A Metric-based Framework for Automatic Taxonomy Induction
Yang, Hui and Callan, Jamie

Article Structure

Abstract

This paper presents a novel metric-based framework for the task of automatic taxonomy induction.

Introduction

Automatic taxonomy induction is an important task in the fields of Natural Language Processing, Knowledge Management, and Semantic Web.

Related Work

There has been a substantial amount of research on automatic taxonomy induction.

The Features

The features used in this work are indicators of semantic relations between terms.

The Metric-based Framework

This section presents the metric-based framework which incrementally clusters terms to form taxonomies.

Experiments

5.1 Data

Conclusions

This paper presents a novel metric-based taxonomy induction framework combining the strengths of lexico-syntactic patterns and clustering.

Topics

co-occurrence

Appears in 13 sentences as: Co-occurrence (2) co-occurrence (12)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. The common types of features include contextual (Lin, 1998), co-occurrence (Yang and Callan, 2008), and syntactic dependency (Pantel and Lin, 2002; Pantel and Ravichandran, 2004).
    Page 2, “Introduction”
  2. The framework integrates contextual, co-occurrence , syntactic dependency, lexi-cal-syntactic patterns, and other features to learn an ontology metric, a score indicating semantic distance, for each pair of terms in a taxonomy; it then incrementally clusters terms based on their ontology metric scores.
    Page 2, “Introduction”
  3. Inspired by the conjunction and appositive structures, Riloff and Shepherd (1997), Roark and Charniak (1998) used co-occurrence statistics in local context to discover sibling relations.
    Page 2, “Related Work”
  4. Besides contextual features, the vectors can also be represented by verb-noun relations (Pereira et al., 1993), syntactic dependency (Pantel and Ravichandran, 2004; Snow et al., 2005), co-occurrence (Yang and Callan, 2008), conjunction and appositive features (Caraballo, 1999).
    Page 3, “Related Work”
  5. The features include contextual, co-occurrence , syntactic dependency, lexical-syntactic patterns, and miscellaneous.
    Page 3, “The Features”
  6. The second set of features is co-occurrence .
    Page 3, “The Features”
  7. In our work, co-occurrence is measured by point-wise mutual information between two terms:
    Page 3, “The Features”
  8. ), we have (3) Document PM], (4) Sentence PM], and (5) Google PM] as the co-occurrence features.
    Page 4, “The Features”
  9. Co-occurrence 0.47 0.56 0.45 0.41 0.41
    Page 8, “Experiments”
  10. Co-occurrence 0.34 0.36 0.34 0.31 0.31
    Page 8, “Experiments”
  11. Both co-occurrence and lexico-syntactic patterns work well for all three types of relations.
    Page 8, “Experiments”

See all papers in Proc. ACL 2009 that mention co-occurrence.

See all papers in Proc. ACL that mention co-occurrence.

Back to top.

WordNet

Appears in 13 sentences as: WordNet (16) WordNetS (1)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. It has been receiving increasing afienfion.because senuuunztaxononues,such as WordNet (Fellbaum, 1998), play an important role in solving knowledge-rich problems, including question answering (Harabagiu et a1., 2003) and textual entailment (Geffet and Dagan, 2005).
    Page 1, “Introduction”
  2. Pattern quality control is also investigated by using WordNet (Girju et al., 2006), graph structures built among terms (Widdows and Dorow, 2002; Kozareva et al., 2008), and pattern clusters (Davidov and Rappoport, 2008).
    Page 3, “Related Work”
  3. The gold standards used in the evaluation are hypernym taxonomies extracted from WordNet and GDP (Open Directory Project), and meronym taxonomies extracted from WordNet .
    Page 6, “Experiments”
  4. In WordNet taxonomy extraction, we only use the word senses within a particular taxonomy to ensure no ambiguity.
    Page 6, “Experiments”
  5. In total, there are 100 hypernym taxonomies, 50 each extracted from WordNet3 and ODP4, and 50 meronym taxonomies from WordNetS .
    Page 6, “Experiments”
  6. 3 WordNet hypernym taxonomies are from 12 topics: gathering, professional, people, building, place, milk, meal, water, beverage, alcohol, dish, and herb.
    Page 6, “Experiments”
  7. 5 WordNet meronym taxonomies are from 15 topics: bed, car, building, lamp, earth, television, body, drama, theatre, water, airplane, piano, book, computer, and watch.
    Page 6, “Experiments”
  8. For each 50 datasets from WordNet hypernyms, WordNet meronyms or ODP hypernyms, we randomly pick 49 of them to generate training data, and test on the remaining dataset.
    Page 7, “Experiments”
  9. Table 3 shows precision, recall, and F1-1easure of each system for WordNet hypernyms isa), WordNet meronyms (part-0f) and GDP .ypernyms (isa).
    Page 7, “Experiments”
  10. WordNet .
    Page 8, “Experiments”
  11. Table 4 shows Fl-measure of using each set of features alone on taxonomy induction for WordNet isa, sibling, and part-of relations.
    Page 8, “Experiments”

See all papers in Proc. ACL 2009 that mention WordNet.

