Timur Seidalin

Morphology Beats Multilingual Embeddings for Kazakh Retrieval: A 300-Query Benchmark with Honest Negative Results

2026 · Zenodo (CERN European Organization for Nuclear Research)

We present a reproducible retrieval benchmark for Kazakh — an agglutinative, low-resource language — comprising 300 queries over 8,370 Wikipedia passages across three query categories: inflected, natural, and vocabulary-gap. We evaluate five retrieval systems: BM25 …