Article Structure
Abstract
Negative expressions are common in natural language text and play a critical role in information extraction.
Introduction
Negation is a grammatical category which comprises various kinds of devices to reverse the truth value of a proposition (Morante and Sporleder, 2012).
Related Work
Earlier studies of negation were almost in linguistics (e.g.
Baselines
Negation focus identification in *SEM’2012 shared tasks is restricted to verbal negations annotated with MNEG in PropBank, with only the constituent belonging to a semantic role selected as negation focus.
Topics
graph model
Appears in 35 sentences as: Graph Model (3) graph model (32) Graph Model: (1) Graph models (1) graph models (1)
In Negation Focus Identification with Contextual Discourse Information
- In this paper, we propose a graph model to enrich intra-sentence features with inter-sentence features from both lexical and topic perspectives.
Page 1, “Abstract”
- Evaluation on the *SEM 2012 shared task corpus indicates the usefulness of contextual discourse information in negation focus identification and justifies the effectiveness of our graph model in capturing such global information.
Page 1, “Abstract”
- In this paper, to well accommodate such contextual discourse information in negation focus identification, we propose a graph model to enrich normal intra—sentence features with various kinds of inter-sentence features from both lexical and topic perspectives.
Page 2, “Introduction”
- Besides, the standard PageRank algorithm is employed to optimize the graph model .
Page 2, “Introduction”
- Section 4 introduces our topic-driven word-based graph model with contextual discourse information.
Page 2, “Introduction”
- In this paper, we first propose a graph model to gauge the importance of contextual discourse
Page 3, “Baselines”
- 4.1 Graph Model
Page 4, “Baselines”
- Graph models have been proven successful in many NLP applications, especially in representing the link relationships between words or sentences (Wan and Yang, 2008; Li et al., 2009).
Page 4, “Baselines”
- In this paper, we propose a graph model to represent the contextual discourse information from both lexical and topic perspectives.
Page 4, “Baselines”
- In particular, a word-based graph model is proposed to represent the explicit relatedness among words in a discourse from the lexical perspective, while a topic-driven word-based model is proposed to enrich the implicit relatedness between words, by adding one more layer to the word-based graph model in representing the global topic distribution of the whole dataset.
Page 4, “Baselines”
- Besides, the PageRank algorithm (Page et al., 1998) is adopted to optimize the graph model .
Page 4, “Baselines”
See all papers in Proc. ACL 2014 that mention graph model.
See all papers in Proc. ACL that mention graph model.
Back to top.
shared task
Appears in 15 sentences as: shared task (13) shared tasks (3)
In Negation Focus Identification with Contextual Discourse Information
- Evaluation on the *SEM 2012 shared task corpus indicates the usefulness of contextual discourse information in negation focus identification and justifies the effectiveness of our graph model in capturing such global information.
Page 1, “Abstract”
- Evaluation on the *SEM 2012 shared task corpus (Morante and Blanco, 2012) justifies our approach over several strong baselines.
Page 2, “Introduction”
- Due to the increasing demand on deep understanding of natural language text, negation recognition has been drawing more and more attention in recent years, with a series of shared tasks and workshops, however, with focus on cue detection and scope resolution, such as the BioNLP 2009 shared task for negative event detection (Kim et al., 2009) and the ACL 2010 Workshop for scope resolution of negation and speculation (Morante and Sporleder, 2010), followed by a special issue of Computational Linguistics (Morante and Sporleder, 2012) for modality and negation.
Page 2, “Related Work”
- However, although Morante and Blanco (2012) proposed negation focus identification as one of the *SEM’2012 shared tasks , only one team (Rosenberg and Bergler, 2012)1 participated in this task.
Page 2, “Related Work”
- Negation focus identification in *SEM’2012 shared tasks is restricted to verbal negations annotated with MNEG in PropBank, with only the constituent belonging to a semantic role selected as negation focus.
Page 2, “Baselines”
- 1 In *SEM’2013, the shared task is changed with focus on "Semantic Textual Similarity".
Page 2, “Baselines”
- For better illustration of the importance of contextual discourse information, Table 1 shows the statistics of intra- and inter-sentence information necessary for manual negation focus identification with 100 instances randomly extracted from the held-out dataset of *SEM'2012 shared task corpus.
