Exploiting Syntactico-Semantic Structures for Relation Extraction
Chan, Yee Seng and Roth, Dan

Article Structure

Abstract

In this paper, we observe that there exists a second dimension to the relation extraction (RE) problem that is orthogonal to the relation type dimension.

Introduction

Relation extraction (RE) has been defined as the task of identifying a given set of semantic binary relations in text.

Relation Extraction Framework

In Figure l, we show the algorithm for training a typical baseline RE classifier (REbase), and for training a RE classifier that leverages the syntactico-semantic structures (RES).

Syntactico-Semantic Structures

In this paper, we performed RE on the ACE-2004 corpus.

Mention Extraction System

As part of our experiments, we perform RE using predicted mentions.

Relation Extraction System

We build a supervised RE system using sentences annotated with entity mentions and predefined target relations.

Experiments

We use the ACE-2004 dataset (catalog LDC2005T09 from the Linguistic Data Consortium) to conduct our experiments.

Analysis

We first show statistics regarding the syntactico-semantic structures.

Conclusion

In this paper, we propose a novel algorithmic approach to RE by exploiting syntactico-semantic structures.

Topics

entity typing

Appears in 13 sentences as: entity type (4) entity types (4) entity typing (6)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. We describe our mention entity typing system in Section 4 and features for the RE system in Section 5.
    Page 2, “Introduction”
  2. Abbreviations: Lm: predicted entity label for mention m using the mention entity typing (MET) classifier described in Section 4; PM ET: prediction probability according to the MET classifier; 75: used for thresholding.
    Page 3, “Relation Extraction Framework”
  3. In (Roth and Yih, 2007), the authors used entity types to constrain the (first dimensional) relation types allowed among them.
    Page 3, “Relation Extraction Framework”
  4. Table 2: Features used in our mention entity typing (MET) system.
    Page 5, “Syntactico-Semantic Structures”
  5. and whether they satisfy certain semantic entity type constraints.
    Page 5, “Syntactico-Semantic Structures”
  6. These mention candidates are then fed to our mention entity typing (MET) classifier for type prediction (more details in Section 6.3).
    Page 5, “Mention Extraction System”
  7. Due to space limitations, we refer the reader to our prior work (Chan and Roth, 2010) for the lexical, structural, mention-level, entity type , and dependency features.
    Page 6, “Relation Extraction System”
  8. ACE-2004 defines 7 coarse-grained entity types , each of which are then refined into 43 fine-
    Page 7, “Experiments”
  9. grained entity types .
    Page 8, “Experiments”
  10. Using the ACE data annotated with mentions and predefined entity types, we build a fine-grained mention entity typing (MET) classifier to disambiguate between 44 labels (43 fine-grained and a null label to indicate not a mention).
    Page 8, “Experiments”
  11. To obtain the coarse-grained entity type predictions from the classifier, we simply check which coarse-grained type the fine-grained prediction belongs to.
    Page 8, “Experiments”

See all papers in Proc. ACL 2011 that mention entity typing.

See all papers in Proc. ACL that mention entity typing.

Back to top.

