Embedded Semantic Lexicon Induction with Joint Global and Local Optimization

Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (*SEM 2017), pages 209–219, Vancouver, Canada, August 3-4, 2017. c⃝2017 Association for Computational Linguistics Embedded Semantic Lexicon Induction with Joint Global and Local Optimization Sujay Kumar Jauhar Language Technologies Institute Carnegie Mellon University Pittsburgh, PA, USA sjauhar@cs.cmu.edu Eduard Hovy Language Technologies Institute Carnegie Mellon University Pittsburgh, PA, USA hovy@cs.cmu.edu Abstract Creating annotated frame lexicons such as PropBank and FrameNet is expen- sive and labor intensive. We present a method to induce an embedded frame lexicon in an minimally supervised fash- ion using nothing more than unlabeled predicate-argument word pairs. We hy- pothesize that aggregating such pair se- lectional preferences across training leads us to a global understanding that captures predicate-argument frame structure. Our approach revolves around a novel inte- gration between a predictive embedding model and an Indian Buffet Process pos- terior regularizer. We show, through our experimental evaluation, that we outper- form baselines on two tasks and can learn an embedded frame lexicon that is able to capture some interesting generalities in re- lation to hand-crafted semantic frames. 1 Introduction Semantic lexicons such as PropBank (Palmer et al., 2005) and FrameNet (Baker et al., 1998) contain information about predicate-argument frame structure. These frames capture knowledge about the affinity of predicates for certain types of arguments, their number and their semantic na- ture, regardless of syntactic realization. For example, PropBank specifies frames in the following manner: • eat → [agent]0, [patient]1 • give → [agent]0, [theme]1, [recipient]2 These frames provide semantic information such as the fact that “eat” is transitive, while “give” is ditransitive, or that the beneficiary of one action is a “patient”, while the other is a “recipient”. This structural knowledge is crucial for a num- ber of NLP applications. Information about frames has been successfully used to drive and improve diverse tasks such as information extrac- tion (Surdeanu et al., 2003), semantic parsing (Das et al., 2010) and question answering (Shen and La- pata, 2007), among others. However, building these frame lexicons is very expensive and time consuming. Thus, it remains difficult to port applications from resource-rich languages or domains to data impoverished ones. The NLP community has tackled this issue along two different lines of unsupervised work. At the local token level, researchers have at- tempted to model frame structure by the selec- tional preference of predicates for certain argu- ments (Resnik, 1997; Séaghdha, 2010). For exam- ple, on this problem a good model might assign a high probability to the word “pasta” occurring as an argument of the word “eat”. Contrastingly, at the global type level, work has focussed on inducing frames by clustering pred- icates and arguments in a joint framework (Lang and Lapata, 2011a; Titov and Klementiev, 2012b). In this case, one is interested in associating pred- icates such as “eat”, “consume”, “devour”, with a joint clustering of arguments such as “pasta”, “chicken”, “burger”. While these methods have been useful for sev- eral problems, they also have shortcomings. Se- lectional preference modelling …

Content truncated to 500 words