Page 3, “Baselines”
- Here, the topics are extracted from all the documents in the *SEM 2012 shared task using the LDA Gibbs Sampling algorithm (Griffiths, 2002).
Page 5, “Baselines”
- In all our experiments, we employ the *SEM'2012 shared task corpus (Morante and Blanco, 2012)2.
Page 5, “Baselines”
- As a freely downloadable resource, the *SEM shared task corpus is annotated on top of PropBank, which uses the WSJ section of the Penn TreeBank.
Page 5, “Baselines”
- For fair comparison, we adopt the same partition as *SEM’2012 shared task in all our experiments, i.e., with 2,302 for training, 530 for development, and 712 for testing.
Page 6, “Baselines”
See all papers in Proc. ACL 2014 that mention shared task.
See all papers in Proc. ACL that mention shared task.
Back to top.
PageRank
Appears in 7 sentences as: PageRank (7)
In Negation Focus Identification with Contextual Discourse Information
- Besides, the standard PageRank algorithm is employed to optimize the graph model.
Page 2, “Introduction”
- Besides, the PageRank algorithm (Page et al., 1998) is adopted to optimize the graph model.
Page 4, “Baselines”
- Finally, the weights of word nodes are calculated using the PageRank algorithm as follows:
Page 4, “Baselines”
- where d is the damping factor as in the PageRank algorithm.
Page 4, “Baselines”
- Finally, the weights of word nodes are calculated using the PageRank algorithm as follows:
Page 5, “Baselines”
- where d is the damping factor as in the PageRank algorithm.
Page 5, “Baselines”
- Given the graph models and the PageRank optimization algorithm discussed above, four kinds of contextual discourse information are extracted as inter-sentence features (Table 2).
Page 5, “Baselines”
See all papers in Proc. ACL 2014 that mention PageRank.
See all papers in Proc. ACL that mention PageRank.
Back to top.
content words
Appears in 6 sentences as: content words (6)
In Negation Focus Identification with Contextual Discourse Information
- Since such correlation is more from the semantic perspective than the grammatical perspective, only content words are considered in our graph model, ignoring functional words (e.g., the, t0,.
Page 4, “Baselines”
- Especially, the content words limited to those with part-of-
Page 4, “Baselines”
- While the above word-based graph model can well capture the relatedness between content words , it can only partially model the focus of a negation eXpression since negation focus is more directly related with topic than content.
Page 4, “Baselines”
- In the topic-driven word-based graph model, the first layer denotes the relatedness among content words as captured in the above word-based graph model, and the second layer denotes the topic distribution, with the dashed lines between these two layers indicating the word-topic model return by LDA.
Page 5, “Baselines”
- where i represents the content words in the focus candidate.
Page 5, “Baselines”
- These two kinds of weights focus on different aspects about the focus candidate with the former on the contribution of content words , which is more beneficial for a long focus candidate, and the latter biased towards the focus candidate which contains some critical word in a discourse.
Page 5, “Baselines”
See all papers in Proc. ACL 2014 that mention content words.
See all papers in Proc. ACL that mention content words.
Back to top.
semantic role
Appears in 6 sentences as: Semantic Role (1) semantic role (8)
In Negation Focus Identification with Contextual Discourse Information
- Current studies (e.g., Blanco and Moldovan, 2011; Rosenberg and Bergler, 2012) sort to various kinds of intra-sentence information, such as lexical features, syntactic features, semantic role features and so on, ignoring less-obvious inter-sentence information.
Page 1, “Introduction”
- Negation focus identification in *SEM’2012 shared tasks is restricted to verbal negations annotated with MNEG in PropBank, with only the constituent belonging to a semantic role selected as negation focus.
Page 2, “Baselines”
- For comparison, we choose the state-of-the-art system described in Blanco and Moldovan (2011), which employed various kinds of syntactic features and semantic role features, as one of our baselines.
Page 3, “Baselines”
- > Semantic features: the syntactic label of semantic role A1; whether A1 contains POS tag DT, JJ, PRP, CD, RB, VB, and WP, as defined in Blanco and Moldovan (2011); whether A1 contains token any, anybody, anymore, anyone, anything, anytime, anywhere, certain, enough, full, many, much, other, some, specifics, too, and until, as defined in Blanco and Moldovan (2011); the syntactic label of the first semantic role in the sentence; the semantic label of the last semantic role in the sentence; the thematic role for AO/Al/AZ/A3/A4 of the negated predicate.