POS tag

Appears in 12 sentences as: POS tag (11) POS tags (4)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. 0 If u* is not empty, we require that it satisfies any of the following POS tag sequences: JJ+ \/ JJ and JJ?
    Page 4, “Syntactico-Semantic Structures”
  2. These are (optional) POS tag sequences that normally start a valid noun phrase.
    Page 4, “Syntactico-Semantic Structures”
  3. 0 We use two patterns to differentiate between premodifier relations and possessive relations, by checking for the existence of POS tags PRP$, WP$, POS, and the word “’s”.
    Page 4, “Syntactico-Semantic Structures”
  4. If the word immediately following v+ is s or its POS tag is “POS”, we accept the mention pair.
    Page 4, “Syntactico-Semantic Structures”
  5. If the POS tag of the last word in 21+ is either PRP$ or WP$, we accept the mention pair.
    Page 4, “Syntactico-Semantic Structures”
  6. These are a combination of 21);, itself, its POS tag , and its integer offset from the last word (lw) in the mention.
    Page 5, “Mention Extraction System”
  7. These features are meant to capture the word and POS tag sequences in mentions.
    Page 5, “Mention Extraction System”
  8. Contextual We extract the word C_1,_1 immediately before mi, the word C+1,+1 immediately after mi, and their associated POS tags P.
    Page 5, “Mention Extraction System”
  9. POS features If there is a single word between the two mentions, we extract its POS tag .
    Page 6, “Relation Extraction System”
  10. Given the hw of m, Pm- refers to the sequence of POS tags in the immediate context of hw (we exclude the POS tag of hw).
    Page 6, “Relation Extraction System”
  11. The offsets i and j denote the position (relative to hw) of the first and last POS tag respectively.
    Page 6, “Relation Extraction System”

See all papers in Proc. ACL 2011 that mention POS tag.

See all papers in Proc. ACL that mention POS tag.

Back to top.

fine-grained

Appears in 7 sentences as: fine-grained (8)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. Out of these, 4,011 are positive relation examples annotated with 6 coarse-grained relation types and 22 fine-grained relation types5 .
    Page 6, “Experiments”
  2. We similarly build a fine-grained classifier to disambiguate between 45 relation labels.
    Page 6, “Experiments”
  3. We built one binary, one coarse-grained, and one fine-grained classifier for each fold.
    Page 7, “Experiments”
  4. The results show that by using syntactico-semantic structures, we obtain significant F-measure improvements of 8.3, 7.2, and 5.5 for binary, coarse-grained, and fine-grained relation predictions respectively.
    Page 7, “Experiments”
  5. Using the ACE data annotated with mentions and predefined entity types, we build a fine-grained mention entity typing (MET) classifier to disambiguate between 44 labels (43 fine-grained and a null label to indicate not a mention).
    Page 8, “Experiments”
  6. To obtain the coarse-grained entity type predictions from the classifier, we simply check which coarse-grained type the fine-grained prediction belongs to.
    Page 8, “Experiments”
  7. The results show that by leveraging syntactico-semantic structures, we obtain significant F-measure improvements of 8.2, 4.6, and 3.6 for binary, coarse-grained, and fine-grained relation predictions respectively.
    Page 8, “Experiments”

See all papers in Proc. ACL 2011 that mention fine-grained.

See all papers in Proc. ACL that mention fine-grained.

Back to top.

parse tree

Appears in 4 sentences as: parse tree (4)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. with lw of m,- in the sentence Syntactic parse-label of parse tree constituent parse that exactly covers m,-
    Page 5, “Syntactico-Semantic Structures”
  2. parse-labels of parse tree constituents covering 771,;
    Page 5, “Syntactico-Semantic Structures”
  3. We extract the label of the parse tree constituent (if it exists) that exactly covers the mention, and also labels of all constituents that covers the mention.
    Page 5, “Mention Extraction System”
  4. From a sentence, we gather the following as candidate mentions: all nouns and possessive pronouns, all named entities annotated by the the NE tagger (Ratinov and Roth, 2009), all base noun phrase (NP) chunks, all chunks satisfying the pattern: NP (PP NP)+, all NP constituents in the syntactic parse tree , and from each of these constituents, all substrings consisting of two or more words, provided the sub-strings do not start nor end on punctuation marks.
    Page 5, “Mention Extraction System”

See all papers in Proc. ACL 2011 that mention parse tree.

See all papers in Proc. ACL that mention parse tree.

