Crafting the Path: Robust Query Rewriting for Information Retrieval

查询重写旨在生成一个新的查询，以补充原始查询，以提高信息检索系统的效果。最近关于查询重写的研究，如query2doc(Q2D)，query2expand(Q2E)和query2cot(Q2C)，依赖于大型语言模型（LLMs）的内部知识，以生成相关段落以添加信息到查询中。然而，在模型的内在参数中没有封装所需知识的情况下，这些方法的效果可能显著下降。在本文中，我们提出了一种新颖的结构化查询重写方法，称为Crafting the Path，专为检索系统量身定制。Crafting the Path包括一个三步骤的过程，为每个步骤中要搜索的段落制作与查询相关的信息。具体而言，Crafting the Path从查询概念理解开始，继续到查询类型识别，最后进行预期答案提取。实验结果表明，我们的方法优于以前的重写方法，特别是在LLMs的不熟悉领域。我们证明了我们的方法不太依赖模型的内部参数知识，并且生成的查询具有更少的事实不准确性。此外，我们观察到Crafting the Path与基线相比具有更少的延迟。

Crafting the Path: Structured Query Rewriting for Information Retrieval Systems

Crafting the Path is a novel structured query rewriting method that involves a three-step process for finding relevant passages to add information to a query. It is less dependent on the internal parameter knowledge of the model and generates queries with fewer factual inaccuracies. It outperforms previous rewriting methods, especially in less familiar domains for LLMs.

The three-step process involves Query Concept Comprehension, Query Type Identification, and Expected Answer Extraction. Experimental results show that Crafting the Path has less latency compared to the baselines. The paper provides a detailed explanation of the methodology and evaluation metrics used. The proposed method can potentially improve the performance of information retrieval systems in various domains.

Recent studies on query rewriting include query2doc (Q2D), query2expand (Q2E) and querey2cot (Q2C), which rely on the internal knowledge of Large Language Models (LLMs). Other related works in the field of information retrieval include 'BERT re-ranker: Reducing redundancy in passage retrieval using contextualized term importance' and 'Learning to Retrieve Reasoning Paths over Wikipedia Graph for Question Answering'.