Detecting Semantic-level Polysemy Ambiguity by Fusing External Semantic Knowledge

Sentence with polysemous words can easily be interpreted as different meanings by different people even in a specific context, which will significantly reduce the quality of requirements documents. However, existing ambiguity detection mainly considers the sentence structure and cannot solve this problem well. In this paper, we consider sentence semantics and perform sentence ambiguity detection by introducing external semantic knowledge to explicitly modeling the different semantics of polysemy. Specifically, we determine the target polysemy in the given sentence according to the ambiguous vocabulary list. Furthermore, based on the fusion strategy we designed, we model the different semantics of polysemous word by introducing external semantic knowledge and predict the possible semantics of the sentence by measuring the the gap between different semantic fusion results and the original sentence. This is also the first work to introduce external semantic knowledge into ambiguity detection. The experimental results illustrate that our accuracy is higher than the baseline accuracy of ambiguity in existing research.