Back to top.

relation extraction

Appears in 4 sentences as: Relation extraction (1) relation extraction (3)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. In this paper, we observe that there exists a second dimension to the relation extraction (RE) problem that is orthogonal to the relation type dimension.
    Page 1, “Abstract”
  2. Relation extraction (RE) has been defined as the task of identifying a given set of semantic binary relations in text.
    Page 1, “Introduction”
  3. In this paper we build on the observation that there exists a second dimension to the relation extraction problem that is orthogonal to the relation type dimension: all relation types are expressed in one of several constrained syntactico-semantic structures.
    Page 1, “Introduction”
  4. In the next section, we describe our relation extraction framework that leverages the syntactico-semantic structures.
    Page 2, “Introduction”

See all papers in Proc. ACL 2011 that mention relation extraction.

See all papers in Proc. ACL that mention relation extraction.

Back to top.

coreferent

Appears in 3 sentences as: coreference (1) coreferent (2)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. In that work, we also highlight that ACE annotators rarely duplicate a relation link for coreferent mentions.
    Page 6, “Experiments”
  2. For instance, assume mentions mi, mj, and mk, are in the same sentence, mentions mi and mj are coreferent , and the annotators tag the mention pair mj, mk, with a particular relation r. The annotators will rarely duplicate the same (implicit)
    Page 6, “Experiments”
  3. Of course, using this scoring method requires coreference information, which is available in the ACE data.
    Page 7, “Experiments”

See all papers in Proc. ACL 2011 that mention coreferent.

See all papers in Proc. ACL that mention coreferent.

Back to top.

F-measure

Appears in 3 sentences as: F-measure (3)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. Using the same evaluation setting, our baseline RE system achieves a competitive 71.4 F-measure .
    Page 6, “Experiments”
  2. The results show that by using syntactico-semantic structures, we obtain significant F-measure improvements of 8.3, 7.2, and 5.5 for binary, coarse-grained, and fine-grained relation predictions respectively.
    Page 7, “Experiments”
  3. The results show that by leveraging syntactico-semantic structures, we obtain significant F-measure improvements of 8.2, 4.6, and 3.6 for binary, coarse-grained, and fine-grained relation predictions respectively.
    Page 8, “Experiments”

See all papers in Proc. ACL 2011 that mention F-measure.

See all papers in Proc. ACL that mention F-measure.

Back to top.

named entity

Appears in 3 sentences as: named entities (1) named entity (3)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. lw: last word in the mention; Bc(w): the brown cluster bit string representing w; NE: named entity
    Page 5, “Syntactico-Semantic Structures”
  2. NE tags We automatically annotate the sentences with named entity (NE) tags using the named entity tagger of (Ratinov and Roth, 2009).
    Page 5, “Mention Extraction System”
  3. From a sentence, we gather the following as candidate mentions: all nouns and possessive pronouns, all named entities annotated by the the NE tagger (Ratinov and Roth, 2009), all base noun phrase (NP) chunks, all chunks satisfying the pattern: NP (PP NP)+, all NP constituents in the syntactic parse tree, and from each of these constituents, all substrings consisting of two or more words, provided the sub-strings do not start nor end on punctuation marks.
    Page 5, “Mention Extraction System”

See all papers in Proc. ACL 2011 that mention named entity.

See all papers in Proc. ACL that mention named entity.

Back to top.

semantic relations

Appears in 3 sentences as: semantic relations (2) semantically related (1)
In Exploiting Syntactico-Semantic Structures for Relation Extraction
  1. RE has been frequently studied over the last few years as a supervised learning task, learning from spans of text that are annotated with a set of semantic relations of interest.
    Page 1, “Introduction”
  2. These four structures cover 80% of the mention pairs having valid semantic relations (we give the detailed breakdown in Section 7) and we show that they are relatively easy to identify using simple rules or patterns.
    Page 3, “Syntactico-Semantic Structures”
  3. Preposition indicates that the two mentions are semantically related via the existence of a preposition, e.g.
    Page 3, “Syntactico-Semantic Structures”

See all papers in Proc. ACL 2011 that mention semantic relations.

See all papers in Proc. ACL that mention semantic relations.

Back to top.

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