Open-Domain Question Answering with Pre-Constructed Question Spaces
Jinfeng Xiao University of Illinois at Urbana-Champaign \AndLidan Wang Adobe Inc. \AndFranck Dernoncourt Adobe Inc. \AndTrung Bui Adobe Inc. \AndTong Sun Adobe Inc. \AndJiawei Han University of Illinois at Urbana-Champaign
Abstract
Open-domain question answering aims at solving the task of locating the answers to user-generated questions in massive collections of documents. There are two families of solutions available: retriever-readers, and knowledge-graph-based approaches. A retriever-reader usually first uses information retrieval methods like TF-IDF to locate some documents or paragraphs that are likely to be relevant to the question, and then feeds the retrieved text to a neural network reader to extract the answer. Alternatively, knowledge graphs can be constructed from the corpus and be queried against to answer user questions. We propose a novel algorithm with a reader-retriever structure that differs from both families. Our reader-retriever first uses an offline reader to read the corpus and generate collections of all answerable questions associated with their answers, and then uses an online retriever to respond to user queries by searching the pre-constructed question spaces for answers that are most likely to be asked in the given way. We further combine retriever-reader and reader-retriever results into one single answer by examining the consistency between the two components. We claim that our
中文速览
开放域问答（Open-domain QA）的核心难题在于：如何从海量文档中又快又准地找到用户问题的答案——现有的"先检索后阅读"方案总在效率与精度之间左右为难，而基于知识图谱的方案又对实体识别错误过于敏感。这篇论文提出了一种反过来的新思路——"读者-检索者"（reader-retriever）结构：在离线阶段，先用深度神经网络把整个语料库读一遍，自动生成所有可回答的问题并与对应答案配对，构建成"问题空间"（question space）；用户提问时，只需在线把问题与预先构建好的问题空间做语义匹配，找出最相近的已有问题，从而返回答案。此外，论文还将这套新结构与传统检索-阅读方案的结果通过一致性判断融合为最终答案，形成称作 R6 的完整模型。在公开数据集上的实验表明，R6 的准确率大幅超越现有最优方法，同时规避了检索精度低和知识图谱脆弱性两大瓶颈，为开放域问答提供了一条更可靠、更具扩展性的新路径。
原文 arXiv:2006.08337;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2006.08337v2