| { |
| "paper_id": "S01-1024", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T15:35:36.791253Z" |
| }, |
| "title": "Sense and Deduction: The Power of Peewees Applied to the SENSEVAL-2 Swedish Lexical Sample Task", |
| "authors": [ |
| { |
| "first": "Torbjorn", |
| "middle": [], |
| "last": "Lagert", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Uppsala University +Hapax Information Systems AB", |
| "location": { |
| "settlement": "Stockholm" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Natalia", |
| "middle": [], |
| "last": "Zinovjevat", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Uppsala University +Hapax Information Systems AB", |
| "location": { |
| "settlement": "Stockholm" |
| } |
| }, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "This paper describes our use of Prolog Word Experts (PWEs) in the SENSEVAL-2 competition. We explain how we specify our PWEs as sequences of transformation rules and how they can be trained on sense tagged corpus data. We give a semantics of PWEs by translating them into first order predicate logic, and we describe how PWEs can be compiled into Prolog procedures. We finally present our results for the Swedish lexical sample task: 63% (fine-grained score) for our best PWE, and a second place in the ranking.", |
| "pdf_parse": { |
| "paper_id": "S01-1024", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "This paper describes our use of Prolog Word Experts (PWEs) in the SENSEVAL-2 competition. We explain how we specify our PWEs as sequences of transformation rules and how they can be trained on sense tagged corpus data. We give a semantics of PWEs by translating them into first order predicate logic, and we describe how PWEs can be compiled into Prolog procedures. We finally present our results for the Swedish lexical sample task: 63% (fine-grained score) for our best PWE, and a second place in the ranking.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "Word experts are small expert system-like modules for processing a particular target word based on neighboring words. Typically, a word expert uses rules that test the identity and relative position of words in the context in order to infer the role of the target word in the passage (Berleant, 1995) . In this paper, we describe the development of various kinds of word experts in a logic programming framework, dealing with word sense disambiguation in the context of the SENSEVAL-2 competition.", |
| "cite_spans": [ |
| { |
| "start": 284, |
| "end": 300, |
| "text": "(Berleant, 1995)", |
| "ref_id": "BIBREF0" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "In a logic programming framework, the task of engineering a word (sense) expert can be specified as follows. Given a suitable representation of a text, we want to define a predicate sense/2 such that sense (P, S) is true iff the word at position P in the text has the senseS. In the remainder of the paper, we will refer to this kind of word expert as a Pro log Word Expert (or PWE for short -\"Peewee\" to its friends). This is to distinguish it from other kinds of word experts, and to emphasize the fact that it is 'programmed in logic'.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "2 The Anatomy of a Peewee", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "99", |
| "sec_num": null |
| }, |
| { |
| "text": "In the present paper, a word expert's knowledge will be expressed, not as Prolog clauses defining sense/2 directly, but as a sequence of transformation rules. For example, here is how we specify a word expert which is able to disambiguate occurrences of interest: 1 word_expert sense := sense:add 6 <-word:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Specifications", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "interest@[O] o sense:6>1 <-word:in@[l] o sense:1>5 <-word:'/.'@[-1] o end.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Specifications", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "The first rule works as a default rule, which simply assigns the most frequent sense to the word interest ( 6 in this case). If no other rules apply, this is the tag that the word will eventually get. The other rules dictate when based on the context -a word should have its tag changed. The second rule is to be read \"replace the tag for sense 6 with the tag for sense 1, if the next word is in\". The third rule says \"replace the tag for sense 1 with the tag for sense 5, if the previous 'word' is '%'.\" The o-symbol is a composition operator, and (R o Rs) basically means that the output of applying the rule R forms the input to the application of the rules Rs. Thus, rules are strictly order-dependent. Note, for example, that the third rule is applicable only if the second rule is.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Specifications", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "Needless to say, the above rules are not at all sufficient for the task of disambiguating all uses of interest. But the number of rules can be increased, and typically a word expert will have access to anything between just a handful of rules and several hundred ones. 2", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Specifications", |
| "sec_num": "2.1" |
| }, |
| { |
| "text": "Interestingly, a sequence of transformation rules can be translated into a set of axioms, expressed in first-order predicate logic, defining relationships between positions in a text, word forms, and senses (Lager, 2000; Lager & Nivre, 2001 ). For example, the meaning of the rules from the previous section can be spelled out as follows:", |
| "cite_spans": [ |
| { |
| "start": 207, |
| "end": 220, |
| "text": "(Lager, 2000;", |
| "ref_id": "BIBREF4" |
| }, |
| { |
| "start": 221, |
| "end": 240, |
| "text": "Lager & Nivre, 2001", |
