japan where speech research is now actively going on. It has been revealed that the ATR's continuous speech database is the most frequently used followed by the equivalent version of the Acoustical Society of Japan. | The Present Status of Speech Database in Japan: Development, Management, and Application to Speech Research |
d15213267 | In the research field of machine translation of patent documents, the issue of acquiring technical term translation equivalent pairs automatically from parallel patent documents is one of those most important. We take an approach of utilizing the phrase table of a state-of-the-art phrase-based statistical machine translation model. In this task, we consider situations where a technical term is observed in many parallel patent sentences and is translated into many translation equivalents. We apply SVM to the task of identifying synonymous translation equivalent pairs and achieve almost 98% precision and over 40% Fmeasure. Then, in order to improve recall, we introduce a semi-automatic framework, where we employ the strategy of selecting more than one seeds for each set of candidates bilingual synonymous term pairs. By manually judging whether each pair of two seeds is synonymous or not, we achieve over 95% precision and 50% recall. | Semi-Automatic Identification of Bilingual Synonymous Technical Terms from Phrase Tables and Parallel Patent Sentences |
d25036575 | This paper describes a Chinese word segmentation system based on unigram language model for resolving segmentation ambiguities. The system is augmented with a set of pre-processors and post-processors to extract new words in the input texts. | Unigram Language Model for Chinese Word Segmentation |
d5260871 | This paper presents hypothesis mixture decoding (HM decoding), a new decoding scheme that performs translation reconstruction using hypotheses generated by multiple translation systems. HM decoding involves two decoding stages: first, each component system decodes independently, with the explored search space kept for use in the next step; second, a new search space is constructed by composing existing hypotheses produced by all component systems using a set of rules provided by the HM decoder itself, and a new set of model independent features are used to seek the final best translation from this new search space. Few assumptions are made by our approach about the underlying component systems, enabling us to leverage SMT models based on arbitrary paradigms. We compare our approach with several related techniques, and demonstrate significant BLEU improvements in large-scale Chinese-to-English translation tasks. | Hypothesis Mixture Decoding for Statistical Machine Translation |
d28248912 | Cette recherche porte sur le chiasme de mots : figure de style jouant sur la réversion (ex. « Bonnet blanc, blanc bonnet »). Elle place le chiasme dans la problématique de sa reconnaissance automatique : qu'est-ce qui le définit et comment un ordinateur peut le trouver ? Nous apportons une description formelle du phénomène. Puis nous procédons à la constitution d'une liste d'exemples contextualisés qui nous sert au test des hypothèses. Nous montrons ainsi que l'ajout de contraintes formelles (contrôle de la ponctuation et omission des mots vides) pénalise très peu le rappel et augmente significativement la précision de la détection. Nous montrons aussi que la lemmatisation occasionne peu d'erreurs pour le travail d'extraction mais qu'il n'en est pas de même pour la racinisation. Enfin nous mettons en évidence que l'utilisation d'un thésaurus apporte quelques résultats pertinents.ABSTRACTTowards an automatic identification of chiasmus of wordsThis article summarises the study of the rhetorical figure "chiasmus" (e.g : "Quitters never win and winners never quit."). We address the problem of its computational identification. How can a computer identify this automatically ? For this purpose this article will provide a formal description of the phenomenon. First, we put together an annotated text for testing our hypothesis. At the end we demonstrate that the use of stopword lists and the identification of the punctuation improve the precision of the results with very little impact on the recall. We discover also that using lemmatization improves the results but stemming doesn't. Finally we see that a French thesaurus provided us with good results on the most elaborate form of chiasmus.MOTS-CLÉS : chiasme, rhétorique, antimétabole, figure de style. | Vers une identification automatique du chiasme de mots |
