SEUPD@CLEF Team RISE at LongEval: Improving Search by Crafting Titles and Matching URLs

This report outlines the development of the RISE group’s Information Retrieval (IR) system for the LongEval-WebRetrieval CLEF 2025 Lab. The objective was to design an efficient, scalable search engine capable of handling large-scale French collections with a focus on consistent performance. The proposed system incorporates a modular architecture, including a parser, an analyzer, an indexer and a searcher, then also query translation and expansion using the Gemini LLM, and a non-neural reranking component to enhance retrieval quality. Emphasis was put on optimizing indexing and searching speed through multi-threading, improving relevance via crafting a title for each document and an URL-based document boosting based on the alignment between user queries and the document’s URL. The evaluation has followed a stepwise enhancement approach, beginning with a Lucene-based baseline.