Свитанько Елизавета Игоревна
Multiple-choice Question Answering in the Russian Language
Прикладная математика и информатика
Natural language processing is a powerful machine learning tool for understanding and handling human language. Over the last decades, Question-Answering systems (QA) have become increasingly popular in the Natural language processing field of machine learning. Such systems can be used in a wide variety of applications, e.g., open-domain question answering or closed-domain question answering, which involves searching for the answers within a fixed domain. In this paper, we explore the idea of using the sentence embedding similarity as a technique to answer questions from the provided corpus. The proposed self-collected dataset consists of pairs of Russian test open-domain questions with multi-choice answers. To the author's knowledge, for the first time, the introduced methodologies are used for Russian language and might be applied to many other morphologically rich Slavonic languages in the future.
Текст работы (работа добавлена 19 мая 2019г.)