Open Domain Question Answering Based on Retriever-Reader Architecture

Open-domain Question Answering (OpenQA) is a trendy research hotspot in the field of Natural Language Processing (NLP). In the past few years, with the in-depth research on Deep Learning (DL) and Pre-training Language Modules (PLM), the research on OpenQA has made rapid progress. OpenQA aims to find relevant documents from a large-scale text corpus and then extract or generate correct answers. In this paper, we introduce the development and research status of OpenQA based on “Retriever-Reader.” Firstly, we briefly expound on the relevant theories and traditional architecture. Then the function and category of the Retriever and Reader and the related existing systems are analyzed. Retriever can be regarded as an Information retrieval (IR) system. The reader can be viewed as a machine reading comprehension system (MRC). The fourth section gives a relatively simple introduction to the latest end-to-end training methods and their related types. Finally, we summarize the content of the paper and look forward to the future development trend.KeywordsOpen-domain question answeringInformation retrievalMachine reading comprehension