| { | |
| "paper_id": "S07-1048", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T15:23:01.022197Z" | |
| }, | |
| "title": "LTH: Semantic Structure Extraction using Nonprojective Dependency Trees", | |
| "authors": [ | |
| { | |
| "first": "Richard", | |
| "middle": [], | |
| "last": "Johansson", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Lund University", | |
| "location": { | |
| "country": "Sweden" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| }, | |
| { | |
| "first": "Pierre", | |
| "middle": [], | |
| "last": "Nugues", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Lund University", | |
| "location": { | |
| "country": "Sweden" | |
| } | |
| }, | |
| "email": "[email protected]" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "We describe our contribution to the SemEval task on Frame-Semantic Structure Extraction. Unlike most previous systems described in literature, ours is based on dependency syntax. We also describe a fully automatic method to add words to the FrameNet lexical database, which gives an improvement in the recall of frame detection.", | |
| "pdf_parse": { | |
| "paper_id": "S07-1048", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "We describe our contribution to the SemEval task on Frame-Semantic Structure Extraction. Unlike most previous systems described in literature, ours is based on dependency syntax. We also describe a fully automatic method to add words to the FrameNet lexical database, which gives an improvement in the recall of frame detection.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "The existence of links between grammatical relations and various forms of semantic interpretation has long been observed; grammatical relations play a crucial role in theories of linking, i.e. the realization of the semantic arguments of predicates as syntactic units (Manning, 1994; Mel'\u010duk, 1988) . Grammatical relations may be covered by many definitions but it is probably easier to use them as an extension of dependency grammars, where relations take the form of arc labels. In addition, some linguistic phenomena such as wh-movement and discontinuous structures are conveniently described using dependency syntax by allowing nonprojective dependency arcs. It has also been claimed that dependency syntax is easier to understand and to teach to people without a linguistic background.", | |
| "cite_spans": [ | |
| { | |
| "start": 268, | |
| "end": 283, | |
| "text": "(Manning, 1994;", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 284, | |
| "end": 298, | |
| "text": "Mel'\u010duk, 1988)", | |
| "ref_id": "BIBREF8" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Despite these advantages, dependency syntax has relatively rarely been used in semantic structure extraction, with a few exceptions. Ahn et al. (2004) used a post-processing step to convert constituent trees into labeled dependency trees that were then used as input to a semantic role labeler. Pradhan et al. (2005) used a rule-based dependency parser, but the results were significantly worse than when using a constituent parser.", | |
| "cite_spans": [ | |
| { | |
| "start": 133, | |
| "end": 150, | |
| "text": "Ahn et al. (2004)", | |
| "ref_id": "BIBREF0" | |
| }, | |
| { | |
| "start": 295, | |
| "end": 316, | |
| "text": "Pradhan et al. (2005)", | |
| "ref_id": "BIBREF10" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "This paper describes a system for frame-semantic structure extraction that is based on a dependency parser. The next section presents the dependency grammar that we rely on. We then give the details on the frame detection and disambiguation, the frame element (FE) identification and classification, and dictionary extension, after which the results and conclusions are given.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "The last few years have seen an increasing interest in dependency parsing (Buchholz and Marsi, 2006) with significant improvements of the state of the art, and dependency treebanks are now available for a wide range of languages. The parsing algorithms are comparatively easy to implement and efficient: some of the algorithms parse sentences in linear time (Yamada and Matsumoto, 2003; Nivre et al., 2006) .", | |
| "cite_spans": [ | |
| { | |
| "start": 74, | |
| "end": 100, | |
| "text": "(Buchholz and Marsi, 2006)", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 358, | |
| "end": 386, | |
| "text": "(Yamada and Matsumoto, 2003;", | |
| "ref_id": "BIBREF14" | |
| }, | |
| { | |
| "start": 387, | |
| "end": 406, | |
| "text": "Nivre et al., 2006)", | |
| "ref_id": "BIBREF9" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Dependency Parsing with the Penn Treebank", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "In the semantic structure extraction system, we used the Stanford part-of-speech tagger (Toutanova et al., 2003) to tag the training and test sentences and MaltParser, a statistical dependency parser (Nivre et al., 2006) , to parse them.", | |
| "cite_spans": [ | |
| { | |
| "start": 88, | |
