Scholarlink

In the Cisco Webex Enterprise Suite, the sheer volume of data produced through messaging, meetings, videos, documents and organizational data necessitates more effective search. In this paper, we propose a method for personalizing enterprise search results by leveraging Large Language Models for context aware topic hierarchy extraction of textual data and combine it with a Relational Graph Attention Network model to capture the relationships between users, topics and content within a cross domain enterprise knowledge graph. Combining the relationship between users, topics, content metadata and organizational hierarchy, enables better personalization, balancing user privacy and search performance with results showing 25% improvement in personalization for video search.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Personalizing Enterprise Search with LLM Populated Attributes in Graph Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Christopher Liu,
  • Varsha Embar

摘要

In the Cisco Webex Enterprise Suite, the sheer volume of data produced through messaging, meetings, videos, documents and organizational data necessitates more effective search. In this paper, we propose a method for personalizing enterprise search results by leveraging Large Language Models for context aware topic hierarchy extraction of textual data and combine it with a Relational Graph Attention Network model to capture the relationships between users, topics and content within a cross domain enterprise knowledge graph. Combining the relationship between users, topics, content metadata and organizational hierarchy, enables better personalization, balancing user privacy and search performance with results showing 25% improvement in personalization for video search.