Large Language Models for Telecom NLP: A Task-Oriented Survey

Large language models (LLMs) have become an important tool for the automation of text-based tasks in the telecommunications domain due to their high success in natural language processing (NLP) in recent years. Tasks such as classification, summarization, question-answering, named entity recognition and sentiment analysis conducted on data sources such as technical documents, user complaints, call center records and standard documents specific to the telecom domain contribute to the improvement of service quality and operational efficiency. However, factors such as domain-specific data scarcity, linguistic diversity and contextual complexity limit the general-purpose use of models and increase the need for domain-adaptive solutions. This study comprehensively covers task-oriented NLP applications in the telecommunications domain and fine-tuned LLM-based methods with open-source datasets used in these applications. In addition, the role of LLMs in improving the quality of interaction is evaluated through applications such as sentiment analysis-based user satisfaction prediction and complaint tendency classification. The review also highlights key limitations and proposes future research directions for the advancement of LLM applications in telecom-specific NLP tasks.