Page 3, “Baselines”
- Along with negation focus annotation, this corpus also contains other annotations, such as POS tag, named entity, chunk, constituent tree, dependency tree, and semantic role .
Page 6, “Baselines”
- > Semantic Role Labeler: We employ the semantic role labeler, as described in Punyaka—nok et al (2008).
Page 6, “Baselines”
See all papers in Proc. ACL 2014 that mention semantic role.
See all papers in Proc. ACL that mention semantic role.
Back to top.
LDA
Appears in 5 sentences as: LDA (5)
In Negation Focus Identification with Contextual Discourse Information
- Here, the topics are extracted from all the documents in the *SEM 2012 shared task using the LDA Gibbs Sampling algorithm (Griffiths, 2002).
Page 5, “Baselines”
- In the topic-driven word-based graph model, the first layer denotes the relatedness among content words as captured in the above word-based graph model, and the second layer denotes the topic distribution, with the dashed lines between these two layers indicating the word-topic model return by LDA .
Page 5, “Baselines”
- where Rel(w,, rm) is the weight of word w in topic rm calculated by the LDA Gibbs Sampling algorithm.
Page 5, “Baselines”
- > Topic Modeler: For estimating transition probability Pt(i,m), we employ GibbsLDA++6, an LDA model using Gibbs Sampling technique for parameter estimation and inference.
Page 6, “Baselines”
- Given the LDA Gibbs Sampling model with parameters 0L = SO/T and [3 = 0.1, we vary T from 20 to 100 with an interval of 10 to find the opti-
Page 6, “Baselines”
See all papers in Proc. ACL 2014 that mention LDA.
See all papers in Proc. ACL that mention LDA.
Back to top.
co-occurrence
Appears in 4 sentences as: co-occurrence (4)
In Negation Focus Identification with Contextual Discourse Information
- One is word co-occurrence (if word w and word wj occur in the same sentence or in the adjacent sentences, Sim(wi,wj) increases 1), and the other is WordNet (Miller, 1995) based similarity.
Page 4, “Baselines”
- 1 Total weight of words in the focus candidate using the co-occurrence similarity.
Page 5, “Baselines”
- 2 Max weight of words in the focus candidate using the co-occurrence similarity.
Page 5, “Baselines”
- In this graph model, the relatedness between words is calculated by word co-occurrence , WordNet-based similarity, and topic-driven similarity.
Page 8, “Baselines”
See all papers in Proc. ACL 2014 that mention co-occurrence.
See all papers in Proc. ACL that mention co-occurrence.
Back to top.
Gibbs Sampling
Appears in 4 sentences as: Gibbs Sampling (4)
In Negation Focus Identification with Contextual Discourse Information
- Here, the topics are extracted from all the documents in the *SEM 2012 shared task using the LDA Gibbs Sampling algorithm (Griffiths, 2002).
Page 5, “Baselines”
- where Rel(w,, rm) is the weight of word w in topic rm calculated by the LDA Gibbs Sampling algorithm.
Page 5, “Baselines”
- > Topic Modeler: For estimating transition probability Pt(i,m), we employ GibbsLDA++6, an LDA model using Gibbs Sampling technique for parameter estimation and inference.
Page 6, “Baselines”
- Given the LDA Gibbs Sampling model with parameters 0L = SO/T and [3 = 0.1, we vary T from 20 to 100 with an interval of 10 to find the opti-
Page 6, “Baselines”
See all papers in Proc. ACL 2014 that mention Gibbs Sampling.
See all papers in Proc. ACL that mention Gibbs Sampling.
Back to top.
natural language
Appears in 4 sentences as: natural language (4)
In Negation Focus Identification with Contextual Discourse Information
- Negative expressions are common in natural language text and play a critical role in information extraction.
Page 1, “Abstract”
- Negation expressions are common in natural language text.
Page 1, “Introduction”
- Horn, 1989; van der Wouden, 1997), and there were only a few in natural language processing with focus on negation recognition in the biomedical domain.
Page 2, “Related Work”
- Due to the increasing demand on deep understanding of natural language text, negation recognition has been drawing more and more attention in recent years, with a series of shared tasks and workshops, however, with focus on cue detection and scope resolution, such as the BioNLP 2009 shared task for negative event detection (Kim et al., 2009) and the ACL 2010 Workshop for scope resolution of negation and speculation (Morante and Sporleder, 2010), followed by a special issue of Computational Linguistics (Morante and Sporleder, 2012) for modality and negation.