| "end": 112, | |
| "text": "(Toutanova et al., 2003)", | |
| "ref_id": "BIBREF11" | |
| }, | |
| { | |
| "start": 200, | |
| "end": 220, | |
| "text": "(Nivre et al., 2006)", | |
| "ref_id": "BIBREF9" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Dependency Parsing with the Penn Treebank", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "We trained the parser on the Penn Treebank (Marcus et al., 1993) . The dependency trees used to train the parser were created from the constituent trees using a conversion program (Johansson and Nugues, 2007) 1 . The converter handles most of the secondary edges in the Treebank and encodes those edges as (generally) nonprojective dependency arcs. Such information is available in the Penn Treebank in the form of empty categories and secondary edges, it is however not available in the output of traditional constituent parsers, although there have been some attempts to apply a post-processing step to predict it, see Ahn et al. (2004) , inter alia. Figures 1 and 2 show a constituent tree from the Treebank and its corresponding dependency tree. Note that the secondary edge from the wh-trace to Why is converted into a nonprojective PRP link.", | |
| "cite_spans": [ | |
| { | |
| "start": 43, | |
| "end": 64, | |
| "text": "(Marcus et al., 1993)", | |
| "ref_id": "BIBREF7" | |
| }, | |
| { | |
| "start": 180, | |
| "end": 208, | |
| "text": "(Johansson and Nugues, 2007)", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 621, | |
| "end": 638, | |
| "text": "Ahn et al. (2004)", | |
| "ref_id": "BIBREF0" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 653, | |
| "end": 668, | |
| "text": "Figures 1 and 2", | |
| "ref_id": "FIGREF0" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Dependency Parsing with the Penn Treebank", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "This section describes how the dependency trees are used to create the semantic structure is divided into two main components: frame detection and disambiguation, and frame element detection and classification.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Semantic Structure Extraction", | |
| "sec_num": "3" | |
| }, | |
| { | |
| "text": "Since many potential target words appear in senses that should not be tagged with a frame, we use a filtering component as a first step in the frame detection. We also removed some words (especially prepositions) that caused significant performance degradation because of lack of training data. With the increasing availability of tagged running text, we expect that we will be able to replace the filtering rules with a classifier in the future.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 have was retained only if it had an object,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 be only if it was preceded by there,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 will was removed in its modal sense,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 of course and in particular were removed,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 the prepositions above, against, at, below, beside, by, in, on, over, and under were removed unless their head was marked as locative,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 after and before were removed unless their head was marked as temporal,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 into, to, and through were removed unless their head was marked as direction,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 as, for, so, and with were always removed,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "\u2022 since the only sense of of was PARTITIVE, we removed it unless it was preceded by only, member, one, most, many, some, few, part, majority, minority, proportion, half, third, quarter, all, or none, or if it was followed by all, group, them, or us.", | |
| "cite_spans": [ | |
| { | |
| "start": 84, | |
| "end": 190, | |
| "text": "only, member, one, most, many, some, few, part, majority, minority, proportion, half, third, quarter, all,", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "We also removed all targets that had been tagged as support verbs for some other target.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Filtering Rules", | |
| "sec_num": "3.1.1" | |
| }, | |
| { | |
| "text": "For the target words left after the filtering, we used a classifier to assign a frame, following Erk (2005) . We trained a disambiguating SVM classifier on all ambiguous words listed in FrameNet. Its accuracy was 84% on the ambiguous words, compared to a first-sense baseline score of 74%.", | |
| "cite_spans": [ | |
| { | |
| "start": 97, | |
| "end": 107, | |
| "text": "Erk (2005)", | |
| "ref_id": "BIBREF2" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Sense Disambiguation", | |
| "sec_num": "3.1.2" | |
| }, | |
| { | |
| "text": "The classifier used the following features: target lemma, target word, subcategorization frame (for verb targets only), the set of dependencies of the target, the set of words of the child nodes, and the parent word of the target.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Sense Disambiguation", | |
