KSE-Web:低资源高棉语语义搜索的混合检索与LLM查询扩展分析
原标题:KSE-Web: An Analysis of Hybrid Retrieval and LLM-Assisted Query Expansion for Low-Resource Khmer Semantic Search
AI 摘要
KSE-Web 研究针对低资源语言高棉语的语义搜索,构建了包含 3K 篇清洗文档和 300 条人工审核查询的数据集,并评估了字符 n-gram BM25、多语言稠密检索、混合检索及基于 Qwen2.5 的 LLM 查询扩展。实验结果显示 BM25 性能最佳(Recall@10 0.943,nDCG@10 0.876),混合检索表现相当,而稠密检索单独使用效果较差;LLM 查询扩展未超越非扩展方法,但更大模型(Qwen2.5-3B)优于小模型,且扩展可能引入主题漂移和噪声。该研究揭示了 LLM 辅助检索在高棉语中的潜力与局限,数据集将在 GitHub 公开。
正文节选
KSE-Web: An Analysis of Hybrid Retrieval and LLM-Assisted Query Expansion for Low-Resource Khmer Semantic Search Abstract As a low-resource language, Khmer presents several retrieval challenges, including limited annotated data, ambiguous word boundaries, weak support in multilingual embedding models, and frequent mixed Khmer–English usage. This paper presents KSE-Web, an analysis of hybrid retrieval and LLM-assisted query expansion for Khmer semantic search. We construct the dataset from approx