| "ref_id": "BIBREF5" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Vp[w(p,interest)-> S1 (p,6)] Vpo,Pl [S1 (po,6) 1\\ Pl=Po+ll\\ w(p1,in) -> S2(Po,l)j Vpo,pl,x[SI(Po,x) l\\p1=po+ll\\ \u2022w(p1,in)-> S2(po,x)] Vpo,PI [S2(po,l) 1\\ Pl=po-11\\ w(p1 ,%) -> S3(po,5)] iipo,Pl ,x[S2(Po,x) 1\\ Pl=Po-11\\ \u2022w(pl ,%) -> Sa(po,x)] Vx,p[S3(p,x)-> S(p,x)]", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "The idea is that for each rule in the sequence a new predicate si is introduced, where the subscript indicates where in the sequence the rule belongs. Semantically, Si relates a position to a sense, and the formulas define this predicate in terms of the predicate Si-1 plus a number of other predicates. Each Si corresponding to a replacement rule is defined by two sentencesone stating the conditions under which a sense tag is replaced with another sense tag, the other one stating the conditions under which the old sense tag is kept.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Given a suitable logical representation of a text, such as", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "w(1,Sue) w(2,developed) w(3,an) w(4,interest) w(5, in) w(6, computers) w(7, and) w(8, bought) w(9, an) w(10, 11.5) w(ll, %) w(12, interest) w(13, in) w(14, Microsoft)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "and given a suitable constructive proof method, the exact identity of the sense of an occurrence of the word interest -say the word at position 12will follow as a logical consequence of the theory formed by taking the union of the previous two sets of formulas. For example, the formula 3x[S(12,x)) is a theorem, for which we can construct (only) the example x --+ 5, and we have thus formally proved that this particular occurrence of interest means \"a share in a company\" . 3", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "http://www.ling.gu.se/~lager/Home/pwe_ui.html 3 The theory can be used in other ways too. Searching", |
| "cite_spans": [ |
| { |
| "start": 46, |
| "end": 47, |
| "text": "3", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Logic", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "What we have here is something that we like to think of as word sense disambiguation as deduction, in analogy to the ideas of parsing as deduction due to Pereira and Warren (1983) .", |
| "cite_spans": [ |
| { |
| "start": 154, |
| "end": 179, |
| "text": "Pereira and Warren (1983)", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "100", |
| "sec_num": null |
| }, |
| { |
| "text": "Since the above formulas have already logic programming form, it is straightforward to translate them into Prolog. For example, the second and the third formulas can be translated as follows:4", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Peewee Compiler", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "s2(P0,1) s2(PO,X) si(P0,6), Pi is P0+1, w(Pi,in). si(PO,X), Pi is PO+i, \\+ w(Pl,in).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Peewee Compiler", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "To write Prolog procedures such as these by hand for many rules would be tedious and prone to errors. Fortunately, since the formalism for transformation rules is compositional, it was straightforward to write a compiler 5 that generates word expert procedures from word expert specifications automatically.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Peewee Compiler", |
| "sec_num": "2.3" |
| }, |
| { |
| "text": "There is an obvious choice of learning method for training Prolog Word Experts, namely Transformation-Based Learning (Brill, 1995) . Of course, the fact that transformation rules can be learned from tagged corpora was a major reason for using them in the first place. The J_L-TBL system -described in detail in (Lager, 1999 )uses the search and database capabilities of the Prolog programming language to implement a generalized form of transformation-based learning. Through its support of a compositional rule/template formalism and 'pluggable' algorithms, the J_L-TBL system can easily be tailored to different learning tasks. 6 Rules that can be learned in Transformation-Based Learning are instances of rule templates. For example, the second of the rules in our example PWE specification is an instance of the following template:", |
| "cite_spans": [ |
| { |
| "start": 117, |
| "end": 130, |
| "text": "(Brill, 1995)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 311, |
| "end": 323, |
| "text": "(Lager, 1999", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 630, |
| "end": 631, |
| "text": "6", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "sense:A>B <-word:C@[1].", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "for a word token with a particular sense (say 5) becomes a matter of constructively proving 3p[S(p,5)].", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "4 There are equivalent but more efficient ways to represent these clauses in Prolog (cf. Lager, 2000 The template is to be read \"replace the tag for sense A with the tag for sense B if the word immediately to the right is C\", where A, B and C are variables. Learning is a matter of repeatedly instantiating rule templates in training data, scoring rules on the basis of counts of positive and negative evidence of them, selecting the highest scoring rule on the basis of this ranking, and applying it to the training data.", |
| "cite_spans": [ |
| { |
| "start": 89, |
| "end": 100, |
