A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
Tian, Zhenhua and Xiang, Hengheng and Liu, Ziqi and Zheng, Qinghua

Article Structure

Abstract

This paper presents an unsupervised random walk approach to alleviate data sparsity for selectional preferences.

Introduction

Selectional preferences (SP) or selectional restrictions capture the plausibility of predicates and their arguments for a given relation.

Related Work 2.1 WordNet-based Approach

Resnik (1996) conducts the pioneer work on corpus-driven SP induction.

RSP: A Random Walk Model for SP

In this section, we briefly introduce how to address SP using random walk.

Topics

data sparsity

Appears in 7 sentences as: data sparsity (7)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. This paper presents an unsupervised random walk approach to alleviate data sparsity for selectional preferences.
    Page 1, “Abstract”
  2. However, this strategy is infeasible for many plausible triples due to data sparsity .
    Page 1, “Introduction”
  3. Then how to use a smooth model to alleviate data sparsity for SP?
    Page 1, “Introduction”
  4. Random walk models have been successfully applied to alleviate the data sparsity issue on collaborative filtering in recommender systems.
    Page 1, “Introduction”
  5. In this paper, we present an extension of using the random walk model to alleviate data sparsity for SP.
    Page 1, “Introduction”
  6. The damp factor d E (0, l), and its value mainly depends on the data sparsity level.
    Page 4, “RSP: A Random Walk Model for SP”
  7. Experiments show it is efficient and effective to address data sparsity for SP.
    Page 9, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention data sparsity.

See all papers in Proc. ACL that mention data sparsity.

Back to top.

co-occurrence

Appears in 6 sentences as: co-occurrence (6)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. (1999) introduce a general similarity-based model for word co-occurrence probabilities, which can be interpreted for SP.
    Page 2, “Related Work 2.1 WordNet-based Approach”
  2. We initiate the links E with the raw co-occurrence counts of seen predicate-argument pairs in a given generalization data.
    Page 3, “RSP: A Random Walk Model for SP”
  3. But in SP, the preferences between the predicates and arguments are implicit: their co-occurrence counts follow the power law distribution and vary greatly.
    Page 4, “RSP: A Random Walk Model for SP”
  4. investigate the correlations between the co-occurrence counts (CT) C(q, a), or smoothed counts with the human plausibility judgements (Lapata et al., 1999; Lapata et al., 2001).
    Page 4, “RSP: A Random Walk Model for SP”
  5. (1999) propose state-of-the-art similarity based model for word co-occurrence probabilities.
    Page 6, “RSP: A Random Walk Model for SP”
  6. (2001), we first collect the co-occurrence counts of predicate-argument pairs in the human plausibility data from AFP and NYT (before removing them as unseen pairs).
    Page 9, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention co-occurrence.

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

Back to top.

NER

Appears in 6 sentences as: NER (7)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. Random confounder (RND) most closes to the realistic case; While nearest confounder ( NER ) is reproducible and it avoids frequency bias (Chambers and Jurafsky, 2010).
    Page 6, “RSP: A Random Walk Model for SP”
  2. In this work, we employ both RND and NER confounders: 1) for RND, we randomly select
    Page 6, “RSP: A Random Walk Model for SP”
  3. 2) for NER , firstly we sort the arguments by their frequency.
    Page 7, “RSP: A Random Walk Model for SP”
  4. This experiment is conducted on the PTB development tests with both RND and NER confounders.
    Page 7, “RSP: A Random Walk Model for SP”
  5. II 8 G, , — aIe — RND macro accuracy (U 80 7 A f — e — RND micro accuracy II —A— NER macro accuracy 78 7 , —V— NER micro accuracy at-76 I I I I I I I I 0.5 1 1.5 2 2.5 3 3.5 4 4.5
    Page 7, “RSP: A Random Walk Model for SP”
  6. We report the results on the PTB final test set, with RND and NER confounders.
    Page 8, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention NER.

See all papers in Proc. ACL that mention NER.

Back to top.

WordNet

Appears in 5 sentences as: WordNet (5)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. For a given predicate q, the system firstly computes its distribution of argument semantic classes based on WordNet .
    Page 2, “Related Work 2.1 WordNet-based Approach”
  2. Clark and Weir (2002) suggest a hypothesis testing method by ascending the noun hierarchy of WordNet .
    Page 2, “Related Work 2.1 WordNet-based Approach”
  3. Cia-ramita and Johnson (2000) model WordNet as a Bayesian network to solve the “explain away” ambiguity.
    Page 2, “Related Work 2.1 WordNet-based Approach”
  4. 2.2 Distributional Models without WordNet
    Page 2, “Related Work 2.1 WordNet-based Approach”
  5. The key idea is to use the latent clusterings to take the place of WordNet semantic classes.
    Page 2, “Related Work 2.1 WordNet-based Approach”

See all papers in Proc. ACL 2013 that mention WordNet.

