Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.06227 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908861003202560 |
|---|---|
| author | Mao, Junyu Hills, Anthony Tseriotou, Talia Liakata, Maria Shamir, Aya Sayda, Dan Atzil-Slonim, Dana Djohari, Natalie Mandal, Arpan Roth, Silke Ugwudike, Pamela Niranjan, Mahesan Middleton, Stuart E. |
| author_facet | Mao, Junyu Hills, Anthony Tseriotou, Talia Liakata, Maria Shamir, Aya Sayda, Dan Atzil-Slonim, Dana Djohari, Natalie Mandal, Arpan Roth, Silke Ugwudike, Pamela Niranjan, Mahesan Middleton, Stuart E. |
| contents | Real-world indicators play an important role in many natural language processing (NLP) applications, such as life-event for mental health analysis and risky behaviour for online safety, yet labelling such information in training datasets is often costly and/or difficult due to their dynamic nature. Large language models (LLMs) show promising potential for automated annotation, yet multi-label prediction remains challenging. In this work, we propose a Confidence-Aware Fine-Grained Debate (CFD) framework that simulates collaborative annotation using fine-grained information to better support automated multi-label enrichment. We introduce two new expert-annotated resources: A mental health Reddit well-being dataset and an online safety Facebook sharenting risk dataset. Experiments show that CFD achieves the most robust enrichment performance compared to a range of baseline approaches. We further evaluate various training-free enrichment incorporation strategies and demonstrate that LLM-enriched indicators consistently improves our downstream tasks. Enriched features incorporated via debate transcripts yield the largest gains, outperforming the non-enriched baseline by 9.9\% on the online safety task. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_06227 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Automated Data Enrichment using Confidence-Aware Fine-Grained Debate among Open-Source LLMs for Mental Health and Online Safety Mao, Junyu Hills, Anthony Tseriotou, Talia Liakata, Maria Shamir, Aya Sayda, Dan Atzil-Slonim, Dana Djohari, Natalie Mandal, Arpan Roth, Silke Ugwudike, Pamela Niranjan, Mahesan Middleton, Stuart E. Computation and Language Machine Learning Real-world indicators play an important role in many natural language processing (NLP) applications, such as life-event for mental health analysis and risky behaviour for online safety, yet labelling such information in training datasets is often costly and/or difficult due to their dynamic nature. Large language models (LLMs) show promising potential for automated annotation, yet multi-label prediction remains challenging. In this work, we propose a Confidence-Aware Fine-Grained Debate (CFD) framework that simulates collaborative annotation using fine-grained information to better support automated multi-label enrichment. We introduce two new expert-annotated resources: A mental health Reddit well-being dataset and an online safety Facebook sharenting risk dataset. Experiments show that CFD achieves the most robust enrichment performance compared to a range of baseline approaches. We further evaluate various training-free enrichment incorporation strategies and demonstrate that LLM-enriched indicators consistently improves our downstream tasks. Enriched features incorporated via debate transcripts yield the largest gains, outperforming the non-enriched baseline by 9.9\% on the online safety task. |
| title | Automated Data Enrichment using Confidence-Aware Fine-Grained Debate among Open-Source LLMs for Mental Health and Online Safety |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2512.06227 |