Mutual Disambiguation for Entity Linking
Charton, Eric and Meurs, Marie-Jean and Jean-Louis, Ludovic and Gagnon, Michel

Article Structure

Abstract

The disambiguation algorithm presented in this paper is implemented in SemLinker, an entity linking system.

Introduction

The Entity Linking (EL) task consists in linking name mentions of named entities (NEs) found in a document to their corresponding entities in a reference Knowledge Base (KB).

Related Work

Entity annotation and linking in natural language text has been extensively studied in NLP research.

Proposed Algorithm

We propose a mutual disambiguation algorithm that improves the accuracy of entity links in a document by using successive corrections applied to an annotation object representing this document.

Experiments and Results

SemLinker has been evaluated on the TAC-KBP 2012 EL task (Charton et al., 2013).

Conclusion

The presented system provides a robust semantic disambiguation method, based on mutual relation of entities inside a document, using a standard annotation engine.

Topics

semantic relatedness

Appears in 9 sentences as: semantic relatedness (6) semantic relation (3) semantically related (1)
In Mutual Disambiguation for Entity Linking
  1. Such techniques are referred to as semantic relatedness (Strube and Ponzetto, 2006), collective disambiguation (Hoffart et al., 2011b), or joint disambiguation (Fahrni et al., 2012).
    Page 1, “Introduction”
  2. For example, if a NE describes a city name like Paris, it is more probable that the correct link for this city name designates Paris (France) rather than Paris (Texas) if a neighbor entity offers candidate links semantically related to Paris (France) like the Seine river or the Champs-Elyse’es.
    Page 1, “Introduction”
  3. The paper makes the following novel propositions: l) the ontology used to evaluate the relatedness of candidates is replaced by internal links and categories from the Wikipedia corpus; 2) the coherence of entities is improved prior to the calculation of semantic relatedness using a co-reference resolution algorithm, and a NE label correction method; 3) the proposed method is robust enough to improve the performance of existing entity linking annotation engines, which are capable of providing a set of ranked candidates for each annotation in a document.
    Page 2, “Introduction”
  4. also introduced the notion of semantic relatedness .
    Page 2, “Related Work”
  5. While all these approaches focus on semantic relation between entities, their potential is limited by the separate mapping of candidate links for each mention.
    Page 2, “Related Work”
  6. Only some of these systems introduce the semantic relatedness in their methods like the AIDA (Hoffart et al., 2011b) system.
    Page 2, “Related Work”
  7. A basic example of semantic relatedness that should be captured is explained hereafter.
    Page 3, “Proposed Algorithm”
  8. The purpose of the MDP is to capture this semantic relatedness information contained in the graph of links extracted from Wikipedia pages related to each candidate annotation.
    Page 3, “Proposed Algorithm”
  9. The calculation combines two scores that we called direct semantic relation score (dsr_score) and common semantic relation score (csr_score):
    Page 4, “Proposed Algorithm”

See all papers in Proc. ACL 2014 that mention semantic relatedness.

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

Back to top.

named entities

Appears in 8 sentences as: named entities (4) Named Entity (2) named entity (2)
In Mutual Disambiguation for Entity Linking
  1. First, named entities are linked to candidate Wikipedia pages by a generic annotation engine.
    Page 1, “Abstract”
  2. Then, the algorithm re—ranks candidate links according to mutual relations between all the named entities found in the document.
    Page 1, “Abstract”
  3. The Entity Linking (EL) task consists in linking name mentions of named entities (NEs) found in a document to their corresponding entities in a reference Knowledge Base (KB).
    Page 1, “Introduction”
  4. Various approaches have been proposed to solve the named entity disambiguation (NED) problem.
    Page 1, “Introduction”
  5. In the context of the Named Entity Recognition (NER) task, such resources can be generic and generative.
    Page 1, “Introduction”
  6. 3.2 Named Entity Label Correction
    Page 3, “Proposed Algorithm”
  7. In this task, mentions of entities found in a document collection must be linked to entities in a reference KB, or to new named entities discovered in the collection.
    Page 4, “Experiments and Results”
  8. We observe that the complete algorithm (co-references, named entity labels and MDP) provides the best results on PER NE links.
    Page 5, “Experiments and Results”

See all papers in Proc. ACL 2014 that mention named entities.

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

Back to top.

best results

Appears in 3 sentences as: best results (3)
In Mutual Disambiguation for Entity Linking
  1. The three best results and the median from TAC-KBP 2012 systems are shown in the remaining columns for the sake of comparison.
    Page 5, “Experiments and Results”
  2. We observe that the complete algorithm (co-references, named entity labels and MDP) provides the best results on PER NE links.
    Page 5, “Experiments and Results”
  3. On GPE and ORG entities, the simple application of MDP without prior corrections obtains the best results .
    Page 5, “Experiments and Results”

See all papers in Proc. ACL 2014 that mention best results.

See all papers in Proc. ACL that mention best results.

Back to top.

knowledge base

Appears in 3 sentences as: Knowledge Base (1) knowledge base (2)
In Mutual Disambiguation for Entity Linking
  1. The Entity Linking (EL) task consists in linking name mentions of named entities (NEs) found in a document to their corresponding entities in a reference Knowledge Base (KB).
    Page 1, “Introduction”
  2. It relies on the Wikipedia-derived YAGO2 (Hoffart et al., 2011a) knowledge base .
    Page 2, “Related Work”
  3. The reference knowledge base is derived from an October 2008 dump of English Wikipedia, which includes 818,741 nodes.
    Page 5, “Experiments and Results”

See all papers in Proc. ACL 2014 that mention knowledge base.

See all papers in Proc. ACL that mention knowledge base.

Back to top.

similarity measures

Appears in 3 sentences as: similarity measures (3)
In Mutual Disambiguation for Entity Linking
  1. They are used as matching sequences to locate corresponding candidate entries in the KB, and then to disambiguate those candidates using similarity measures .
    Page 1, “Introduction”
  2. This is usually done using similarity measures (such as cosine similarity, weighted J accard distance, KL divergence...) that evaluate the distance between a bag of words related to a candidate annotation, and the words surrounding the entity to annotate in the text.
    Page 1, “Introduction”
  3. It proposes a disambiguation method that combines popularity-based priors, similarity measures , and coherence.
    Page 2, “Related Work”

See all papers in Proc. ACL 2014 that mention similarity measures.

See all papers in Proc. ACL that mention similarity measures.

Back to top.

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