d259370542 | Temporal Knowledge Graph (TKG) reasoning aims to predict future facts based on historical data. However, due to the limitations in construction tools and data sources, many important associations between entities may be omitted in TKG. We refer to these missing associations as latent relations. Most of the existing methods have some drawbacks in explicitly capturing intra-time latent relations between co-occurring entities and inter-time latent relations between entities that appear at different times. To tackle these problems, we propose a novel Latent relations Learning method for TKG reasoning, namely L 2 TKG. Specifically, we first utilize a Structural Encoder (SE) to obtain representations of entities at each timestamp. We then design a Latent Relations Learning (LRL) module to mine and exploit the intraand inter-time latent relations. Finally, we extract the temporal representations from the output of SE and LRL for entity prediction. Extensive experiments on four datasets demonstrate the effectiveness of L 2 TKG. | Learning Latent Relations for Temporal Knowledge Graph Reasoning |
d259370805 | Recent advances in neural theorem-proving resort to large language models and tree searches. When proving a theorem, a language model advises single-step actions based on the current proving state and the tree search finds a sequence of correct steps using actions given by the language model. However, prior works often conduct constant computation efforts for each proving state while ignoring that the hard states often need more exploration than easy states. Moreover, they evaluate and guide the proof search solely depending on the current proof state instead of considering the whole proof trajectory as human reasoning does. Here, to accommodate general theorems, we propose a novel Dynamic-Tree Driven Theorem Solver (DT-Solver) by guiding the search procedure with state confidence and proof-level values. Specifically, DT-Solver introduces a dynamic-tree Monte-Carlo search algorithm, which dynamically allocates computing budgets for different state confidences, guided by a new proof-level value function to discover proof states that require substantial exploration. Experiments on two popular theorem-proving datasets, PISA and Mathlib, show significant performance gains by our DT-Solver over the state-of-the-art approaches, with a 6.65% improvement on average in terms of success rate. And especially under low computing resource settings (11.03% improvement on average). | DT-Solver: Automated Theorem Proving with Dynamic-Tree Sampling Guided by Proof-level Value Function |
d16695586 | This paper presents the methodology and results of a project for the large-scale analysis of public messages in political discourse on Facebook, the dominant social media site in Hungary. We propose several novel social psychologymotivated dimensions for natural language processing-based text analysis that go beyond the standard sentiment-based analysis approaches. Communion describes the moral and emotional aspects of an individual's relations to others, while agency describes individuals in terms of the efficiency of their goalorientated behavior. We treat these by custom lexicons that identify positive and negative cues in text. We measure the level of optimism in messages by examining the ratio of events talked about in the past, present and future by looking at verb tenses and temporal expressions. For assessing the level of individualism, we build on research that correlates it to pronoun dropping. We also present results that demonstrate the viability of our measures on 1.9 million downloaded public Facebook comments by examining correlation to party preferences in public opinion poll data. | Beyond Sentiment: Social Psychological Analysis of Political Facebook Comments in Hungary |
d8630854 | Two-step TAG Parsing Revisited | |
d225063191 | Aspect-level Sentiment Analysis Based on Hierarchical Attention and Gate Networks 轁轳 轡 輌轮轥輭轧轲轡轩轮轥轤 轴轡轳轫輬 轡轳轰轥轣轴輭转轥轶轥转 轳轥轮轴轩轭轥轮轴 轡轮轡转轹轳轩轳 轷轨软轳轥 轰轵轲轰软轳轥 轩轳 轴软 轩轤轥轮轴轩车轹 轴轨轥 轳轥轮轴轩轭轥轮轴 轰软转轡轲轩轴轹 轣软轲轲轥轳轰软轮轤轩轮轧 轴软 轴轨轥 轳轰轥轣轩輌轣 轡轳轰轥轣轴 轩轮 轡 轳轥轮轴轥轮轣轥 轨轡轳 轲轥轣轥轩轶轥轤 轭软轲轥 轡轮轤 轭软轲轥 轡轴轴轥轮轴轩软轮 轩轮 轴轨轥 轢轵轳轩轮轥轳轳 轡轮轤 轡轣轡轤轥轭轩轣輮 轔轨轥 轭软轳轴 软车 轴轨轥 轲轥转轡轴轥轤 轷软轲轫轳 车软轣轵轳 软轮 轴轨轥 轤轥轳轩轧轮 软车 轡轴轴轥轮轴轩软轮 轮轥轴轷软轲轫 轴软 轨轩轧轨转轩轧轨轴 轴轨轥 轤轩輋轥轲轥轮轴 轣软轮轴轲轩轢轵轴轩软轮轳 软车 轷软轲轤轳 轩轮 轣软轮轴轥轸轴 轡轮轤 轭轡轫轥 轣软轮轴轥轸轴 轡轮轤 轡轳轰轥轣轴 轩轮轴轥轲轡轣轴轩轶轥轤輮 轉轮 轴轨轩轳 轰轡轰轥轲輬 轷轥 轰轵轴 车软轲轷轡轲轤 轴轨轥 轨轩轥轲轡轲轣轨轩轣轡转 轡轴轴轥轮轴轩软轮 轡轮轤 轧轡轴轥 轮轥轴轷软轲轫轳 轴软 轰轲软轣轥轳轳 轡轳轰轥轣轴輭转轥轶轥转 轳轥轮轴轩轭轥轮轴 轡轮轡转轹轳轩轳 轴轡轳轫輮 轷轥 软轢轴轡轩轮 轴轨轥 轮轥轷 轲轥轰轲轥轳轥轮轴轡轴轩软轮 软车 轴轨轥 轣软轮轴轥轸轴 轴轨轲软轵轧轨 轴轨轥 轡轴轴轥轮轴轩软轮 轮轥轴轷软轲轫 轢轥轴轷轥轥轮 轴轨轥 轣软轮轴轥轸轴 轡轮轤 轴轨轥 轮轥轷 轲轥轰轲轥轳轥轮轴轡轴轩软轮 软车 轴轨轥 轡轳轰轥轣轴 轷轨轩轣轨 轷轥轩轧轨轴 轢轹 轴轨轥 轣软轮轴轥轸轴輮 轁轴 轴轨轥 轳轡轭轥 轴轩轭轥輬 轴轨轥 轵轳轥车轵转 轩轮车软轲轭轡轴轩软轮 轩轮 轴轨轥 轣软轮轴轥轸轴 轩轳 轳轥转轥轣轴轥轤 轴轨轲软轵轧轨 轴轨轥 轧轡轴轥 轮轥轴轷软轲轫轳 轴软 轥轮轲轩轣轨 轴轨轥 轲轥轰轲轥轳轥轮轴轡轴轩软轮 软车 轴轨轥 轣软轮轴轥轸轴輮 轔轨轥 轤轥轳轩轧轮 软车 轨轩轥轲轡轲轣轨轩轣轡转 轡轴轴轥轮轴轩软轮 轧轡轴轥 轮轥轴轷软轲轫轳 轭轡轫轥 轴轨轥 轲轥轰轲轥轳轥轮轴轡轴轩软轮 软车 轴轨轥 轣软轮轴轥轸轴 轡轮轤 轴轨轥 轡轳轰轥轣轴 轭软轲轥 轡轣轣轵轲轡轴轥輮轔轨轥 轥轸轰轥轲轩轭轥轮轴轡转 轲轥轳轵转轴轳 软轮 轴轨轥 轓轥轭轅轶轡转 輲輰輱輴 轔轡轳轫輴 轡轮轤 轔轷轩轴轴轥轲 轳轨软轷 轴轨轥 轶轡转轩轤轩轴轹 软车 轴轨轥 轭软轤轥转輮 | |
