TY - GEN
T1 - Overview of the NLPCC 2025 Shared Task 8
T2 - 14th National CCF Conference on Natural Language Processing and Chinese Computing, NLPCC 2025
AU - Jin, Zhengda
AU - Wang, Bingbing
AU - Tu, Geng
AU - Xu, Ruifeng
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2026
Y1 - 2026
N2 - This paper provides a detailed overview of the NLPCC 2025 Shared Task 8, which focused on personalized emotional support conversation. Addressing the dual challenges of acquiring high-quality, nuanced datasets for emotional support and the limitations of generic Large Language Model (LLM) responses, this task aims to encourage the development of systems capable of delivering empathetic and contextually relevant support by taking individual user characteristics into account. The central objective is to train models on a newly created dataset, Uni-Conv, to generate supportive dialogue that acknowledges and responds to users’ unique profiles and situations. This paper outlines the motivation behind the task, the creation and specific details of the UniConv dataset, the evaluation framework employed, the approaches taken by the participating teams, and a summary of the achieved results. Looking ahead, we anticipate that continued exploration and refinement of personalized modeling approaches will pave the way for the development of next-generation emotional support systems that are more attuned to user needs and more human-centered. A total of 3 teams participated in the task, submitting 12 system results.
AB - This paper provides a detailed overview of the NLPCC 2025 Shared Task 8, which focused on personalized emotional support conversation. Addressing the dual challenges of acquiring high-quality, nuanced datasets for emotional support and the limitations of generic Large Language Model (LLM) responses, this task aims to encourage the development of systems capable of delivering empathetic and contextually relevant support by taking individual user characteristics into account. The central objective is to train models on a newly created dataset, Uni-Conv, to generate supportive dialogue that acknowledges and responds to users’ unique profiles and situations. This paper outlines the motivation behind the task, the creation and specific details of the UniConv dataset, the evaluation framework employed, the approaches taken by the participating teams, and a summary of the achieved results. Looking ahead, we anticipate that continued exploration and refinement of personalized modeling approaches will pave the way for the development of next-generation emotional support systems that are more attuned to user needs and more human-centered. A total of 3 teams participated in the task, submitting 12 system results.
KW - Emotional Support Conversation
KW - Individual Characteristics
KW - Large Language Models
UR - https://www.scopus.com/pages/publications/105046008807
U2 - 10.1007/978-981-95-3352-7_43
DO - 10.1007/978-981-95-3352-7_43
M3 - 会议稿件
AN - SCOPUS:105046008807
SN - 9789819533510
T3 - Lecture Notes in Computer Science
SP - 512
EP - 520
BT - Natural Language Processing and Chinese Computing - 14th National CCF Conference, NLPCC 2025, Proceedings
A2 - Mao, Xian-Ling
A2 - Ren, Zhaochun
A2 - Yang, Muyun
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 7 August 2025 through 9 August 2025
ER -