Topic and Description Reasoning Generation based on User-Contributed Comments
Keywords: topic modeling, topic reasoning, large language models
Abstract: We propose Topic and Description Reasoning Generation (TDRG), a text inference and generation method based on user-contributed comments with large language models (LLMs). Unlike summarization methods, TDRG can infer the topic according to comments contributed by different users, and generate a readable description that addresses the issue of the lack of interpretability in traditional topic modeling for text mining. In this paper, we adopted zero-shot and fine-tuning methods to generate topics and descriptions for comments. We use a human-annotated YouTube comment dataset to evaluate performance. Our results demonstrate that the potential of large language models of reasoning the topic and description. Generated topic titles and descriptions are similar to human references in textual semantics, but the words used are different from those of humans.
Primary Area: transfer learning, meta learning, and lifelong learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 11186
Loading