Turkish Text Retrieval Experiments Using Lemur Toolkit

TL;DR

This paper evaluates Turkish text retrieval using the Lemur Toolkit, demonstrating that language-specific preprocessing improves retrieval quality and that the Language Modeling approach performs best with such preprocessing.

Contribution

It is the first to compare multiple retrieval models on Turkish using Lemur Toolkit and highlights the effectiveness of language-specific preprocessing techniques.

Findings

Language-specific preprocessing improves retrieval quality

Language Modeling approach outperforms other models with preprocessing

All models benefit from preprocessing techniques

Abstract

We used Lemur Toolkit, an open source toolkit designed for Information Retrieval (IR) research, for our automated indexing and retrieval experiments on a TREC-like test collection for Turkish. We study and compare three retrieval models Lemur supports, especially Language modeling approach to IR, combined with language specific preprocessing techniques. Our experiments show that all retrieval models benefits from language specific preprocessing in terms of retrieval quality. Also Language Modeling approach is the best performing retrieval model when language specific preprocessing applied.