System of automatic homography resolution based on the semantic connection words of adjacent sentences in a text passage

Authors: Chemerilov V. V., Fadeev A. S.

Annotation: This paper describes a technique to automatically resolve homographs in interrogative sentences based on a semantic connection with the answer. The results of the semantic analysis of the Semsin parser are used as input data. Based on this technique, a software algorithm was developed that allows to mark automatically homographs with an SSML tag that changes the acoustic characteristics of synthesized speech. On a special test sample consisting of phrase units, an experiment conducted to evaluate the quality of the algorithm, which gave positive results. In the course of work, shortcomings related to errors in the semantic analysis of the Semsin parser and the inability to recognize the context within a single sentence were revealed.

Keywords: russian text-to-speech systems, semantic connection of sentences, homography resolution, homographs, interrogative sentence, super phrase units, semsin parser