Entity-Based Local Coherence Modelling Using Topological Fields
Cheung, Jackie Chi Kit and Penn, Gerald

Article Structure

Abstract

One goal of natural language generation is to produce coherent text that presents information in a logical order.

Introduction

One type of coherence modelling that has captured recent research interest is local coherence modelling, which measures the coherence of a document by examining the similarity between neigh-bouring text spans.

Topics

coreference

Appears in 14 sentences as: coreference (15) coreferential (1) coreferents (1)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. The authors remark that extracted sentences with VFs that are referentially related to previous context (e. g., they contain a coreferential noun phrase or a discourse relation like “therefore”) are reinserted at higher accuracies.
    Page 2, “Introduction”
  2. The main focus of that work, however, was to adapt the model for use in a low-resource situation when perfect coreference information is not available.
    Page 4, “Introduction”
  3. Table 3: Accuracy of automatic annotations of noun phrases with coreferents .
    Page 6, “Introduction”
  4. These results show the upper bound of entity-based local coherence modelling with perfect coreference information.
    Page 7, “Introduction”
  5. The results we obtain are higher than the results for the English corpora of Barzilay and Lapata (2008) (87.2% on the Earthquakes corpus and 90.4% on the Accidents corpus), but this is probably due to corpus differences as well as the availability of perfect coreference information in our experimentsl.
    Page 7, “Introduction”
  6. In our experiments, we create an entity for every single noun phrase node that we encounter, then merge the entities that are linked by coreference .
    Page 7, “Introduction”
  7. 1Barzilay and Lapata (2008) use the coreference system of Ng and Cardie (2002) to obtain coreference annotations.
    Page 7, “Introduction”
  8. We are not aware of similarly well-tested, publicly available coreference resolution systems that handle all types of anaphora for German.
    Page 7, “Introduction”
  9. We considered adapting the BART coreference resolution toolkit (Versley et a1., 2008) to German, but a number of language-dependent decisions regarding preprocessing, feature engineering, and the learning paradigm would need to be made in order to achieve reasonable performance comparable to state-of-the—art English coreference resolution systems.
    Page 7, “Introduction”
  10. Each article also contains the coreference chain of the subject of the biography (the biographee).
    Page 8, “Introduction”
  11. Because this corpus does not come with general coreference information except for the coreference chain of the biographee, we use the semantic classes instead.
    Page 8, “Introduction”

See all papers in Proc. ACL 2010 that mention coreference.

See all papers in Proc. ACL that mention coreference.

Back to top.

noun phrases

Appears in 14 sentences as: noun phrases (15)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. The NF (Nachfeld or “post-field”) contains prosodically heavy elements such as postposed prepositional phrases or relative clauses, and occasionally postposed noun phrases .
    Page 2, “Introduction”
  2. The model of Elsner and Charniak (2007) uses syntactic cues to model the discourse-newness of noun phrases .
    Page 4, “Introduction”
  3. Since noun phrases can be embedded in other noun phrases , overlaps can occur.
    Page 5, “Introduction”
  4. In this case, the dominating noun phrase takes the smallest order number among its dominated noun phrases .
    Page 5, “Introduction”
  5. cases distinguishes five types of entities corresponding to the four morphological cases of German in addition to another category for noun phrases which are not complements of the main verb.
    Page 5, “Introduction”
  6. t opf / pp distinguishes entities in the topological fields VF, MF, and NF, contains a separate category for PP, and a category for all other noun phrases .
    Page 5, “Introduction”
  7. Prepositional objects are treated the same as other noun phrases here.
    Page 5, “Introduction”
  8. Other noun phrases in MF are marked as NONTOPIC.
    Page 5, “Introduction”
  9. Categories for NF and miscellaneous noun phrases also exist.
    Page 5, “Introduction”
  10. topic+pron refines the above by taking into account a word order restriction in German that pronouns appear before full noun phrases in the MF field.
    Page 5, “Introduction”
  11. Morphological case is distinct from grammatical role, as noun phrases can function as adjuncts in possessive constructions and preposi-
    Page 6, “Introduction”

See all papers in Proc. ACL 2010 that mention noun phrases.

See all papers in Proc. ACL that mention noun phrases.

