Research On Semantic Role Labeling For Chinese Nominal Predicates

The objective of shallow semantic parsing is to identify semantic roles for a predicate (either a verb or a noun) in a sentence and, including its agent, patient, time and location etc. As a particular case of shallow semantic parsing, semantic role labeling (SRL) has been widely applied in many natural language processing applications, such as information extraction, question answering, and machine translation. SRL can be divided into SRL for verbal predicates (verbal SRL, in short) and SRL for nominal predicates (nominal SRL, in short). This thesis focuses exclusively on nominal SRL, concerning the following issues:1. Realizing a nominal SRL system. First, different pruning rules are devised for internal roles and external roles respectively in order to filter out impossible chunks as well as keep the possible chunks for semantic roles. Then,feature-based machine learing approaches are adopted to identify the types of semantic roles.2. Exploring the task of nominal SRL, specifically, the impact of Chinese verbial predicates on nominal SRL is investigated.The experiments show that the nominal SRL system based on basic features, internal roles and external roles achieves F1-measure of 69.47 While it achieves 70.88 of F1-measure when Chinese verbial predicates are added, suggessting that Chinese verbial predicates enhance the nominial SRL.