AmharicIR+Instr: A Two-Dataset Resource for Neural Retrieval and Instruction Tuning
Tilahun Yeshambel, Moncef Garouani, Josiane Mothe
TL;DR
This work tackles the scarcity of high-quality data for Amharic neural retrieval and instruction tuning by releasing two datasets: 1,091 manually verified Amharic query–document triplets for dense retrieval benchmarking and 6,285 Amharic GPT-style prompt–response pairs for instruction following. It introduces a reusable, native-speaker–validated data-construction methodology for low-resource languages, combining expert authoring, web harvesting, and LLM generation with explicit supervision, normalization, and deduplication. Baseline experiments using SPLADE, RoBERTa, and ColBERT demonstrate competitive retrieval performance and validate the utility of the Amharic triplet dataset for training diverse neural architectures, with RoBERTa generally performing best. By standardizing formats and providing documented splits, the releases enable reproducible Amharic IR and generation research and offer a generalizable approach applicable to other low-resource languages.
Abstract
Neural retrieval and GPT-style generative models rely on large, high-quality supervised data, which is still scarce for low-resource languages such as Amharic. We release an Amharic data resource consisting of two datasets that supports research on (i) neural retrieval-ranking and (ii) instruction-following text generation. The retrieval-ranking dataset contains 1,091 manually verified query-positive-negative document triplets drawn from diverse Amharic sources and constructed to support contrastive training and benchmarking of neural retrievers (e.g., DPR, ColBERT-style late interaction and SPLADE-style sparse neural retrieval). Triplets are created through a combination of expert-curated queries, web-derived queries, and LLM-assisted generation, with positive/negative documents selected from the web or synthesized by LLMs and then validated by native speakers. The instruction prompt-response dataset comprises 6,285 Amharic prompt-response pairs spanning multiple domains and instruction types, generated with several LLMs and refined through manual review and correction for grammaticality, relevance, fluency, and factual plausibility. We release both datasets with standardized splits and formats (CSV,JSON,JSONL) to enable reproducible work on Amharic retrieval, ranking, and generative modelling. These datasets also come with a methodology that can be generalized to other low-resource languages.