A Rule Based Open Information Extraction Method Using Cascaded Finite-State Transducer

Hailun Lin, Yuanzhuo Wang, Peng Zhang, Weiping Wang, Yinliang Yue, Zheng Lin

Abstract: In this paper, we present R-OpenIE, a rule based open information extraction method using cascaded finite-state transducer. R-OpenIE defines contextual constraint declarative rules to generate relation extraction templates, which frees from the influence of syntactic parser errors, and it uses cascaded finite-state transducer model to match the satisfied relational tuples. It is noted that R-OpenIE creates inverted index for each matched state during the matching process of cascaded finite-state transducer, which improves the efficiency of pattern matching. The experimental results have shown that our R-OpenIE can achieve good adaptability and efficiency for open information extraction.