See all papers in Proc. ACL that mention WordNet.

Back to top.

topic models

Appears in 4 sentences as: topic model (1) Topic Modeling (1) topic models (2)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. Recently, more sophisticated methods are innovated for SP based on topic models , where the latent variables (topics) take the place of semantic classes and distributional clusterings (Seaghdha, 2010; Ritter et al., 2010).
    Page 2, “Related Work 2.1 WordNet-based Approach”
  2. Since RSP falls into the unsupervised distributional approach, we compare it with previous similarity-based methods and unsupervised generative topic model 3.
    Page 6, “RSP: A Random Walk Model for SP”
  3. C) Seaghdha (2010) applies topic models for the SP induction with three variations: LDA, Rooth-LDA, and Dual-LDA; Ritter et al.
    Page 6, “RSP: A Random Walk Model for SP”
  4. We use the Mat-lab Topic Modeling Toolbox4 for the inference of latent topics.
    Page 6, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention topic models.

See all papers in Proc. ACL that mention topic models.

Back to top.

clusterings

Appears in 3 sentences as: clusterings (3)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. The key idea is to use the latent clusterings to take the place of WordNet semantic classes.
    Page 2, “Related Work 2.1 WordNet-based Approach”
  2. Where the latent clusterings are automatically derived from distributional data based on EM algorithm.
    Page 2, “Related Work 2.1 WordNet-based Approach”
  3. Recently, more sophisticated methods are innovated for SP based on topic models, where the latent variables (topics) take the place of semantic classes and distributional clusterings (Seaghdha, 2010; Ritter et al., 2010).
    Page 2, “Related Work 2.1 WordNet-based Approach”

See all papers in Proc. ACL 2013 that mention clusterings.

See all papers in Proc. ACL that mention clusterings.

Back to top.

conditional probability

Appears in 3 sentences as: conditional probabilities (1) conditional probability (2)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. We suppose this could bring at least two benefits: 1) a proper measure on the preferences can make the discovering of nearby predicates with similar preferences to be more accurate; 2) while propagation, we propagate the scored preferences, rather than the raw counts or conditional probabilities , which could be more proper and agree with the nature of SP smooth.
    Page 4, “RSP: A Random Walk Model for SP”
  2. Some introduce conditional probability (CP) p(a|q) for the decision of preference judgements (Chambers and Jurafsky, 2010; Erk et al., 2010; Seaghdha, 2010).
    Page 4, “RSP: A Random Walk Model for SP”
  3. In this mode, we always set Pr(q, a) as the conditional probability p(a|q) for the propagation function, despite what \11 is used for the distance function.
    Page 5, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention conditional probability.

See all papers in Proc. ACL that mention conditional probability.

Back to top.

development set

Appears in 3 sentences as: development set (3)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. We split the test set equally into two parts: one as the development set and the other as the final test set.
    Page 5, “RSP: A Random Walk Model for SP”
  2. Parameters Tuning: The parameters are tuned on the PTB development set , using AFP as the generalization data.
    Page 7, “RSP: A Random Walk Model for SP”
  3. This experiment is conducted on the PTB development set with RND confounders.
    Page 7, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention development set.

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

Back to top.

latent variables

Appears in 3 sentences as: latent variable (1) latent variables (2)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. Recently, more sophisticated methods are innovated for SP based on topic models, where the latent variables (topics) take the place of semantic classes and distributional clusterings (Seaghdha, 2010; Ritter et al., 2010).
    Page 2, “Related Work 2.1 WordNet-based Approach”
  2. Without introducing semantic classes and latent variables , Keller and Lapata (2003) use the web to obtain frequencies for unseen bigrams smooth.
    Page 2, “Related Work 2.1 WordNet-based Approach”
  3. LDA-SP: Another kind of sophisticated unsupervised approaches for SP are latent variable models based on Latent Dirichlet Allocation (LDA).
    Page 6, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention latent variables.

See all papers in Proc. ACL that mention latent variables.

Back to top.

LDA

Appears in 3 sentences as: LDA (3)
In A Random Walk Approach to Selectional Preferences Based on Preference Ranking and Propagation
  1. LDA-SP: Another kind of sophisticated unsupervised approaches for SP are latent variable models based on Latent Dirichlet Allocation ( LDA ).
    Page 6, “RSP: A Random Walk Model for SP”
  2. C) Seaghdha (2010) applies topic models for the SP induction with three variations: LDA , Rooth-LDA, and Dual-LDA; Ritter et al.
    Page 6, “RSP: A Random Walk Model for SP”
  3. In this work, we compare with C) Seaghdha’s original LDA approach to SP.
    Page 6, “RSP: A Random Walk Model for SP”

See all papers in Proc. ACL 2013 that mention LDA.

See all papers in Proc. ACL that mention LDA.

Back to top.

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