EAGER: Dynamic Contextual Explanation of Search Results

该研究项目旨在调查和开发Defuddle，这是一种方法和系统，可以分析搜索引擎排名列表顶部的文档，以找到人类可读的解释，解释为什么要为这个查询检索文档，以及与现有技术不同的是，这些文档如何相互关联。结果解释方面的进步将使人们在搜索网络或任何其他文本文档集合时更容易理解发生了什么。考虑到搜索是最常见的在线活动之一，减少的挫败感和节省的时间将是实质性的：对电子商务、信息收集、教育和其他无数搜索用途的影响将是深远的。Defuddle项目将培养1名研究生和1名博士后。它也将被纳入本科和研究生课程的课程作业中。Defuddle根据查询上下文和其他排名靠前的项提供每个搜索结果项的解释。它把搜索引擎当作一个“黑盒子”，这样技术就可以被广泛应用，并且独立于任何特定的引擎。Defuddle专注于设计和训练适当的神经模型来解释搜索结果，而不使用任何预定义的解释模板或显式查询方面。Defuddle为其他与搜索相关的任务奠定了基础，这些任务将从解释中受益匪浅：提供多样化的搜索结果，以便最初显示的结果涵盖所有查询方面；通过解释为什么用户可能更喜欢一个答案而不是另一个答案来改进会话搜索；特别重要的是，提供一种机制来识别有偏见的结果，或者提醒用户有偏见，或者可能调整排名列表来补偿。在人类受试者审查允许的情况下，在Defuddle项目中开发的数据集和软件将在项目结束时或之前免费提供。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This research project aims to investigate and develop Defuddle, an approach and a system that analyzes documents at the top of a search engine’s ranked list to find human-readable explanations for why documents were retrieved for this query and, unlike existing technology, for how the documents relate to each other. The resulting advances in result explanation will make it easier for people to make sense of what happens when they search the web or any other collection of text documents. Given that search is among the most common online activities, the reduced frustration and time savings will be substantial: the impact on e-commerce, information gathering, education, and the other myriad uses of search will be far-reaching. The Defuddle project will train one graduate student and one postdoctoral researcher. It will also be incorporated into coursework at both the undergraduate and graduate courses. Defuddle provides explanations of each search result item in the context of the query and in light of the other top-ranked items. It treats the search engine as a “black box” so that the techniques can be applied broadly and independent of any particular engine. Defuddle focuses on designing and training appropriate neural models for explaining search results, without using any predefined explanation templates or explicit query facets. Defuddle lays the foundation for other search-related tasks that will greatly benefit from explanations: providing diversified search results such that the initially shown results cover the full range of query aspects; improving conversational search by elucidating why a user might prefer one answer over another; and particularly importantly, providing a mechanism to identify biased results, either alerting a user to the bias or possibly adjusting the ranked list to compensate. As permitted by human subjects review, data sets and software developed in the Defuddle project will be made freely available at or before the conclusion of the project.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.