TeleDoCTR: Domain-Specific and Contextual Troubleshooting for Telecommunications

TL;DR

TeleDoCTR tackles end-to-end telecom ticket troubleshooting by integrating domain-specific ranking and generative models for routing, retrieval, and fault-analysis generation. It employs cross-modal domain-adapted embeddings, parameter-efficient fine-tuning with $QLoRA$/$LoRA$ adapters, and a Retrieval-Augmented Generation (RAG) module enhanced with demonstrations to ground analyses in historical telecom data. Empirical results on a large real-world telecom dataset show TeleDoCTR outperforms baselines in ticket routing accuracy, retrieval recall, and fault-analysis quality (e.g., ROUGE/BERTScore), demonstrating improved efficiency and reliability in telecom operations. The approach emphasizes deployment-ready design with multi-task adapters, RL-guided alignment, and demonstration-based RAG to support iterative, context-rich troubleshooting across complex telecom infrastructures.

Abstract

Ticket troubleshooting refers to the process of analyzing and resolving problems that are reported through a ticketing system. In large organizations offering a wide range of services, this task is highly complex due to the diversity of submitted tickets and the need for specialized domain knowledge. In particular, troubleshooting in telecommunications (telecom) is a very time-consuming task as it requires experts to interpret ticket content, consult documentation, and search historical records to identify appropriate resolutions. This human-intensive approach not only delays issue resolution but also hinders overall operational efficiency. To enhance the effectiveness and efficiency of ticket troubleshooting in telecom, we propose TeleDoCTR, a novel telecom-related, domain-specific, and contextual troubleshooting system tailored for end-to-end ticket resolution in telecom. TeleDoCTR integrates both domain-specific ranking and generative models to automate key steps of the troubleshooting workflow which are: routing tickets to the appropriate expert team responsible for resolving the ticket (classification task), retrieving contextually and semantically similar historical tickets (retrieval task), and generating a detailed fault analysis report outlining the issue, root cause, and potential solutions (generation task). We evaluate TeleDoCTR on a real-world dataset from a telecom infrastructure and demonstrate that it achieves superior performance over existing state-of-the-art methods, significantly enhancing the accuracy and efficiency of the troubleshooting process.

Figures (6)

• Figure 1: The overview of TeleDoCTR, which has four key components: (1) domain-specific rankers (ranking); (2) a finetuned generative model for ticket routing (classification); (3) a finetuned generative model with multiple fault analysis generation and ranking (generation); and (4) an enhanced RAG-based fault analysis generation with demonstrations selection (generation).
• Figure 2: Instruction-tuning template for finetuning the team label generation model.
• Figure 3: Instruction-tuning template for finetuning the fault analysis generation model.
• Figure 4: Fault analysis generation prompt for the first round of the RAG-based module.
• Figure 5: Team labels distribution.