| "text": "Lager, 2000", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "3 Peewees at SENSEVAL-2", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "The lexical sample task for Swedish in SENSEVAL-2 involved 40 lemmas: 20 nouns, 15 verbs and 5 adjectives. Together they represented 145 senses and 304 sub-senses. 8, 718 annotated instances were provided as training material and 1,525 unannotated instances were provided for testing. Furthermore, a lexicon -the GLDB (Gothenburg Lexical Database)complete with morphological information, definitions, language examples, etc. was available.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "Our team explored three approaches. For each lemma, we trained:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "\u2022 PWE-smpl: a simple PWE capable of arriving at a single sense for each instance of that lemma in the testing material.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "\u2022 PWE-disj: a committee of PWEs (i.e. a set of PWEs) capable of arriving at (possibly) multiple senses for each instance of that lemma, by collecting the individual results into a set.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "\u2022 PWE-vote: a committee of PWEs capable of arriving at a single sense for each instance of that lemma, by applying a simple voting procedure.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "As it turned out, the second of these approaches produced a rather unimpressive result, and we will therefore spend very little time discussing it. Indeed, had we been able to run the scoring software ourselves (which we were not), we would have left them outside the competition altogether.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Peewee Training", |
| "sec_num": "2.4" |
| }, |
| { |
| "text": "For the training of our simplest form of sense disambiguation expert, the following set of seven templates was used: The idea was to exploit. a fact noted by many researchers in the field: that the sense of an occurrence of a word can fairly successfully be determined from just looking at the two previous words and the two following words ( cf. Ide & Veronis, 1998) . The choice of the above set of templates is based on a fairly thorough trailand-error process and works well for most words that we have tried.", |
| "cite_spans": [ |
| { |
| "start": 347, |
| "end": 367, |
| "text": "Ide & Veronis, 1998)", |
| "ref_id": "BIBREF2" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Simple Peewees", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "The idea here was to train five different PWEs for each lemma, and then to use a simple voting mechanism to arrive at a final decision. The PWEs were different only in that they used different sets of templates during the training. Templates looking forwards only, templates looking backwards only, and templates looking both forwards and backwards. Furthermore, one member in each committee was trained for using a bag-of-words approach to disambiguation, based on templates of the following form: Finally, one PWE in each committee had access to a list of words extracted from the language examples provided by the GLDB.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Peewee Committees", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "In this section we describe the actions that we took in order to submit our entry in the competition.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Procedure", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\u2022 In a preparatory step, the XML formatted training data was parsed and subsequently converted into the format required by the p-TBL system.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Procedure", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\u2022 The training was performed, and resulted in one PWE specification per lemma.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Procedure", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "Training took between 5 seconds and a couple of minutes per lemma, depending on the amount of training data available for the lemma in question.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Procedure", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\u2022 The PWE specifications were compiled into a set of PWE procedures, by means of the PWE compiler.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Procedure", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "\u2022 Simple procedures were written to print the results to a file in the prescribed format, and the PWEs were then run on the test data. This took only a couple of seconds for the whole test corpus.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Procedure", |
| "sec_num": "3.3" |
| }, |
| { |
| "text": "In the following table we show the results of our entry in the competition, copied from the SENSEVAL-2 homepage. 7 Five groups and altogether eight systems participated in the Swedish lexical sample task. In terms of ranking, our PWE-vote came in second, after Yarowski's JHU system, and before the Goteborg team's best entry. However, we hasten to add that the step from Yarowski's (nearly 70%, fine grained evaluation) to our results is a very significant 7%, and that the step down to Goteborg's result is very small and probably statistically insignificant. Our simple Peewees shared the fourth place with Resnik et al.'s UMD-SST. As can be seen from the table, the PWE committees did slightly better than a single simple PWE. It is however dubious whether the small difference was really worth the trouble. It is quite possible that training a single PWE on the combination of corpus data and the examples from the GLDB would have lead to a result almost as good, and with less work.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "3.4" |
| }, |
| { |
| "text": "It seems we can conclude that an approach to word sense disambiguation based on Transformation-Based Learning is competitive with approaches based on Memory-Based Learning as used by the Goteborg team, and support vector machine (SVM) learning, used by the University of Maryland team, This is good news for those aiming at building NLP systems in which transformation rules play a major role.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "4" |
| }, |
| { |
