A Structured SVM Semantic Parser Augmented by Semantic Tagging with Conditional Random Field

Proceedings of PACLIC 19, the 19th Asia-Pacific Conference on Language, Information and Computation. A Structured SVM Semantic Parser Augmented by Semantic Tagging with Conditional Random Field Minh Le Nguyen Graduate Information School Japan Advanced Institute of Science and Technology Address: 1-1, Asahidai, Nomi, Ishikawa 923-1292 Japan E-mail: nguyenml@jaist.ac.jp Akira Shimazu Graduate Information School Japan Advanced Institute of Science and Technology Address: 1-1, Asahidai, Nomi, Ishikawa 923-1292 Japan E-mail: shimazu@jaist.ac.jp Hieu Xuan Phan Graduate Information School Japan Advanced Institute of Science and Technology Address: 1-1, Asahidai, Nomi, Ishikawa 923-1292 Japan E-mail: hieuxuan@jaist.ac.jp Abstract This paper presents a novel method of semantic parsing that maps a natural language (NL) sentence to a logical form. We propose a semantic parsing method by conducting separately two steps as follows; 1) The first step is to predict semantic tags for a given input sentence. 2) The second step is to build a semantic representation structure for the sentence using the sequence of semantic tags. We formulate the problem of semantic tagging as a sequence learning using a conditional random field models (CRFs). We then represent a tree structure of a given sentence in which syntactic and semantic information are integrated in that tree. The learning problem is to map a given input sentence to a tree structure using a structure support vector model. Experimental results on the CLANG corpus show that the semantic tagging performance achieved a sufficiently high result. In addition, the precision and recall of mapping NL sentences to logical forms i.e. the meaning representation in CLANG show an improvement in comparison with the previous work. 1. Introduction Semantic parsing is an interesting problem in NLP as it would very likely be part of any interesting NLP applications. For example, the ability to map natural language to a formal query or command language is critical to developing more user-friendly interfaces. Traditional approaches to constructing database interfaces require an expert to hand-craft an appropriate semantic parser (Allen, 95). However, such hand-crafted parsers are time consuming to develop and suffer from problems with robustness and incompleteness, even for domain specific applications. Recent approaches have focused on using machine learning methods on the corpus of sentences and semantic annotations to map natural language sentences (NL) to a complete formal language including cite (Miller, 96)(Zelle,96)(L.R.Tang, 2003). These machine learning methods are the application of inductive logic programming (Zelle, 96) (L.R. Tang, 2003), the adaptation of transformation based learning (R.J. Kate, et.al. 2005) to semantic parsing, and the modification of a head-driven parsing model (R.Ge and R.J. Mooney 2005). This paper addresses a novel approach to the problem of semantic parsing by considering it as a structure classification problem with a support vector machine model (SVM). We describe a tree structure in which semantic information and syntactic information are incorporated and the semantic representation of a sentence, such as first order logic or CLANG, can be interpreted by the tree. With this flexible structure, the problem of semantic parsing is converted to the problem of how we can …

Content truncated to 500 words