SEMAFOR: Frame Argument Resolution with Log-Linear Models

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 264–267, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics SEMAFOR: Frame Argument Resolution with Log-Linear Models Desai Chen Nathan Schneider Dipanjan Das Noah A. Smith School of Computer Science, Carnegie Mellon University, Pittsburgh, PA 15213, USA {desaic@andrew,dipanjan@cs,nschneid@cs,nasmith@cs}.cmu.edu Abstract This paper describes the SEMAFOR sys- tem’s performance in the SemEval 2010 task on linking events and their partici- pants in discourse. Our entry is based upon SEMAFOR 1.0 (Das et al., 2010a), a frame-semantic probabilistic parser built from log-linear models. The extended sys- tem models null instantiations, including non-local argument reference. Performance is evaluated on the task data with and with- out gold-standard overt arguments. In both settings, it fares the best of the submitted systems with respect to recall and F1. 1 Introduction The theory of frame semantics (Fillmore, 1982) holds that meaning is largely structured by holis- tic units of knowledge, called frames. Each frame encodes a conventionalized gestalt event or sce- nario, often with conceptual dependents (partic- ipants, props, or attributes) filling roles to elab- orate the specific instance of the frame. In the FrameNet lexicon (Fillmore et al., 2003), each frame defines core roles tightly coupled with the particular meaning of the frame, as well as more generic non-core roles (Ruppenhofer et al., 2006). Frames can be evoked with linguistic pred- icates, known as lexical units (LUs); role fillers can be expressed overtly and linked to the frame via (morpho)syntactic constructions. However, a great deal of conceptually-relevant content is left unexpressed or is not explicitly linked to the frame via linguistic conventions; rather, it is expected that the listener will be able to infer the appro- priate relationships pragmatically. Certain types of implicit content and implicit reference are for- malized in the theory of null instantiations (NIs) (Fillmore, 1986; Ruppenhofer, 2005). A complete frame-semantic analysis of text thus incorporates covert and overt predicate-argument information. In this paper, we describe a system for frame- semantic analysis, evaluated on a semantic role labeling task for explicit and implicit arguments (§2). Extending the SEMAFOR 1.0 frame- semantic parser (Das et al., 2010a; outlined in §3), we detect null instantiations via a simple two-stage pipeline: the first stage predicts whether a given role is null-instantiated, and the second stage (§4) predicts how it is null-instantiated, if it is not overt. We report performance on the SemEval 2010 test set under the full-SRL and NI-only conditions. 2 Data The SemEval 2007 task on frame-semantic pars- ing (Baker et al., 2007) provided a small (about 50,000 words and 2,000 sentences) dataset of news text, travel guides, and bureaucratic accounts of weapons stockpiles. Sentences in this dataset were fully annotated with frames and their argu- ments. The SemEval 2010 task (Ruppenhofer et al., 2010) adds annotated data in the fiction do- main: parts of two Sherlock Holmes stories by Arthur Conan Doyle. The SemEval 2010 train- ing set consists of the SemEval 2007 data plus one document from the new …

Content truncated to 500 words