d18296816 | In this paper, we propose a novel method for generating a coarse-grained sense inventory from Wikipedia using a machine learning framework. Structural and content-based features are employed to induce clusters of articles representative of a word sense. Additionally, multilingual features are shown to improve the clustering accuracy, especially for languages that are less comprehensive than English. We show the effectiveness of our clustering methodology by testing it against both manually and automatically annotated datasets. | Sense Clustering Using Wikipedia |
d219303897 | ||
d1163931 | The analysis of sound and sonic devices in poetry is the focus of much poetic scholarship, and poetry scholars are becoming increasingly interested in the role that computation might play in their research. Since the nature of such sonic analysis is unique, the associated tasks are not supported by standard text analysis techniques. We introduce a formalism for analyzing sonic devices in poetry. In addition, we present RhymeDesign, an open-source implementation of our formalism, through which poets and poetry scholars can explore their individual notion of rhyme. | RhymeDesign: A Tool for Analyzing Sonic Devices in Poetry |
d5446291 | This article describes an approach to Lexical-Functional Grammar (LFG) generation that is based on the fact that the set of strings that an LFG grammar relates to a particular acyclic f-structure is a context-free language. We present an algorithm that produces for an arbitrary LFG grammar and an arbitrary acyclic input f-structure a context-free grammar describing exactly the set of strings that the given LFG grammar associates with that f-structure. The individual sentences are then available through a standard context-free generator operating on that grammar. The context-free grammar is constructed by specializing the context-free backbone of the LFG grammar for the given f-structure and serves as a compact representation of all generation results that the LFG grammar assigns to the input. This approach extends to other grammatical formalisms with explicit context-free backbones, such as PATR, and also to formalisms that permit a context-free skeleton to be extracted from richer specifications. It provides a general mathematical framework for understanding and improving the operation of a family of chart-based generation algorithms. | LFG Generation by Grammar Specialization |
d170476009 | ||
d16894077 | We present a novel scheme for wordbased Japanese typed dependency parser which integrates syntactic structure analysis and grammatical function analysis such as predicate-argument structure analysis. Compared to bunsetsu-based dependency parsing, which is predominantly used in Japanese NLP, it provides a natural way of extracting syntactic constituents, which is useful for downstream applications such as statistical machine translation. It also makes it possible to jointly decide dependency and predicate-argument structure, which is usually implemented as two separate steps. We convert an existing treebank to the new dependency scheme and report parsing results as a baseline for future research. We achieved a better accuracy for assigning function labels than a predicate-argument structure analyzer by using grammatical functions as dependency label. | Word-based Japanese typed dependency parsing with grammatical function analysis |
d258999673 | Standard decoding approaches for conditional text generation tasks typically search for an output hypothesis with high model probability, but this may not yield the best hypothesis according to human judgments of quality. Reranking to optimize for downstream metrics can better optimize for quality, but many metrics of interest are computed with pre-trained language models, which are slow to apply to large numbers of hypotheses. We explore an approach for reranking hypotheses by using Transformers to efficiently encode lattices of generated outputs, a method we call EEL. With a single Transformer pass over the entire lattice, we can approximately compute a contextualized representation of each token as if it were only part of a single hypothesis in isolation. We combine this approach with a new class of token-factored rerankers (TFRs) that allow for efficient extraction of high reranker-scoring hypotheses from the lattice. Empirically, our approach incurs minimal degradation error compared to the exponentially slower approach of encoding each hypothesis individually. When applying EEL with TFRs across three text generation tasks, our results show both substantial speedup compared to naive reranking and often better performance on downstream metrics than comparable approaches. | EEL: Efficiently Encoding Lattices for Reranking |
