BiCon-Gate: Consistency-Gated De-colloquialisation for Dialogue Fact-Checking
For dialogue fact-checking systems, this work provides a method to handle colloquial language, yielding measurable gains in retrieval and verification accuracy.
The paper tackles automated fact-checking in dialogue by addressing colloquial language through a staged de-colloquialisation method with a consistency gate (BiCon-Gate). On the DialFact benchmark, it improves evidence retrieval and fact verification, outperforming baselines including a one-shot LLM rewrite.
Automated fact-checking in dialogue involves multi-turn conversations where colloquial language is frequent yet understudied. To address this gap, we propose a conservative rewrite candidate for each response claim via staged de-colloquialisation, combining lightweight surface normalisation with scoped in-claim coreference resolution. We then introduce BiCon-Gate, a semantics-aware consistency gate that selects the rewrite candidate only when it is semantically supported by the dialogue context, otherwise falling back to the original claim. This gated selection stabilises downstream fact-checking and yields gains in both evidence retrieval and fact verification. On the DialFact benchmark, our approach improves retrieval and verification, with particularly strong gains on SUPPORTS, and outperforms competitive baselines, including a decoder-based one-shot LLM rewrite that attempts to perform all de-colloquialisation steps in a single pass.