| "sec_num": "3.1.2" | |
| }, | |
| { | |
| "text": "The subcategorization frame feature was formed by concatenating the dependency labels of the children, excluding subject, parentheticals, punctuation and coordinations. For instance, for kidnap in Figure 2, the feature is PRP+OBJ.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 197, | |
| "end": 203, | |
| "text": "Figure", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Sense Disambiguation", | |
| "sec_num": "3.1.2" | |
| }, | |
| { | |
| "text": "Coverage is one of the main weaknesses of the current FrameNet lexical database -it lists only 10,197 lexical units, compared to 207,016 wordsense pairs in WordNet 3.0 (Fellbaum, 1998). We tried to remedy this problem by training classifiers to find words that are related to the words in a frame.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Extending the Lexical Database", | |
| "sec_num": "3.1.3" | |
| }, | |
| { | |
| "text": "We designed a feature representation for each lemma in WordNet, which uses a sequence of identifiers for each synset in its hypernym tree. All senses of the lemma were used, and the features were weighted with respect to the relative frequency of the sense. Using this feature representation, we trained an SVM classifier for each frame that tells whether a lemma belongs to that frame or not.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Extending the Lexical Database", | |
| "sec_num": "3.1.3" | |
| }, | |
| { | |
| "text": "The FrameNet dictionary could thus be extended by 18,372 lexical units. If we assume a Zipf distribution and that the lexical units already in FrameNet are the most common ones, this would increase the coverage by up to 9%. In the test set, the new lexical units account for 53 out of the 808 target words our system detected (6.5%). We roughly estimated the precision to 70% by manually inspecting 100 randomly selected words in the extended dictionary. This strategy is most successful when the frame is equivalent to one or a few synsets (and their subtrees). For instance, for the frame MEDI-CAL_CONDITION, we can add the complete subtree of the synset pathological state, resulting in 641 new lemmas referring to all sorts of diseases. On the other hand, the strategy also works well for motion verbs (which often exhibit complex patterns of polysemy): 137 lemmas could be added to the SELF_MOTION frame. Examples of frames with frequent errors are LEADERSHIP, which includes many insects (probably because the most frequent sense of queen in SemCor is the queen bee), and FOOD, which included many chemical substances as well as inedible plants and animals.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Extending the Lexical Database", | |
| "sec_num": "3.1.3" | |
| }, | |
| { | |
| "text": "Following convention, we divided the FE extraction into two subtasks: argument identification and argument classification. We did not try to assign multiple labels to arguments. Figure 3 shows an overview. In addition to detecing the FEs, the argument identification classifier detects the dependency nodes that should be tagged on the layers other than the frame element layer: SUPP, COP, NULL, EXIST, and ASP. The ANT and REL labels could be inserted using simple rules. Similarly to Xue and Palmer (2004) we could filter away many nodes before the argument identification step by assuming that the arguments for a given predicate correspond to a subset of the dependents of the target or of its transitive heads.", | |
| "cite_spans": [ | |
| { | |
| "start": 486, | |
| "end": 507, | |
| "text": "Xue and Palmer (2004)", | |
| "ref_id": "BIBREF13" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 178, | |
| "end": 186, | |
| "text": "Figure 3", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Frame Element Extraction", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "Both classifiers were implemented using SVMs and use the following features: target lemma, voice (for verb targets only), subcategorization frame (for verb targets only), the set of dependencies of the target, part of speech of the target node, path through the dependency tree from the target to the node, position (before, after, or on), word and part of speech for the head, word and part of speech for leftmost and rightmost descendent.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Frame Element Extraction", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "In the path feature, we removed steps through verb chains and coordination. For instance, in the sentece I have seen and heard it, the path from heard to I is only SBJ\u2193 and to it OBJ\u2193.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Frame Element Extraction", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "In addition to the frame-semantic information, the SemEval task also scores named entities. We used YamCha (Kudo and Matsumoto, 2003) to detect named entities, and we trained it on the SemEval full-text training sets. Apart from the word and part of speech, we used suffixes up to length 5 as features. We think that results could be improved further by using an external NE tagger.", | |
