New framework KeySI tunes text embeddings with keyword-based human feedback

Researchers have introduced KeySI, an interaction framework designed to tune text embeddings using human feedback. This system allows users to provide feedback at the feature level by organizing keywords into concept groups, which KeySI then uses to generate supervision signals for model tuning. By focusing on keywords rather than requiring individual document inspection, KeySI aims to lower the barrier for adapting embedding models for domain-specific semantics. A prototype implementation visualizes keywords and document embeddings, supports interactive grouping, and has been evaluated through user studies and quantitative experiments. AI

IMPACT This framework could simplify the adaptation of embedding models for specialized domains, reducing the need for extensive labeled data and technical expertise.

RANK_REASON The cluster contains a research paper detailing a new framework for tuning text embeddings. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• ScienceCast

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →