Improved Lexical Acquisition through DPP-based Verb Clustering
Reichart, Roi and Korhonen, Anna

Article Structure

Abstract

Subcategorization frames (SCFs), selectional preferences (SPs) and verb classes capture related aspects of the predicate-argument structure.

Introduction

Verb classes (VCs), subcategorization frames (SCFs) and selectional preferences (SPs) capture different aspects of predicate-argument structure.

Previous Work

SCF acquisition Most current works induce SCFs from the output of an unleXicalized parser (i.e.

The Unified Framework

In this section we present our unified framework.

Evaluation

Data sets and gold standards We evaluated the SCFs and verb clusters on gold standard datasets.

Conclusions and Future Work

In this paper we have presented the first unified framework for the induction of verb clusters, subcategorization frames and selectional preferences from corpus data.

Topics

gold standard

Appears in 9 sentences as: gold standard (9) gold standards (2)
In Improved Lexical Acquisition through DPP-based Verb Clustering
  1. Our evaluation against a well-known VC gold standard shows that our clustering model outperforms the state-of-the-art verb clustering algorithm of Sun and Korhonen
    Page 2, “Introduction”
  2. Our evaluation against a well-known SCF gold standard and in the context of SP disambiguation tasks shows results that are superior to strong baselines, demonstrating the benefit our approach.
    Page 2, “Introduction”
  3. Data sets and gold standards We evaluated the SCFs and verb clusters on gold standard datasets.
    Page 5, “Evaluation”
  4. This yielded 101 additional verbs which we added to the gold standard with the initial 183 verbs.
    Page 6, “Evaluation”
  5. Since 176 out of the 183 initial verbs are represented in this corpus, our final gold standard consists of 34 classes containing 277 verbs, of which 176 have SCF gold standard and has been evaluated for this task.
    Page 6, “Evaluation”
  6. Clustering Evaluation We first evaluate the quality of the clusters induced by our algorithm (DPP-cluster) compared to the gold standard VCs (table 1).
    Page 6, “Evaluation”
  7. We do this by gathering the GR combinations for each of the verbs in our gold standard , assuming they are frames and gathering their frequencies.
    Page 7, “Evaluation”
  8. The induced clusters performed well in evaluation against a VerbNet -based gold standard and proved useful in improving the quality of SCFs and SPs over strong baselines.
    Page 8, “Conclusions and Future Work”
  9. In future work we plan to apply our approach on larger scale data sets and gold standards and to evaluate it in different domains, languages and in the context of NLP tasks such as syntactic parsing and SRL.
    Page 9, “Conclusions and Future Work”

See all papers in Proc. ACL 2013 that mention gold standard.

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

Back to top.

joint model

Appears in 6 sentences as: joint model (3) Joint Modeling (1) joint modeling (1) joint models (1)
In Improved Lexical Acquisition through DPP-based Verb Clustering
  1. (2012) presented a joint model for inducing simple syntactic frames and VCs.
    Page 2, “Introduction”
  2. (2012) introduced a joint model for SCF and SP acquisition.
    Page 2, “Introduction”
  3. Joint Modeling A small number of works have recently investigated joint approaches to SCFs, SPs and VCs.
    Page 3, “Previous Work”
  4. Although evaluation of these recent joint models has been partial, the results have been encouraging and fur-
    Page 3, “Previous Work”
  5. DPPs are particularly suitable for joint modeling as they come with various simple and intuitive ways to combine individual model kernel matrices into a joint kernel.
    Page 4, “The Unified Framework”
  6. A natural extension of our unified framework is to construct a joint model in which the predictions for all three tasks inform each other at all stages of the prediction process.
    Page 9, “Conclusions and Future Work”

See all papers in Proc. ACL 2013 that mention joint model.

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

Back to top.

iteratively

Appears in 5 sentences as: iteratively (5)
In Improved Lexical Acquisition through DPP-based Verb Clustering
  1. Cluster set construction In its while loop, the algorithm iteratively generates fixed-size cluster sets such that each data point belongs to exactly one cluster in one set.
    Page 5, “The Unified Framework”
  2. Then, it gradually extends the clusters by iteratively mapping the samples, in decreasing order of probability, to the existing clusters (the mlMappz'ng function).
    Page 5, “The Unified Framework”
  3. By iteratively extending the clusters with high probability subsets, we thus expect each cluster set to consist of clusters that demonstrate these properties.
    Page 5, “The Unified Framework”
  4. The agglomerative clustering then iteratively combines cluster sets such that in each iteration two sets are combined to one set with K clusters.
    Page 5, “The Unified Framework”
  5. Agglomerative Clustering Finally, the AgglomerativeClustering function builds a hierarchy of cluster sets, by iteratively combining cluster set pairs.
    Page 5, “The Unified Framework”

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

See all papers in Proc. ACL that mention iteratively.

Back to top.

im

Appears in 4 sentences as: im (4)
In Improved Lexical Acquisition through DPP-based Verb Clustering
  1. (O’Donovan et al., 2005; Schulte im Walde, 2006; Erk, 2007; Preiss et al., 2007; Van de Cruys, 2009; Reisinger and Mooney, 2011; Sun and Korhonen, 2011; Lippincott et al., 2012).
    Page 1, “Introduction”
  2. Schulte im Walde et al.
    Page 2, “Introduction”
  3. K—means and spectral) algorithms (Schulte im Walde, 2006; Joanis et al., 2008; Sun et al., 2008; Li and Brew, 2008; Korhonen et al., 2008; Sun and Korhonen, 2009; Vlachos et al., 2009; Sun and Korhonen, 2011).
    Page 3, “Previous Work”
  4. Finally, the model of Schulte im Walde et a1.
    Page 3, “Previous Work”

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

See all papers in Proc. ACL that mention im.

Back to top.

F-score

Appears in 3 sentences as: F-score (3)
In Improved Lexical Acquisition through DPP-based Verb Clustering
  1. For four out of five conditions its F-score performance outperforms the baselines by 42-83%.
    Page 6, “Evaluation”
  2. These are the Most Frequent SCF (O’Donovan et al., 2005) which uniformly assigns to all verbs the two most frequent SCFs in general language, transitive (SUBJ-DOBJ) and intransitive (SUBJ) (and results in poor F-score ), and a filtering that removes frames with low corpus frequencies (which results in low recall even when trying to provide the maximum recall for a given precision level).
    Page 7, “Evaluation”
  3. The task we address is therefore to improve the precision of the corpus statistics baseline in a way that does not substantially harm the F-score .
    Page 7, “Evaluation”

See all papers in Proc. ACL 2013 that mention F-score.

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

Back to top.

probabilistic models

Appears in 3 sentences as: probabilistic models (3)
In Improved Lexical Acquisition through DPP-based Verb Clustering
  1. We show how to utilize Determinantal Point Processes (DPPs), elegant probabilistic models that are defined over the possible subsets of a given dataset and give higher probability mass to high quality and diverse subsets, for clustering.
    Page 1, “Abstract”
  2. Our framework is based on Determinantal Point Processes (DPPs, (Kulesza, 2012; Kulesza and Taskar, 2012c)), elegant probabilistic models that are defined over the possible subsets of a given dataset and give higher probability mass to high quality and diverse subsets.
    Page 2, “Introduction”
  3. Determinantal point processes (DPPs) are elegant probabilistic models of repulsion that offer efficient and exact algorithms for sampling, marginalization, conditioning, and other inference tasks.
    Page 3, “The Unified Framework”

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

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

Back to top.