CROWN: Conversational Passage Ranking by Reasoning over Word Networks
Magdalena Kaiser Rishiraj Saha Roy Gerhard Weikum
Abstract
Information needs around a topic often cannot be satisfied in a single turn; users typically ask follow-up questions referring to the same theme. A system must be capable of understanding the conversational context of a request to retrieve correct answers. In this paper, we present our submission to the TREC Conversational Assistance Track (CAsT) 2019, in which such a conversational setting is explored. We propose an unsupervised method for conversational passage ranking by formulating the passage score for a query as a combination of similarity and coherence. To be specific, passages are preferred that contain words semantically similar to the words used in the question, and where such words appear close by. We built a word proximity network (WPN) from a large corpus, where words are nodes and there is an edge between two nodes if they co-occur in the same passages in a statistically significant way, within a context window. Our approach, named CROWN, achieved above-average performance on the TREC CAsT data with respect to AP@5 and nDCG@1000.
中文速览
对话式搜索（conversational search）中，用户往往连续追问同一主题，后续问题常含指代或省略，系统需要理解多轮上下文才能返回准确段落。作者提出了一种名为 CROWN 的无监督方法，核心思路是把段落得分拆成"相似性"和"连贯性"两部分：先用词向量余弦相似度衡量段落词汇与问题词汇的语义接近程度，再借助从大规模语料库预先构建的词邻近网络（word proximity network，WPN）——该网络以词语共现的统计显著性（NPMI）为边权——奖励那些与查询相关的词对在同一段落中紧密共现的段落，同时结合多轮查询扩展策略引入历史轮次上下文。在 TREC CAsT 2019 评测上，CROWN 在 AP@5 和 nDCG@1000 两项指标上均超过大会提供的 Indri 基线，并达到参赛系统的平均水平以上，验证了在无需标注数据的条件下利用词语网络结构处理多轮对话检索的可行性。
原文 arXiv:1911.02850;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1911.02850v3