Question-answering method and system

Abstract

The invention discloses a question-answering method and a question-answering system. The method comprises the following steps: taking a HitpotQA data set as a training sample for training to form a question-answering model; inputting a proposed question into the question and answer model; searching a question answer from a knowledge source and marking content related to the answer in a searched text paragraph by the the question and answer model ; and editing the marked content into answers and outputting the answers. According to the method, the question and answer model is formed by training the training sample through the HitpotQA data set, the question and answer model has the human-like logical reasoning capability, the content related to the answer can be rapidly labeled from the searched text paragraph, the labeled content is edited into the answer, and the knowledge question and answer response speed and the answer giving accuracy are greatly improved.

Description

Technical field[0001]The present invention relates to the field of natural language processing, and specific to a question and answer method and a question and answer system.Background technique[0002]In the time of this information explosion, there are too many new knowledge every day to generate in the form of text (also other modal), and anyone cannot digest the knowledge alone, so when we encounter problems, the search engine input is usually used. Question to search the answer. But the search engine search is often a list of links, we also need to open these link list one by one, and then get what we want from these link contents according to our own reasoning capabilities or we think the correct answer.[0003]In order to quickly find the correct answer, the search engine provides a variety of algorithms to achieve this goal, and achieved great progress. For example, when you search for "How many countries in the world", the search engine will directly display the correct answer,...

AI Technical Summary

Problems solved by technology

But these algorithms still have limitations. When you search for a slightly complicated question, such as, "How long do I need to ride a bicycle to consume the calories of a Big Mac", the search engine cannot directly give an accurate answer. Sometimes you still need to find the correct answer from the list of links given by the search engine

Images