The Number of Terms and Documents for Pseudo-Relevant Feedback for Ad-hoc Information Retrieval

TL;DR

This paper investigates how varying the number of documents and terms affects the performance of pseudo-relevant feedback in Arabic ad-hoc information retrieval, highlighting the importance of parameter selection for effective query reformulation.

Contribution

It provides an analysis of the impact of document and term count parameters on ARF effectiveness specifically for Arabic documents, which was previously underexplored.

Findings

Optimal performance depends on selecting sufficient documents D.

A small set of relevant terms T can significantly improve retrieval.

Some queries are inherently difficult to improve with ARF.

Abstract

In Information Retrieval System (IRS), the Automatic Relevance Feedback (ARF) is a query reformulation technique that modifies the initial one without the user intervention. It is applied mainly through the addition of terms coming from the external resources such as the ontologies and or the results of the current research. In this context we are mainly interested in the local analysis technique for the ARF in ad-hoc IRS on Arabic documents. In this article, we have examined the impact of the variation of the two parameters implied in this technique, that is to say, the number of the documents {\guillemotleft}D{\guillemotright} and the number of terms {\guillemotleft}T{\guillemotright}, on an Arabic IRS performance. The experimentation, carried out on an Arabic corpus text, enables us to deduce that there are queries which are not easily improvable with the query reformulation. In addition, the success of the ARF is due mainly to the selection of a sufficient number of documents D and to the extraction of a very reduced set of relevant terms T for retrieval.