Shallow Semantic Parsing of Persian Sentences

Shallow Semantic Parsing of Persian Sentences* Azadeh Kamel Ghalibaf a, Saeed Rahati a, and Azam Estaji b a Department of Artificial Intelligence, Azad University of Mashhad, Ostad Yousefi - Ghasem Abad - Mashhad - Iran (0098511-6627512) Azadeh_Kamel@hotmail.com bDepartment of Linguistics, Ferdowsi University of Mashhad, Azadi Square- Mashhad-Iran- PostCode: 9177948974 Estaji@um.ac.ir Abstract. Extracting semantic roles is one of the major steps in representing text meaning. It refers to finding the semantic relations between a predicate and syntactic constituents in a sentence. In this paper we present a semantic role labeling system for Persian, using memory-based learning model and standard features. We show that good semantic parsing results can be achieved with a small 1300-sentence training set. In order to extract features, we developed a shallow syntactic parser which divides the sentence into segments with certain syntactic units. The input data for both systems is drawn from Hamshahri corpus which is hand-labeled with required syntactic and semantic information. The results show an F-score of 90.3% on argument boundary detection task and an F-score of 87.4% on semantic role labeling task using Gold-standard parses. An overall system performance shows an F-score of 83.8% on complete semantic role labeling system i.e. boundary plus classification. Keywords: Semantic Role Labeling, Shallow Semantic Parsing, Shallow Syntactic Parsing, Memory-Based Learning. Copyright 2009 by Azadeh Kamel Ghalibaf, Saeed Rahati, and Azam Estaji 1 Introduction Semantic role labeling (SRL), also called shallow semantic parsing, involves identifying which groups of words (phrases) act as the arguments to a given predicate. These arguments must be labeled with the role they play in relation to the predicate (verb), indicating how the proposition should be semantically interpreted (Hacioglu, 2004). A number of algorithms have been proposed for automatically assigning such shallow semantic structure to English sentences. But little is understood about how these algorithms may perform in other languages, and in general the role of language-specific idiosyncrasies in the extraction of semantic content, and how to train these algorithms when large hand-labeled training sets are not available (Sun and Jurafsky, 2004). So, to design an optimal model for a Persian SRL system we should take into account specific linguistic aspects of the language. Regarding the remarkable amount of research that has already been done in English, we can capitalize from it to design a basic and effective SRL system. The idea is to use the technology for English and verify if it is suitable for Persian. Our proposed SRL system implements a two-phase architecture to first identify the arguments by a shallow syntactic parser or chunker, and then to label them with appropriate semantic role, with respect to the predicate of the sentence. We treat both phases as a multi- class classification problem, where the classifier is trained in a supervised manner, from 150 23rd Pacific Asia Conference on Language, Information and Computation, pages 150–159 human-annotated data, using memory-based learning. To our knowledge it is the first corpus based SRL system for Persian. Memory-based language processing is based on the idea that NLP problems …

Content truncated to 500 words