Ant Group Insurance Technology Wins First Place in Fine‑Grained Dialogue Social Bias Detection at NLPCC 2023

At the 11th CCF Natural Language Processing and Chinese Computing Conference (NLPCC), Ant Group's insurance technology team won first place in the "Fine‑Grained Dialogue Social Bias Identification" evaluation, leveraging years of dialogue understanding expertise in the insurance domain.

The NLPCC competition featured seven tasks, including knowledge‑based QA, speech entity linking, multimodal product summarization, and the bias detection task, which aims to locate and identify socially biased language in conversations to improve service fairness.

With the rapid rise of intelligent service assistants, Ant Group emphasizes that AI‑driven language understanding systems must be trained on large dialogue datasets while carefully preprocessing data to eliminate biased language, thereby reducing algorithmic prejudice and enhancing emotional intelligence.

The team has also developed its own pre‑trained language model, AntInsBert, now applied across product recommendation, underwriting, claims processing, and service quality inspection. Their "Claim Brain" technology, opened to insurers in 2020, can recognize over 107 types of medical and claim documents, increasing claim processing efficiency by 70%.