d208331031 | ||
d52141191 | Philippinization of English has come full circle: It has penetrated not only the center but also the periphery. This paper demonstrates a trend of nativization of English in a rural area as seen in a local daily. Thirty newspaper articles from The Stalwart Journal, a provincewide weekly circulating bilingual journal in the island province of Masbate, Bicol Region in the Philippines, were examined to identify the local lexical intrusions in the English text. The borrowing and assimilation of local lexical terms were analyzed and categorized. These lexical items were found in various categories: people, cultural events, cultural groups, public and private organizations, government programs, program units, government agencies, places, broadcast and social media, transportation, food, animals, human descriptions, public services, and other items. English nativization is shown in the borrowing and switching to local or native lexis in the news articles of the local daily. | Intrusions of Masbate Lexicon in Local Bilingual Tabloid |
d6315014 | This work presents a flexible and efficient discriminative training approach for statistical machine translation. We propose to use the RPROP algorithm for optimizing a maximum expected BLEU objective and experimentally compare it to several other updating schemes. It proves to be more efficient and effective than the previously proposed growth transformation technique and also yields better results than stochastic gradient descent and AdaGrad. We also report strong empirical results on two large scale tasks, namely BOLT Chinese→English and WMT German→English, where our final systems outperform results reported by Setiawan and Zhou (2013) and on matrix.statmt.org. On the WMT task, discriminative training is performed on the full training data of 4M sentence pairs, which is unsurpassed in the literature.2. In terms of time and memory efficiency, RPROP clearly outperforms GT. The latter needs to update a much larger number of features due to its renormalization component. On the IWSLT data, RPROP is 6.4 times faster than GT and requires a third of the memory.3. On the WMT German→English task, we perform discriminative training on 4M sentence 1516 pairs, which, to the best of our knowledge, is 2.4 times the size of the largest training set reported in previous work (1.66M sentences in(Simianer et al., 2012)). This proves the scalability of our approach.4. On two large scale tasks our experiments show good improvements over strong baselines which include recurrent language modeling components. On the Chinese→English DARPA BOLT task, we achieve nearly twice the improvement reported in (Setiawan and Zhou, 2013) on the same test sets which results in a superior final system. Finally, the best single system reported on matrix.statmt.org is outperformed by 0.8 BLEU points on the WMT German→English newstest2013 set.Our experiments also prove that leave-one-out impacts translation quality.This paper is organized as follows. We review related work in Section 2 and present the translation system in Section 3. In Section 4 we describe the different discriminative update strategies applied in this work and Section 5 derives the complete maximum expected BLEU training algorithm. Finally, experimental results are given in Section 6 and we conclude with Section 7. | A Comparison of Update Strategies for Large-Scale Maximum Expected BLEU Training |
d5816453 | In this paper, we extend current state-of-theart research on unsupervised acquisition of scripts, that is, stereotypical and frequently observed sequences of events. We design, evaluate and compare different methods for constructing models for script event prediction: given a partial chain of events in a script, predict other events that are likely to belong to the script. Our work aims to answer key questions about how best to (1) identify representative event chains from a source text, (2) gather statistics from the event chains, and (3) choose ranking functions for predicting new script events. We make several contributions, introducing skip-grams for collecting event statistics, designing improved methods for ranking event predictions, defining a more reliable evaluation metric for measuring predictiveness, and providing a systematic analysis of the various event prediction models. | Skip N-grams and Ranking Functions for Predicting Script Events |
