A FrameNet Annotation Feasibility Research For Chinese Text Frame Semantic Representation

From the perspective of Frame Semantics, this paper examines an English-based frame semantic annotation method on Chinese text. The author carries out FrameNet text annotation discussions both in theory and practice to prove FrameNet feasibility for the Chinese language. Frame Theory was originally proposed by Minsky to represent knowledge in Artificial Intelligence (AI). Fillmore brought the concept into linguistic studies and gradually developed it into a new research field, Frame Semantics. Since 1997, he has chaired the Berkeley FrameNet Project as Principal and Project Director. The project is ongoingly undertaken in the International Computer Science Institute (ICSI) in Berkeley, California, USA. The output of this project is a FrameNet generated from semantically-annotated sentences. Constituents in these sentences are tagged on layers of Frame Element, Grammar Function and Phrase Type etc, in order to find collocation rules for the central verb (Lexical Unit) with its surrounding words (Frame Elements) in each frame.In the realm of Knowledge Representation (KR), the Chinese language has also made breakthroughs in rule-based sentence analyses and corpus-based text syntactic annotation, yet we haven't attempted the above-mentioned semantic annotation to generate FrameNet so far. Thus, this paper focuses on whether the FrameNet annotation method can be employed to represent Chinese text knowledge. Actually, all supportive language layer requirements in Berkeley FrameNet, concerning lexical category, syntax and semantics, are prepared in Chinese linguistic studies, too. In recent years, findings derived from Chinese Basic Sentence Models on Verb Valency, have laid foundations for Chinese sentence frame semantic annotation already as these Basic Sentence Models integrate syntactic and semantic information together into sentence models. From a theoretical point of view, domestic and foreign achievements concerning text KR are similar in origin, uniform in contents and likewise in practice, though differ in terms. FrameNet annotation method, once adapted to character features of the Chinese language, can be applied on Chinese texts as well.In order to prove the application feasibility of FrameNet semantic representation method, the author carries out annotation practice for 24 basic Chinese sentence constituent configuration types, on layers of Syntax, Semantics, Part-of-Speech, Category ID and ID Code. As Basic Sentence Models, the 24 sample sentences, each...