Back to top.

word order

Appears in 11 sentences as: word order (12)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. Previous work on English, a language with relatively fixed word order , has identified factors that contribute to local coherence, such as the grammatical roles associated with the entities.
    Page 1, “Introduction”
  2. For instance, freer-word-order languages exhibit word order patterns which are dependent on discourse factors relating to information structure, in addition to the grammatical roles of nominal arguments of the main verb.
    Page 1, “Introduction”
  3. We thus expect word order information to be particularly important in these languages in discourse analysis, which includes coherence modelling.
    Page 1, “Introduction”
  4. They apply their model to pronominal anaphora resolution, identifying potential antecedents of subsequent anaphora by considering syntactic and word order information, classifying constituents by their familiarity to the reader.
    Page 1, “Introduction”
  5. They find that their approach correctly resolves more pronominal anaphora than a grammatical role-based approach which ignores word order , and the difference between the two approaches is larger in German corpora than in English ones.
    Page 1, “Introduction”
  6. Topological fields are a useful abstraction of word order, because while Germanic word order is relatively free with respect to grammatical functions, the order of the topological fields is strict and unvarying.
    Page 2, “Introduction”
  7. The main goal of this study is to compare word order , grammatical role and topological field information, which is encoded into the entity grid at each occurrence of an entity.
    Page 4, “Introduction”
  8. The correlation between word order and information structure is well known, and has formed the basis of some theories of syntax such as the Prague School’s (Sgall et al., 1986).
    Page 5, “Introduction”
  9. topic+pron refines the above by taking into account a word order restriction in German that pronouns appear before full noun phrases in the MF field.
    Page 5, “Introduction”
  10. We suggest that the utility of topological fields in local coherence modelling comes from the interaction between word order and information structure in freer-word-order languages.
    Page 9, “Introduction”
  11. Crucially, topological fields take into account issues such as coordination, appositives, sentential fragments and differences in clause types, which word order alone does not.
    Page 9, “Introduction”

See all papers in Proc. ACL 2010 that mention word order.

See all papers in Proc. ACL that mention word order.

Back to top.

natural language

Appears in 9 sentences as: Natural Language (1) natural language (8)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. One goal of natural language generation is to produce coherent text that presents information in a logical order.
    Page 1, “Abstract”
  2. Then, we incorporate the model enhanced with topological fields into a natural language generation system that generates constituent orders for German text, and show that the added coherence component improves performance slightly, though not statistically significantly.
    Page 1, “Abstract”
  3. Local coherence modelling has been shown to be useful for tasks like natural language generation and summarization, (Barzilay and Lee, 2004) and genre classification (Barzilay and Lapata, 2008).
    Page 1, “Introduction”
  4. We then embed these topological field annotations into a natural language generation system to show the utility of local coherence information in an applied setting.
    Page 2, “Introduction”
  5. Filippova and Strube (2007c) also examine the role of the VF in local coherence and natural language generation, focusing on the correlation between VFs and sentential topics.
    Page 3, “Introduction”
  6. This is particularly useful in natural language understanding tasks.
    Page 4, “Introduction”
  7. In contrast, our work focuses on improving performance by annotating entities with additional linguistic information, such as topological fields, and is geared towards natural language generation systems where perfect information is available.
    Page 4, “Introduction”
  8. 4 Local Coherence for Natural Language Generation
    Page 7, “Introduction”
  9. Table 5: Results of adding coherence features into a natural language generation system.
    Page 9, “Introduction”

See all papers in Proc. ACL 2010 that mention natural language.

See all papers in Proc. ACL that mention natural language.

Back to top.

statistically significant

Appears in 6 sentences as: statistically significant (5) statistically significantly (2)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. Then, we incorporate the model enhanced with topological fields into a natural language generation system that generates constituent orders for German text, and show that the added coherence component improves performance slightly, though not statistically significantly .
    Page 1, “Abstract”
  2. We add contextual features using topological field transitions to the model of Filippova and Strube (2007b) and achieve a slight improvement over their model in a constituent ordering task, though not statistically significantly .
    Page 2, “Introduction”
  3. Two-tailed sign tests were calculated for each result against the best performing model in each column (1: p = 0.101; 2: p = 0.053; +: statistically significant, p < 0.05; ++: very statistically significant , p < 0.01 ).
    Page 6, “Introduction”
  4. We embed entity topological field transitions into their probabilistic model, and show that the added coherence component slightly improves the performance of the baseline NLG system in generating constituent orderings in a German corpus, though not to a statistically significant degree.
    Page 8, “Introduction”
  5. These improvements are not statistically significant by Mc-Nemar’s test.
    Page 9, “Introduction”
  6. We find that the coherence-enhanced model slightly outperforms the baseline system, but this was not statistically significant .
    Page 9, “Introduction”

See all papers in Proc. ACL 2010 that mention statistically significant.

