An unsupervised syntax disambiguation method combined with the context-sensitive probability

To address the limitations of probabilistic context free grammar, the context information is used and a probabilistic estimation function of syntax structure combined the cooccurrence information of part of speech and syntax category is proposed in this paper. The inside-outside algorithm is used to obtain the probabilities of the syntactic rules and the structure co-occurrences from the raw materials, which can address the bottleneck of supervised learning that large-scale treebank is expensive to create. The experimental results show that the proposed method can effectively improve the precision of syntax disambiguation.