核心概念
SURVEYAGENT is a novel conversational system designed to provide personalized and efficient research survey assistance to researchers by integrating knowledge management, recommendation, and query answering capabilities.
摘要
The paper introduces SURVEYAGENT, a conversational system that aims to facilitate the literature review process for researchers. SURVEYAGENT is equipped with three key modules:
-
Knowledge Management Module:
- Allows SURVEYAGENT to find papers and organize them into collections based on user research interests.
- Provides actions for searching, retrieving, and managing paper collections.
-
Recommendation Module:
- Enables SURVEYAGENT to search for papers through keyword-based queries and recommend similar papers.
- Combines term-based recommendations from arXiv Sanity with semantic filtering using large language models (LLMs).
-
Query Answering Module:
- Empowers SURVEYAGENT to assist users with various queries about specified papers, such as question answering, summarization, and providing reviews.
- Addresses the challenge of long context in academic papers by using a chunk-based approach to leverage long-context LLMs.
The paper presents quantitative experiments to validate SURVEYAGENT's capabilities in action planning, paper recommendation, and query answering. It also includes qualitative case studies demonstrating SURVEYAGENT's effectiveness in streamlining the research literature review process.