报告题目：Enhancing Transparency and Rationality in AI Inference Process: From Neuro-Symbolic AI to Neuro-Conceptual AI

报 告 人：康鑫  副教授  日本德岛大学

报告时间：2025年3月31日（星期一）15:00-16:00

报告地点：扬子津校区信息学院电工中心N104，线上（腾讯会议258-977-111）

主办单位：信息工程学院（人工智能学院）、江苏省知识管理与智能服务工程研究中心、科学技术处

欢迎广大师生参加!

报告摘要：

This presentation highlights the development of Neuro-Conceptual AI, a next-generation framework that advances transparency and rationality in AI reasoning by integrating neural perception with conceptual knowledge representation. Building upon Neuro-Symbolic AI, which combines neural networks with symbolic logic, Neuro-Conceptual AI introduces structured semantic understanding and domain-grounded reasoning, enabling AI systems to make interpretable and trustworthy decisions in emotionally complex contexts.

We showcase two representative applications: a sentiment analysis model that fuses Transformer-based neural features with LDA-driven symbolic topics through Logic Tensor Networks (LTN), and TAM-SenticNet, a framework for early depression detection that integrates time-aware affective memory with symbolic inference via SenticNet. Both systems exemplify how Neuro-Conceptual AI can transform black-box predictions into logically structured and semantically rich explanations, opening new pathways for reliable affective computing in healthcare, social media analysis, and beyond.

报告人简介：

康鑫，日本德岛大学先进技术与科学研究生院信息科学与智能系统学科副教授，博士（工学）。研究兴趣涵盖情感计算、自然语言处理、图像处理、多模态机器学习、神经符号人工智能（Neuro-Symbolic AI）、神经概念人工智能（Neuro-Conceptual AI）、因果机器学习、主动学习，以及人工智能的可解释性与可信性等方向，特别关注其在心理健康支持和金融论证分析等实际应用中的融合与发展。在IEEE TAFFC、IEEE/ACM TASLP、Applied Soft Computing、Neurocomputing、Big Data Mining and Analytics 等国际期刊，以及 ACL、ICASSP、IJCNN、PRCV等国际会议上发表论文百余篇。现任期刊Cognitive Robotics、Chinese Journal of Information Fusion编委，并担任多个顶级期刊与国际会议的审稿人，IEEE、ACM、ACL、CAAI 等学会会员。