| "text": "As we have seen, there is meaning in the life of Peewees, and sound mathematical meaning at that! Also, given the link between first order logic and a logic programming language such as Prolog, the implementation follows very directly from the specification. The existence of a compiler from Peewee specifications into Frolog procedures makes Peewees very convenient to work with in a Prolog environment.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "4" |
| }, |
| { |
| "text": "This word was of course not used the Swedish task, but is used here for expository reasons. The sense tags are numbers: l=\"readiness to give attention\", 5=\"a company share\", 6= \"money paid for the use of money\", etc.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "A demo of a more potent PWE is available at:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Note that the coarse-grained evaluation was not applicable to the Swedish task. Also, it should be noted that our results in the first round of evaluation were slightly worse than the results reported here. However, this was due to a spelling error which could be corrected by the conference organizers and thus did not involve any resubmission of test results.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "We thank the SENSEVAL-2 organizers for making all this possible, and in particular Jerker Jarborg and Dimitrios Kokkinakis in Goteborg for their work on preparing for the Swedish lexical sample task.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgements", |
| "sec_num": "5" |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Engineering \"Word Experts\" for Word Disambiguation", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Berleant", |
| "suffix": "" |
| } |
| ], |
| "year": 1995, |
| "venue": "Natural Language Engineering", |
| "volume": "", |
| "issue": "4", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Berleant, D. (1995) Engineering \"Word Ex- perts\" for Word Disambiguation. Natural Language Engineering, 1(4).", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Transformation-Based Error-Driven Learning and Natural Language Processing: A Case Study in Part of Speech Tagging", |
| "authors": [ |
| { |
| "first": "E", |
| "middle": [], |
| "last": "Brill", |
| "suffix": "" |
| } |
| ], |
| "year": 1995, |
| "venue": "Computational Linguistics", |
| "volume": "21", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Brill, E. (1995) Transformation-Based Error- Driven Learning and Natural Language Pro- cessing: A Case Study in Part of Speech Tag- ging. Computational Linguistics 21.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Introduction to the Special Issue on Word Sense Disambiguation: The State of the Art", |
| "authors": [ |
| { |
| "first": "N", |
| "middle": [], |
| "last": "Ide", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Veronis", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "Computational Linguistics", |
| "volume": "24", |
| "issue": "1", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ide, N. and Veronis, J. (1998) Introduction to the Special Issue on Word Sense Disambigua- tion: The State of the Art. Computational Linguistics 24( 1).", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "The J-L-TBL System: Logic Programming Tools for Transformation-Based Learning", |
| "authors": [ |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Lager", |
| "suffix": "" |
| } |
| ], |
| "year": 1999, |
| "venue": "Proceedings of CoNLL 'g9", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Lager, T. (1999) The J-L-TBL System: Logic Programming Tools for Transformation- Based Learning. In Proceedings of CoNLL 'g9, Bergen, Norway.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "A Logic Programming Approach to Word Expert Engineering", |
| "authors": [ |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Lager", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "Proceedings of ACIDCA 2000: Workshop on Corpora and Nat ural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Lager, T. (2000) A Logic Programming Ap- proach to Word Expert Engineering. In Pro- ceedings of ACIDCA 2000: Workshop on Corpora and Nat ural Language Processing, Monastir, Tunisia, March 22-24 2000.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "Part of Speech Tagging from a Logical Point of View", |
| "authors": [ |
| { |
| "first": "T", |
| "middle": [], |
| "last": "Lager", |
| "suffix": "" |
| }, |
| { |
| "first": "J", |
| "middle": [], |
| "last": "Nivre", |
| "suffix": "" |
| } |
| ], |
| "year": 2001, |
| "venue": "Logical Aspects of Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Lager, T. and Nivre, J. (2001) Part of Speech Tagging from a Logical Point of View. In de Groote, P., Morrill, G., Retor, C. (eds.) Logical Aspects of Computational Linguistics. Springer-Verlag, LNAI. VOL. 2099.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Parsing as Deduction", |
| "authors": [ |
| { |
| "first": "F", |
| "middle": [], |
| "last": "Pereira", |
| "suffix": "" |
| }, |
| { |
| "first": "D", |
| "middle": [ |
| "H D" |
| ], |
| "last": "Warren", |
| "suffix": "" |
| } |
| ], |
| "year": 1983, |
| "venue": "Proceedings of the 21th Meeting of the ACL", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Pereira, F. and Warren, D. H. D. (1983) Pars- ing as Deduction, In Proceedings of the 21th Meeting of the ACL.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "num": null, |
| "text": "sense:A>B <-inBag:W~[O]. sense:A>B <-inBag:W1~[0] & inBag:W2~[0].", |
| "uris": null, |
| "type_str": "figure" |
| }, |
| "TABREF1": { |
| "num": null, |
| "text": ":A>B <-word:C~[-1] & word:D~(-2]. sense:A>B <-word:C~[-1) & word:D~[1).", |
| "content": "<table><tr><td>sense:A>B <-word:C~[1].</td></tr><tr><td>sense:A>B <-word:C~[1,2].</td></tr><tr><td>sense:A>B <-word:C~[1) & word:D~[2].</td></tr></table>", |
| "type_str": "table", |
| "html": null |
| } |
| } |
| } |
| } |