Page 2, “Related Work”
See all papers in Proc. ACL 2014 that mention natural language.
See all papers in Proc. ACL that mention natural language.
Back to top.
POS tag
Appears in 4 sentences as: POS tag (3) POS tags (1)
In Negation Focus Identification with Contextual Discourse Information
- Following is a list of features adopted in the two baselines, for both BaselineC4'5 and BaselineSVM, > Basic features: first token and its part-of-speech (POS) tag of the focus candidate; the number of tokens in the focus candidate; relative position of the focus candidate among all the roles present in the sentence; negated verb and its POS tag of the negative expression;
Page 3, “Baselines”
- > Syntactic features: the sequence of words from the beginning of the governing VP to the negated verb; the sequence of POS tags from the beginning of the governing VP to the negated verb; whether the governing VP contains a CC; whether the governing VP contains a RB.
Page 3, “Baselines”
- > Semantic features: the syntactic label of semantic role A1; whether A1 contains POS tag DT, JJ, PRP, CD, RB, VB, and WP, as defined in Blanco and Moldovan (2011); whether A1 contains token any, anybody, anymore, anyone, anything, anytime, anywhere, certain, enough, full, many, much, other, some, specifics, too, and until, as defined in Blanco and Moldovan (2011); the syntactic label of the first semantic role in the sentence; the semantic label of the last semantic role in the sentence; the thematic role for AO/Al/AZ/A3/A4 of the negated predicate.
Page 3, “Baselines”
- Along with negation focus annotation, this corpus also contains other annotations, such as POS tag , named entity, chunk, constituent tree, dependency tree, and semantic role.
Page 6, “Baselines”
See all papers in Proc. ACL 2014 that mention POS tag.
See all papers in Proc. ACL that mention POS tag.
Back to top.
named entity
Appears in 3 sentences as: named entities (1) Named Entity (1) named entity (2)
In Negation Focus Identification with Contextual Discourse Information
- > Basic features: the named entity and its type in the focus candidate; relative position of the focus candidate to the negative expression (before or after).
Page 3, “Baselines”
- Along with negation focus annotation, this corpus also contains other annotations, such as POS tag, named entity , chunk, constituent tree, dependency tree, and semantic role.
Page 6, “Baselines”
- > Named Entity Recognizer: We employ the Stanford NER5 (Finkel et al., 2005) to obtain named entities .
Page 6, “Baselines”
See all papers in Proc. ACL 2014 that mention named entity.
See all papers in Proc. ACL that mention named entity.
Back to top.
topic distribution
Appears in 3 sentences as: topic distribution (3)
In Negation Focus Identification with Contextual Discourse Information
- In particular, a word-based graph model is proposed to represent the explicit relatedness among words in a discourse from the lexical perspective, while a topic-driven word-based model is proposed to enrich the implicit relatedness between words, by adding one more layer to the word-based graph model in representing the global topic distribution of the whole dataset.
Page 4, “Baselines”
- In order to reduce the gap, we propose a topic-driven word-based model by adding one more layer to refine the word-based graph model over the global topic distribution , as shown in Figure 2.
Page 4, “Baselines”
- In the topic-driven word-based graph model, the first layer denotes the relatedness among content words as captured in the above word-based graph model, and the second layer denotes the topic distribution , with the dashed lines between these two layers indicating the word-topic model return by LDA.
Page 5, “Baselines”
See all papers in Proc. ACL 2014 that mention topic distribution.
See all papers in Proc. ACL that mention topic distribution.
Back to top.
WordNet
Appears in 3 sentences as: WordNet (3)
In Negation Focus Identification with Contextual Discourse Information
- One is word co-occurrence (if word w and word wj occur in the same sentence or in the adjacent sentences, Sim(wi,wj) increases 1), and the other is WordNet (Miller, 1995) based similarity.
Page 4, “Baselines”
- 3 Total weight of words in the focus candidate using the WordNet similarity.
Page 5, “Baselines”
- 4 Max weight of words in the focus candidate using the WordNet similarity.
Page 5, “Baselines”
See all papers in Proc. ACL 2014 that mention WordNet.
See all papers in Proc. ACL that mention WordNet.
Back to top.