PROM: Personal Knowledge Graph Construction with Large Language Models

Aishan Maoliniyazi, Chaohong Ma, Xiaofeng Meng, Bingbing Xu

2025

Abstract

The growing volume of digital information requires effective Personal Knowledge Management. Personal Knowledge Graphs (PKGs), which model knowledge as connected entities and relationships, show potential. Chats or natural voice conversations contain abundant context information about users’ thoughts and preferences, which is beneficial for constructing PKGs. However, constructing PKGs from unstructured natural conversations is still challenging. The main obstacle comes from two aspects: inherently complex and context-dependent conversations. In this paper, we present PROM, a novel framework of personal knowledge graph construction with LLMs. PROM effectively constructs PKGs from natural conversations. Particularly, PROM constructs PKGs with rich knowledge information, preserves context information for knowledge provenance, and fuses different kinds of contexts for structural and semantic coherence. Specifically, PROM constructs knowledge triples (subject, predicate, object) from conversational text and integrates them into a coherent PKG with the help of LLMs. We propose a multi-strategy knowledge fusion technique to resolve conflicts and unify information from different sources for structural and semantic consistency. Moreover, we design an API proxy engine to facilitate consistent knowledge extraction from different LLM backends. The proxy system is flexible and cost-effective. It can adapt different triple extraction strategies from LLMs and unify the results with a knowledge fusion strategy. We evaluate PROM in different scenarios. The experiments show that PROM is able to construct comprehensive and context-aware PKGs from unstructured conversations and can support personal knowledge discovery.

Paper Citation

in Harvard Style

Maoliniyazi A., Ma C., Meng X. and Xu B. (2025). PROM: Personal Knowledge Graph Construction with Large Language Models. In Proceedings of the 17th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 2: KMIS; ISBN 978-989-758-769-6, SciTePress, pages 301-312. DOI: 10.5220/0013830900004000

in Bibtex Style

@conference{kmis25,
author={Aishan Maoliniyazi and Chaohong Ma and Xiaofeng Meng and Bingbing Xu},
title={PROM: Personal Knowledge Graph Construction with Large Language Models},
booktitle={Proceedings of the 17th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 2: KMIS},
year={2025},
pages={301-312},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013830900004000},
isbn={978-989-758-769-6},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 17th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 2: KMIS
TI - PROM: Personal Knowledge Graph Construction with Large Language Models
SN - 978-989-758-769-6
AU - Maoliniyazi A.
AU - Ma C.
AU - Meng X.
AU - Xu B.
PY - 2025
SP - 301
EP - 312
DO - 10.5220/0013830900004000
PB - SciTePress