CLARINET: Augmenting Language Models to Ask Clarification Questions for Retrieval

TL;DR

CLARINET is a system that enhances language models to generate clarification questions, improving retrieval success by reducing ambiguity in user queries through an end-to-end fine-tuned approach.

Contribution

It introduces a novel method to augment large language models for asking effective clarification questions in retrieval tasks, outperforming traditional heuristics and baseline prompts.

Findings

Outperforms heuristics like information gain by 17%.

Outperforms vanilla LLM prompts by 39%.

Effective in real-world book search dataset.

Abstract

Users often make ambiguous requests that require clarification. We study the problem of asking clarification questions in an information retrieval setting, where systems often face ambiguous search queries and it is challenging to turn the uncertainty in the retrieval model into a natural language question. We present CLARINET, a system that asks informative clarification questions by choosing questions whose answers would maximize certainty in the correct candidate. Our approach works by augmenting a large language model (LLM) to condition on a retrieval distribution, finetuning end-to-end to generate the question that would have maximized the rank of the true candidate at each turn. When evaluated on a real-world retrieval dataset of users searching for books, our system outperforms traditional heuristics such as information gain on retrieval success by 17% and vanilla-prompted LLMs by 39% relative.