| "cite_spans": [ | |
| { | |
| "start": 107, | |
| "end": 133, | |
| "text": "(Kudo and Matsumoto, 2003)", | |
| "ref_id": "BIBREF5" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Named Entity Recognition", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "The system was evaluated on three texts. Table 1 shows the results for frame detection averaged over the test texts. In the Setting colums, the first shows whether Exact or Partial frame matching was used by the evaluation script, and the second whether Labels or Dependencies were used. Table 2 compares the results of the system using the extended dictionary with one using the orignal FrameNet dictionary, using the Partial matching and Labels scoring. The extended dictionary introduces some noise and thus lowers the precision slightly, but the effects on the recall are positive. Table 3 shows the aver- aged precision, recall, and F 1 measures for different evaluation parameters. The third column shows whether named entities were used (Y) or not (N). Interestingly, the scores are higher for the semantic dependency graphs than for flat labels, while the two other teams generally had higher scores for flat labels. We believe that the reason for this is that we used a dependency parser, and that the rules that we used to convert dependency nodes into spans may have produced some errors. It is possible that the figures would have been slightly higher if our program produced semantic dependency graphs directly. ", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 41, | |
| "end": 48, | |
| "text": "Table 1", | |
| "ref_id": "TABREF0" | |
| }, | |
| { | |
| "start": 288, | |
| "end": 295, | |
| "text": "Table 2", | |
| "ref_id": "TABREF1" | |
| }, | |
| { | |
| "start": 586, | |
| "end": 593, | |
| "text": "Table 3", | |
| "ref_id": "TABREF2" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Results", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "We have presented a system for frame-semantic structure extraction that achieves promising results. While most previous systems have been based on constituents, our system relies on a dependency parser. We also described an automatic method to add new units to the FrameNet lexical database.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion and Future Work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "To improve labeling quality, we would like to apply constraints to the semantic output so that semantic type and coreness rules are obeyed. In addition, while the system described here is based on pipelined classification, recent research on semantic role labeling has shown that significant performance improvements can be gained by exploiting interdependencies between arguments (Toutanova et al., 2005) . With an increasing amount of running text annotated with frame semantics, we believe that this insight can be extended to model interdependencies between frames as well.", | |
| "cite_spans": [ | |
| { | |
| "start": 381, | |
| "end": 405, | |
| "text": "(Toutanova et al., 2005)", | |
| "ref_id": "BIBREF12" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion and Future Work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Our motivation for using dependency grammar is that we hope that it will eventually make semantic structure extraction easier to implement and more theoretically well-founded. How to best design the dependency syntax is also still an open question.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion and Future Work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Ideally, all arguments would be direct dependents of the predicate node and we could get rid of the sparse and brittle Path feature in the classifier.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion and Future Work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Available at http://nlp.cs.lth.se/pennconverter", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "The university of Amsterdam at Senseval-3: Semantic roles and logic forms", | |
| "authors": [ | |
| { | |
| "first": "David", | |
| "middle": [], | |
| "last": "Ahn", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Sisay", | |
| "middle": [], | |
| "last": "Fissaha", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Valentin", | |
| "middle": [], | |
| "last": "Jijkoun", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Maarten", | |
| "middle": [], | |
| "last": "De Rijke", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Proceedings of SENSEVAL-3", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "David Ahn, Sisay Fissaha, Valentin Jijkoun, and Maarten de Rijke. 2004. The university of Amsterdam at Senseval-3: Semantic roles and logic forms. In Pro- ceedings of SENSEVAL-3.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "CoNLL-X shared task on multilingual dependency parsing", | |
| "authors": [ | |
| { | |
| "first": "Sabine", | |
| "middle": [], | |
| "last": "Buchholz", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Erwin", | |
| "middle": [], | |
| "last": "Marsi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of the CoNLL-X", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sabine Buchholz and Erwin Marsi. 2006. CoNLL-X shared task on multilingual dependency parsing. In Proceedings of the CoNLL-X.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Frame assignment as word sense disambiguation", | |
| "authors": [ | |
| { | |
| "first": "Katrin", | |
| "middle": [], | |
| "last": "Erk", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of IWCS 6", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Katrin Erk. 2005. Frame assignment as word sense dis- ambiguation. In Proceedings of IWCS 6.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "WordNet: An electronic lexical database", | |
| "authors": [], | |
| "year": 1998, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Christiane Fellbaum, editor. 1998. WordNet: An elec- tronic lexical database. MIT Press.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Extended constituent-to-dependency conversion for English", | |
| "authors": [ | |
| { | |
| "first": "Richard", | |
| "middle": [], | |
| "last": "Johansson", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Pierre", | |
| "middle": [], | |
| "last": "Nugues", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "Proceedings of NODALIDA 2007", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Richard Johansson and Pierre Nugues. 2007. Extended constituent-to-dependency conversion for English. In Proceedings of NODALIDA 2007. To appear.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Fast methods for kernel-based text analysis", | |
| "authors": [ | |
| { | |
| "first": "Taku", | |
| "middle": [], | |
| "last": "Kudo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yuji", | |
| "middle": [], | |
| "last": "Matsumoto", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Taku Kudo and Yuji Matsumoto. 2003. Fast methods for kernel-based text analysis. In ACL-2003.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Ergativity: Argument structure and grammatical relations", | |
| "authors": [ | |
| { | |
| "first": "Chistopher", | |
| "middle": [], | |
| "last": "Manning", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1994, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chistopher Manning. 1994. Ergativity: Argument struc- ture and grammatical relations.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Building a large annotated corpus of English: the Penn Treebank", | |
| "authors": [ | |
| { | |
| "first": "Mitchell", | |
| "middle": [ | |
| "P" | |
| ], | |
| "last": "Marcus", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Beatrice", | |
| "middle": [], | |
| "last": "Santorini", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Mary", | |
| "middle": [ | |
| "Ann" | |
| ], | |
| "last": "Marcinkiewicz", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "Computational Linguistics", | |
| "volume": "19", | |
| "issue": "2", | |
| "pages": "313--330", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mitchell P. Marcus, Beatrice Santorini, and Mary Ann Marcinkiewicz. 1993. Building a large annotated cor- pus of English: the Penn Treebank. Computational Linguistics, 19(2):313-330.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Dependency Syntax: Theory and Practice", | |
| "authors": [ | |
| { | |
| "first": "Igor", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Igor A. Mel'\u010duk. 1988. Dependency Syntax: Theory and Practice. State University Press of New York, Albany.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Malt-Parser: A data-driven parser generator for dependency parsing", | |
| "authors": [ | |
| { | |
| "first": "Joakim", | |
| "middle": [], | |
| "last": "Nivre", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Johan", | |
| "middle": [], | |
| "last": "Hall", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jens", | |
| "middle": [], | |
| "last": "Nilsson", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of LREC", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Joakim Nivre, Johan Hall, and Jens Nilsson. 2006. Malt- Parser: A data-driven parser generator for dependency parsing. In Proceedings of LREC.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Semantic role labeling using different syntactic views", | |
| "authors": [ | |
| { | |
| "first": "Sameer", | |
| "middle": [], | |
| "last": "Pradhan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Wayne", | |
| "middle": [], | |
| "last": "Ward", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kadri", | |
| "middle": [], | |
| "last": "Hacioglu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "James", | |
| "middle": [], | |
| "last": "Martin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Daniel", | |
| "middle": [], | |
