FrameNet-Based Shallow Semantic Parsing with a POS Tagger
Search this Article
In this paper we propose a FrameNet-based shallow semantic parsing without syntactic parsing. Previous studies on shallow semantic parsing utilize the results of syntactic parsing of input sentences as input data. However, syntactic parsing has well-known shortfalls, such as large amount of computation and insufficient accuracy etc… Further-more, when use of syntactic parsing is premised, it limits applicable languages, since good syntactic parser is rarely available. To prevent such undesirable consequences in shallow semantic parsing, we propose to use POS tagger instead of syntactic parsing. Our experiments using FrameNetll data as training and test data showed the same level performance as existing methods using syntactic parsing.
- IEICE technical report. Artificial intelligence and knowledge-based processing
IEICE technical report. Artificial intelligence and knowledge-based processing 104(488), 7-10, 2004-12-07
The Institute of Electronics, Information and Communication Engineers