CLiFF (Clustering & Language model integration for FAQ Formation)
Published:
Developed CLiFF, an automated natural language processing pipeline that generates frequently asked questions from technical support tickets. Using a fine-tuned large language model, sentence embeddings, and semantic clustering, the system identifies recurring support issues and transforms groups of related tickets into reusable question-and-answer content.
The pipeline includes ticket preprocessing, issue-and-resolution summarization, semantic subclustering, cluster-quality ranking, and FAQ generation. The system was developed for technical support data in high-performance computing environments and supported a peer-reviewed publication.