| "last": "Jurafsky", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "ACL-2005", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sameer Pradhan, Wayne Ward, Kadri Hacioglu, James Martin, and Daniel Jurafsky. 2005. Semantic role la- beling using different syntactic views. In ACL-2005.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Feature-rich part-of-speech tagging with a cyclic dependency network", | |
| "authors": [ | |
| { | |
| "first": "Kristina", | |
| "middle": [], | |
| "last": "Toutanova", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dan", | |
| "middle": [], | |
| "last": "Klein", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Christopher", | |
| "middle": [], | |
| "last": "Manning", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yoram", | |
| "middle": [], | |
| "last": "Singer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "Proceedings of HLT-NAACL", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kristina Toutanova, Dan Klein, Christopher Manning, and Yoram Singer. 2003. Feature-rich part-of-speech tagging with a cyclic dependency network. In Pro- ceedings of HLT-NAACL.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Joint learning improves semantic role labeling", | |
| "authors": [ | |
| { | |
| "first": "Kristina", | |
| "middle": [], | |
| "last": "Toutanova", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Aria", | |
| "middle": [], | |
| "last": "Haghighi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Christopher", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Manning", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of ACL 2005", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kristina Toutanova, Aria Haghighi, and Christopher D. Manning. 2005. Joint learning improves semantic role labeling. In Proceedings of ACL 2005.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Calibrating features for semantic role labeling", | |
| "authors": [ | |
| { | |
| "first": "Nianwen", | |
| "middle": [], | |
| "last": "Xue", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Martha", | |
| "middle": [], | |
| "last": "Palmer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Proc. of EMNLP", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Nianwen Xue and Martha Palmer. 2004. Calibrating fea- tures for semantic role labeling. In Proc. of EMNLP.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Statistical dependency analysis with support vector machines", | |
| "authors": [ | |
| { | |
| "first": "Hiroyasu", | |
| "middle": [], | |
| "last": "Yamada", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yuji", | |
| "middle": [], | |
| "last": "Matsumoto", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2003, | |
| "venue": "Proceedings of IWPT-03", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Hiroyasu Yamada and Yuji Matsumoto. 2003. Statistical dependency analysis with support vector machines. In Proceedings of IWPT-03.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "num": null, | |
| "type_str": "figure", | |
| "text": "Why would intelligent beings kidnap seven Soviet mailmen *T* ?Figure 1: A constituent tree from the Penn Treebank.Why would intelligent beings kidnap seven Soviet mailmen ? Converted dependency tree." | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "num": null, | |
| "type_str": "figure", | |
| "text": "Figure 3: FE extraction steps." | |
| }, | |
| "TABREF0": { | |
| "content": "<table><tr><td colspan=\"4\">Setting Recall Precision</td><td>F 1</td></tr><tr><td>E</td><td>L</td><td>0.528</td><td>0.688</td><td>0.597</td></tr><tr><td>P</td><td>L</td><td>0.581</td><td>0.758</td><td>0.657</td></tr><tr><td>E</td><td>D</td><td>0.549</td><td>0.715</td><td>0.621</td></tr><tr><td>P</td><td>D</td><td>0.601</td><td>0.784</td><td>0.681</td></tr></table>", | |
| "type_str": "table", | |
| "num": null, | |
| "html": null, | |
| "text": "Results for frame detection." | |
| }, | |
| "TABREF1": { | |
| "content": "<table><tr><td colspan=\"3\">Dictionary Recall Precision</td><td>F 1</td></tr><tr><td>Original</td><td>0.550</td><td>0.767</td><td>0.634</td></tr><tr><td>Extended</td><td>0.581</td><td>0.758</td><td>0.657</td></tr></table>", | |
| "type_str": "table", | |
| "num": null, | |
| "html": null, | |
| "text": "Comparison of dictionaries." | |
| }, | |
| "TABREF2": { | |
| "content": "<table><tr><td>Setting</td><td colspan=\"2\">Recall Precision</td><td>F 1</td></tr><tr><td>E L Y</td><td>0.372</td><td>0.532</td><td>0.438</td></tr><tr><td>P L Y</td><td>0.398</td><td>0.570</td><td>0.468</td></tr><tr><td>E D Y</td><td>0.389</td><td>0.557</td><td>0.458</td></tr><tr><td>P D Y</td><td>0.414</td><td>0.594</td><td>0.488</td></tr><tr><td>E L N</td><td>0.364</td><td>0.530</td><td>0.432</td></tr><tr><td>P L N</td><td>0.391</td><td>0.570</td><td>0.464</td></tr><tr><td>E D N</td><td>0.384</td><td>0.561</td><td>0.456</td></tr><tr><td>P D N</td><td>0.411</td><td>0.600</td><td>0.488</td></tr></table>", | |
| "type_str": "table", | |
| "num": null, | |
| "html": null, | |
| "text": "Results for frame and FE detection." | |
| } | |
| } | |
| } | |
| } |