d1588782 | We address the task of computing vector space representations for the meaning of word occurrences, which can vary widely according to context. This task is a crucial step towards a robust, vector-based compositional account of sentence meaning. We argue that existing models for this task do not take syntactic structure sufficiently into account.We present a novel structured vector space model that addresses these issues by incorporating the selectional preferences for words' argument positions. This makes it possible to integrate syntax into the computation of word meaning in context. In addition, the model performs at and above the state of the art for modeling the contextual adequacy of paraphrases. | A Structured Vector Space Model for Word Meaning in Context |
d8382317 | First-order factoid question answering assumes that the question can be answered by a single fact in a knowledge base (KB). While this does not seem like a challenging task, many recent attempts that apply either complex linguistic reasoning or deep neural networks achieve 65%-76% accuracy on benchmark sets. Our approach formulates the task as two machine learning problems: detecting the entities in the question, and classifying the question as one of the relation types in the KB. We train a recurrent neural network to solve each problem. On the SimpleQuestions dataset, our approach yields substantial improvements over previously published results -even neural networks based on much more complex architectures. The simplicity of our approach also has practical advantages, such as efficiency and modularity, that are valuable especially in an industry setting. In fact, we present a preliminary analysis of the performance of our model on real queries from Comcast's X1 entertainment platform with millions of users every day. | No Need to Pay Attention: Simple Recurrent Neural Networks Work! (for Answering "Simple" Questions) |
d34416192 | It is well-known that Chinese has SVO as predominant word order, with variant orders OSV and SOV marking topic and focus, intimately linked to the topic-prominence of the language. Assuming early setting of the head parameter in syntactic acquisition and the peripheral positions of topic and focus in clausal structure, one might hypothesize that Chinese-speaking children will acquire the predominant SVO order early, but develop the variant orders later. Such acquisition findings, if true, would seemingly go against the idea of a topic prominence parameter. This paper explores the development of topic prominence by examining word order in early child Chinese, based on naturalistic longitudinal corpora as well as cross-sectional experiments that investigated the relative accessibility of different word orders.Our naturalistic data show that the word order of two-year-old Chinese children reflects adherence to canonical mapping of thematic roles to structural positions, as well as sensitivity to the unselectivity of subject and object. While sentences of OSV order appeared around two years of age, double nominative structures were virtually absent before three, suggesting that as a typological characteristic, topic-prominence is not acquired early. Our experimental results show that Mandarin-speaking children by three years of age have established SVO solidly as the dominant word order, on both comprehension and production, but still find the topicalized and fronting orders (OSV, SOV) difficult, indicating that the structures of the left periphery may be acquired at a later stage, and at different times. The implications of these acquisition findings for the topic prominence parameter will be explored. PACLIC 24 Proceedings 15 | The acquisition of word order in a topic-prominent language: Corpus findings and experimental investigation |
d45221071 | Dans cet article, nous montrons comment l'utilisation conjointe d'une technique d'alignement de phrases parallèles à la demande et d'estimation de modèles de traduction à la volée permet une réduction en temps très notable (jusqu'à 93% dans nos expériences) par rapport à un système à l'état de l'art, tout en offrant un compromis en termes de qualité très intéressant dans certaines configurations. En particulier, l'exploitation immédiate de documents traduits permet de compenser très rapidement l'absence d'un corpus de développement.Abstract. In this article, we show how using both on-demand alignment of parallel sentences and on-the-fly estimation of translation models can yield massive reduction (up to 93% in our experiments) in development time as compared to a state-of-the-art system, while offering an interesting tradeoff as regards translation quality under some configurations. We show in particular that the absence of a development set can be quickly compensated by immediately using translated documents.Mots-clés : traduction automatique statistique ; développement efficace ; temps de calcul. | 21 ème Traitement Automatique des Langues Naturelles |
