Evaluation of a Standalone Language-independent Dialogue Framework

摘要

This paper presents an innovative dialogue agent designed for textual casual chatting, which can handle any language. The system acquires knowledge from a non-annotated corpus and then represents all the language aspects into a graph. Using previously acquired knowledge it splits sentences into sub-nodes to proceed to a flexible output generation. Moreover, it uses graph clustering to generate node categories without using any grammar-related tags, and uses these categories to induce new knowledge. The system uses the same processing regardless of the language, that makes the system able to handle any language without any adaptation task. In addition, since the system uses only a limited number of resources, it can be set up as a standalone system in order to preserve the user privacy. We carried out dialogue correctness experiments in Chinese, English and Japanese and obtained results comparable to a more language-specific multilingual system.