Article Details

Abstract

As the scale and service complexity of communication networks continue to grow, the expansion of network fault Knowledge Graphs faces challenges of data scarcity and quality fluctuations. A self-planning and self-verification method for knowledge graph generation based on Chain-of-Thought (CoT) for operation and maintenance scenarios is proposed. This method employs a small amount of high-quality seed data and structured prompts to drive Large Language Models, and utilizes a three-stage generation process of analysis-planning-synthesis. It employs a two-stage quality optimization strategy comprising screening based on synthetic data similarity distribution and correction based on content verification, aimed at automatically identifying and correcting low-quality samples. Experimental validation in low-resource operator scenarios demonstrates that this method effectively filters out semantically irrelevant and redundant samples, significantly enhancing the consistency and accuracy of synthetic data. The model design quality score improves by over 15% compared to the baseline, enabling controllable scalability of the operation and maintenance Knowledge Graph.

Keywords

synthetic data; chain-of-thought; Large Language Model; network operation and maintenance of operator; Knowledge Graph construction

References

1. [1] Du Chunxiang, Lan Kun, Cui Meng Research on the Response Strategies of Telecommunications Operators under the Background of AI Large Models [J]. Telecommunications Science, 2024 (S2): 1-8
2. [2] Che Wanxiang, Dou Zhicheng, Feng Yansong, etc Natural language processing in the era of big models: challenges, opportunities, and developments [J]. Chinese Science: Information Science, 2023, 53 (9): 1645-1687
3. [3] WANG Y，KORDI Y，MISHRA S，et al. Self-Instruct: Aligning Language Models With Self-Generated Instructions [C]//Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Toronto： Association for Computational Linguistics，2023：13484-13508.
4. [4] CAO L. GraphReason: Enhancing Reasoning Capabilities of Large Language Models through A Graph-Based Verification Approach [C]//Mishra B D, Durrett G, Jansen P, et al. (Eds.). Proceedings of the 2nd Workshop on Natural Language Reasoning and Structured Explanations (@ACL 2024). Bangkok：Association for Computational Linguistics，2024：1-12.
5. [5] LIU Y，PENG X，DU T，et al. ERA-CoT: Improving Chain-of-Thought through Entity Relationship Analysis [C]//Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Bangkok：Association for Computational Linguistics，2024：8780-8794.
6. [6] IBRAHIM N，ABOULELA S，IBRAHIM A，et al. A Survey On Augmenting Knowledge Graphs (KGs) With Large Language Models (LLMs): Models, Evaluation Metrics, Benchmarks, And Challenges [J]. Discover Artificial Intelligence (S2731-0809)，2024，4（1）：76.
7. [7] LIANG Y，WANG J，ZHU H，et al. Prompting Large Language Models With Chain-Of-Thought For Few-Shot Knowledge Base Question Generation [C]//Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. Singapore: Association for Computational Linguistics，2023：4329-4343.
8. [8] JI Z，YU T，XU Y，et al. Towards Mitigating LLM Hallucination Via Self Reflection [C]//Findings of the Association for Computational Linguistics: EMNLP 2023. Singapore：Association for Computational Linguistics，2023：1827-1843.
9. [9] Xu Zhiwei, Li Hailong, Li Bo, etc Overview of AIGC Large Model Evaluation: Enabling Technologies, Security Hazards, and Responses [J]. Computer Science and Exploration, 2024, 18 (9): 2293-2325
10. [10] EL EMAM K，MOSQUERA L，HOPTROFF R. Practical Synthetic Data Generation: Balancing Privacy and the Broad Availability of Data [M]. Sebastopol: O’ Reilly Media，2020：50-60. ISBN：9781492072737.
11. [11] SCHOBER P，BOER C，SCHWARTE L A. Correlation Coefficients：Appropriate Use And Interpretation [J]. Anesthesia & Analgesia (S0003-2999)，2018，126（5）：1763-1768.
12. [12] Huang Qiaojuan, Cao Duan, Wang Ya, etc Extension Method of Event Knowledge Graph Based on Large Language Model [J]. Journal of Software, 2025, 36 (9): 4153-4186

Funding

Hefei Major Science and Technology Research and Development Project (2024SZD005)

About the authors

Chen Zhuo (1979-), male, Han ethnicity, from Zhengzhou, Henan Province, is an expert in the Network Department of China Mobile Communications Group Guangdong Co., Ltd. He is a master's student with research interests in operator network technology and intelligent software; Corresponding author: Jin Shuyi (1997-), female, Han nationality, from Anqing, Anhui Province, master's student, research direction: knowledge computing, natural language processing, and large-scale modeling.