See all papers in Proc. ACL that mention WordNet.

Back to top.

hypernym

Appears in 9 sentences as: Hypernym (2) hypernym (5) hypernyms (3)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. (2004) extended isa relation acquisition towards terascale, and automatically identified hypernym patterns by minimal edit distance.
    Page 2, “Related Work”
  2. Hypernym Patterns Sibling Patterns
    Page 4, “The Features”
  3. We have (11) Hypernym Patterns based on patterns proposed by (Hearst, 1992) and (Snow et al., 2005), (12) Sibling Patterns which are basically conjunctions, and (13) Part-of Patterns based on patterns proposed by (Girju et al., 2003) and (Cimiano and Wenderoth, 2007).
    Page 4, “The Features”
  4. The gold standards used in the evaluation are hypernym taxonomies extracted from WordNet and GDP (Open Directory Project), and meronym taxonomies extracted from WordNet.
    Page 6, “Experiments”
  5. In total, there are 100 hypernym taxonomies, 50 each extracted from WordNet3 and ODP4, and 50 meronym taxonomies from WordNetS.
    Page 6, “Experiments”
  6. 3 WordNet hypernym taxonomies are from 12 topics: gathering, professional, people, building, place, milk, meal, water, beverage, alcohol, dish, and herb.
    Page 6, “Experiments”
  7. 4 GDP hypernym taxonomies are from 16 topics: computers, robotics, intranet, mobile computing, database, operating system, linux, tex, software, computer science, data communication, algorithms, data formats, security multimedia, and artificial intelligence.
    Page 6, “Experiments”
  8. For each 50 datasets from WordNet hypernyms, WordNet meronyms or ODP hypernyms , we randomly pick 49 of them to generate training data, and test on the remaining dataset.
    Page 7, “Experiments”
  9. Table 3 shows precision, recall, and F1-1easure of each system for WordNet hypernyms isa), WordNet meronyms (part-0f) and GDP .ypernyms (isa).
    Page 7, “Experiments”

See all papers in Proc. ACL 2009 that mention hypernym.

See all papers in Proc. ACL that mention hypernym.

Back to top.

edge weights

Appears in 3 sentences as: edge weight (1) edge weights (3)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. Formally, it is a function d :C><C a [Rh where C is the set of terms in T. An ontology metric d on a taxonomy T with edge weights w
    Page 4, “The Metric-based Framework”
  2. for any term pair (ox,cy)EC is the sum of all edge weights along the shortest path between the pair:
    Page 4, “The Metric-based Framework”
  3. In the training data, an ontology metric d(c,,,cy) for a term pair (obey) is generated by assuming every edge weight as 1 and summing up all the edge weights along the shortest path from C, to Cy.
    Page 6, “The Metric-based Framework”

See all papers in Proc. ACL 2009 that mention edge weights.

See all papers in Proc. ACL that mention edge weights.

Back to top.

gold standards

Appears in 3 sentences as: gold standard (1) gold standards (2)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. The gold standards used in the evaluation are hypernym taxonomies extracted from WordNet and GDP (Open Directory Project), and meronym taxonomies extracted from WordNet.
    Page 6, “Experiments”
  2. We evaluate the quality of automatic generated taxonomies by comparing them with the gold standards in terms of precision, recall and F1-measure.
    Page 7, “Experiments”
  3. Fl-measure is calculated as 2*P*R/ (P+R), where P is precision, the percentage of correctly returned relations out of the total returned relations, R is recall, the percentage of correctly returned relations out of the total relations in the gold standard .
    Page 7, “Experiments”

See all papers in Proc. ACL 2009 that mention gold standards.

See all papers in Proc. ACL that mention gold standards.

Back to top.

language model

Appears in 3 sentences as: language model (2) language models (1)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. It is built into a unigram language model without smoothing for each term.
    Page 3, “The Features”
  2. This feature function measures the Kullback—Leibler divergence (KL divergence) between the language models associated with the two inputs.
    Page 3, “The Features”
  3. Similarly, the local context is built into a unigram language model without smoothing for each term; the feature function outputs KL divergence between the models.
    Page 3, “The Features”

See all papers in Proc. ACL 2009 that mention language model.

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

Back to top.

semantic relations

Appears in 3 sentences as: semantic relation (1) semantic relations (2)
In A Metric-based Framework for Automatic Taxonomy Induction
  1. Existing work on automatic taxonomy induction has been conducted under a variety of names, such as ontology learning, semantic class learning, semantic relation classification, and relation extraction.
    Page 1, “Introduction”
  2. They have been applied to extract various types of lexical and semantic relations , including isa, part-of, sibling, synonym, causal, and many others.
    Page 2, “Related Work”
  3. The features used in this work are indicators of semantic relations between terms.
    Page 3, “The Features”

See all papers in Proc. ACL 2009 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