Linguistic Knowledge-driven Approach to Chinese Comparative Elements Extraction

The BI ( 比 )-structure, which highlights a contrasting characteristic between two items, is the key comparative sentence structure in Chinese. In this paper, we explore the methods of extracting the 6 constituents of the BI-structure. Previous studies are often restricted to probabilistic classification methods, where the feature used hardly embodies linguistic knowledge, therefore unintuitive. As an alternative, we propose the use of two linguistic knowledge-driven approaches, namely the POS chunking-based and TBL-based methods. The first model effectively captures grammatical restrictions over POS sequential patterns. The second model set up on new and lesser templates performs better than Brill’s (1995). Experimental results show that the proposed models are simple and effective methods for Chinese comparative element extraction task.