A comparative study on generalization of semantic roles in FrameNet

Abstract

A number of studies have presented machine-learning approaches to semantic role labeling with availability of corpora such as FrameNet and PropBank. These corpora define the semantic roles of predicates for each frame independently. Thus, it is crucial for the machine-learning approach to generalize semantic roles across different frames, and to increase the size of training instances. This paper explores several criteria for generalizing semantic roles in FrameNet: role hierarchy, human-understandable descriptors of roles, semantic types of filler phrases, and mappings from FrameNet roles to thematic roles of VerbNet. We also propose feature functions that naturally combine and weight these criteria, based on the training data. The experimental result of the role classification shows 19.16% and 7.42% improvements in error reduction rate and macro-averaged F1 score, respectively. We also provide in-depth analyses of the proposed criteria.

Frame-Semantic Parsing

“…The LTH system of Johansson and Nugues (2007), our baseline ( §2.4), performed the best in the SemEval'07 task. Matsubayashi et al (2009) trained a loglinear model on the SemEval'07 data to evaluate argument identification features exploiting various types of taxonomic relations to generalize over roles. A line of work has sought to extend the coverage of FrameNet by exploiting VerbNet, WordNet, and Wikipedia (Shi and Mihalcea, 2005;Giuglea and Moschitti, 2006;Pennacchiotti et al, 2008;Tonelli and Giuliano, 2009), and projecting entries and annotations within and across languages (Boas, 2002;Fung and Chen, 2004;Padó and Lapata, 2005;Fürstenau and Lapata, 2009).…”

A frame-based approach for capturing semantics from Arabic text for text-to-sign language MT

“…Many attempts based on FrameNet tenet have been experimented, from Multi-Lingual Lexicon Databases (MLLD) building (Boas 2005;Fung and Chen 2006;Pado and Lapata 2005;Fürstenau and Lapata 2009), information extraction (Moschitti et al 2003;Surdeanu et al 2003), text entailment (Burchardt and Frank 2006;Burchardt et al 2009;Tatu and Moldovan 2005), text categorization (Moschitti 2008), question answering (Narayanan and Harabagiu 2004; Shen and Lapata 2007;Frank et al 2007;Moschitti et al 2007), paraphrase recognition (Padó and Erk 2005), machine translation (Boas 2002;Wu and Fung 2009) and shallow semantic analysis and role labeling (Gildea and Jurafsky 2002;Thompson et al 2003;Fleischman et al 2003;Shi and Mihalcea 2005;Erk and Padó 2006;Giuglea and Moschitti 2006;Johansson and Nugues 2007;Matsubayashi et al 2009;Fürstenau and Lapata 2009;Deschacht and Moens 2009;Lang and Lapata 2011;Titov and Klementiev 2012). 6 Frame based approach to Arabic text semantics In our project, we describe techniques for automatic semantic analysis of Arabic texts with deeper grain level than SRL analysis using a dependency-based deep analysis and following a frame semantics approach. We use a dependency syntax parser which integrates the BAMA analyzer with a lexical semantics analyzer based on AWN database in a single model.…”

Frame blending in specialized language

“…In FrameNet, semantic roles are verb-specific and frame-specific, and are hierarchically related within frames (Matsubayashi et al 2009). Although the role hierarchy in individual frames was expected to generalize semantic roles, this does not seem to have happened (Baldewein et al 2004).…”

“…The set of semantic roles in this study largely coincides with the most general ones in FrameNet. Others are taken from the inventory proposed by Gildea and Jurafsky (2002) and still others from the set of semantic roles in VerbNet (http://verbs.colorado.edu/semlink), which can also be partially mapped onto FrameNet roles (Matsubayashi et al 2009). Table 2 is a non-exhaustive list of these roles.…”