d9143314 | Semantic language model is a technique that utilizes the semantic structure of an utterance to better rank the likelihood of words composing the sentence. When used in a conversational system, one can dynamically integrate the dialog state and domain semantics into the semantic language model to better guide the speech recognizer executing the decoding process. We describe one such application that employs semantic language model to cope with spontaneous speech in a robust manner. The semantic language model, though can be manually crafted without data, can benefit significantly from data driven machine learning techniques. An example based approach is also described here to demonstrate a viable approach. | Use and Acquisition of Semantic Language Model |
d6574418 | Adjuncts are characteristically optional, but many, such as adverbs and adjectives, are strictly ordered. In Minimalist Grammars (MGs), it is straightforward to account for optionality or ordering, but not both. I present an extension of MGs, MGs with Adjunction, which accounts for optionality and ordering simply by keeping track of two pieces of information at once: the original category of the adjoined-to phrase, and the category of the adjunct most recently adjoined. By imposing a partial order on the categories, the Adjoin operation can require that higher adjuncts precede lower adjuncts, but not vice versa, deriving order. | Order and Optionality: Minimalist Grammars with Adjunction |
d2109061 | We describe the SUPERSENSELEARNER system that participated in the English allwords disambiguation task. The system relies on automatically-learned semantic models using collocational features coupled with features extracted from the annotations of coarse-grained semantic categories generated by an HMM tagger. | UNT-Yahoo: SuperSenseLearner: Combining SenseLearner with SuperSense and other Coarse Semantic Features |
d1860368 | A research program is described in which a particular representational format for meaning is tested as broadly as possible. | Testing The Psychological Reality of a Representational Model |
d9454249 | On Deftly Introducing Procedural Elements into Unification Parsing | |
d17185822 | This paper presents the approach of the GTI Research Group to SemEval-2016 task 4 on Sentiment Analysis in Twitter, or more specifically, subtasks A (Message Polarity Classification), B (Tweet classification according to a two-point scale) and D (Tweet quantification according to a two-point scale). We followed a supervised approach based on the extraction of features by a dependency parsing-based approach using a sentiment lexicon and Natural Language Processing techniques. | Training a Naive Bayes Classifier using Features of an Unsupervised System |
d22418524 | Vers un treebank du français parlé | |
d16438200 | ||
d218974366 | ||
d67856536 | This paper describes our submission for the SemEval-2019 Suggestion Mining task. A simple Convolutional Neural Network (CNN) classifier with contextual word representations from a pre-trained language model is used for sentence classification. The model is trained using tri-training, a semi-supervised bootstrapping mechanism for labelling unseen data. Tri-training proved to be an effective technique to accommodate domain shift for cross-domain suggestion mining (Subtask B) where there is no hand labelled training data. For in-domain evaluation (Subtask A), we use the same technique to augment the training set. Our system ranks thirteenth in Subtask A with an F 1 -score of 68.07 and third in Subtask B with an F 1 -score of 81.94. | Zoho at SemEval-2019 Task 9: Semi-supervised Domain Adaptation using Tri-training for Suggestion Mining |
d67856251 | We introduce a new method to tag Multiword Expressions (MWEs) using a linguistically interpretable language-independent deep learning architecture. We specifically target discontinuity, an under-explored aspect that poses a significant challenge to computational treatment of MWEs. Two neural architectures are explored: Graph Convolutional Network (GCN) and multi-head self-attention. GCN leverages dependency parse information, and self-attention attends to long-range relations. We finally propose a combined model that integrates complementary information from both, through a gating mechanism. The experiments on a standard multilingual dataset for verbal MWEs show that our model outperforms the baselines not only in the case of discontinuous MWEs but also in overall F-score. 1 | Bridging the Gap: Attending to Discontinuity in Identification of Multiword Expressions |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.