See all papers in Proc. ACL that mention statistically significant.

Back to top.

maximum entropy

Appears in 5 sentences as: maximum entropy (5)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. This is done using a maximum entropy model (call it MAXENT).
    Page 8, “Introduction”
  2. Then, the remaining constituents are ordered using a second maximum entropy model (MAXENTZ).
    Page 8, “Introduction”
  3. The maximum entropy model for both steps rely on the following features:
    Page 8, “Introduction”
  4. All of the features but one in the original maximum entropy models rely on local properties of the clause.
    Page 8, “Introduction”
  5. Table 5 shows the results of adding these contextual features into the maximum entropy models.
    Page 9, “Introduction”

See all papers in Proc. ACL 2010 that mention maximum entropy.

See all papers in Proc. ACL that mention maximum entropy.

Back to top.

development set

Appears in 4 sentences as: development set (4)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. There are 216 documents and 4126 original-permutation pairs in the training set, and 24 documents and 465 pairs in the development set .
    Page 4, “Introduction”
  2. Transition length, salience, and a regularization parameter are tuned on the development set .
    Page 4, “Introduction”
  3. We only report results using the setting of transition length g 4, and no salience threshold, because they give the best performance on the development set .
    Page 4, “Introduction”
  4. We found that looking back up to two sentences produces the best results (by tuning on the development set ).
    Page 8, “Introduction”

See all papers in Proc. ACL 2010 that mention development set.

See all papers in Proc. ACL that mention development set.

Back to top.

coreference resolution

Appears in 3 sentences as: coreference resolution (4)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. We are not aware of similarly well-tested, publicly available coreference resolution systems that handle all types of anaphora for German.
    Page 7, “Introduction”
  2. We considered adapting the BART coreference resolution toolkit (Versley et a1., 2008) to German, but a number of language-dependent decisions regarding preprocessing, feature engineering, and the learning paradigm would need to be made in order to achieve reasonable performance comparable to state-of-the—art English coreference resolution systems.
    Page 7, “Introduction”
  3. The model also shows promise for other discourse-related tasks such as coreference resolution and discourse parsing.
    Page 9, “Introduction”

See all papers in Proc. ACL 2010 that mention coreference resolution.

See all papers in Proc. ACL that mention coreference resolution.

Back to top.

feature vector

Appears in 3 sentences as: feature vector (3)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. We tabulate the transitions of entities between different syntactic positions (or their nonoccurrence) in sentences, and convert the frequencies of transitions into a feature vector representation of transition probabilities in the document.
    Page 4, “Introduction”
  2. We solve this problem in a supervised machine learning setting, where the input is the feature vector representations of the two versions of the document, and the output is a binary value indicating the document with the original sentence ordering.
    Page 4, “Introduction”
  3. Transition length — the maximum length of the transitions used in the feature vector representation of a document.
    Page 4, “Introduction”

See all papers in Proc. ACL 2010 that mention feature vector.

See all papers in Proc. ACL that mention feature vector.

Back to top.

manual annotations

Appears in 3 sentences as: manual annotation (1) manual annotations (2)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. First, we show in a sentence ordering experiment that topological field information improves the entity grid model of Barzilay and Lapata (2008) more than grammatical role and simple clausal order information do, particularly when manual annotations of this information are not available.
    Page 1, “Abstract”
  2. sentations to automatic extraction in the absence of manual annotations .
    Page 6, “Introduction”
  3. Note, however, that the models based on automatic topological field annotations outperform even the grammatical role-based models using manual annotation (at marginal significance, p < 0.1).
    Page 6, “Introduction”

See all papers in Proc. ACL 2010 that mention manual annotations.

See all papers in Proc. ACL that mention manual annotations.

Back to top.

vector representation

Appears in 3 sentences as: vector representation (2) vector representations (1)
In Entity-Based Local Coherence Modelling Using Topological Fields
  1. We tabulate the transitions of entities between different syntactic positions (or their nonoccurrence) in sentences, and convert the frequencies of transitions into a feature vector representation of transition probabilities in the document.
    Page 4, “Introduction”
  2. We solve this problem in a supervised machine learning setting, where the input is the feature vector representations of the two versions of the document, and the output is a binary value indicating the document with the original sentence ordering.
    Page 4, “Introduction”
  3. Transition length — the maximum length of the transitions used in the feature vector representation of a document.
    Page 4, “Introduction”

See all papers in Proc. ACL 2010 that mention vector representation.

See all papers in Proc. ACL that mention vector representation.

Back to top.

PDF available from www.aclweb.org.
Send us an email