Customer Support Chatbot Training
I led a project to annotate over 15,000 customer support transcripts for a travel company. Tasks included labeling user intents (e.g., booking inquiries, cancellations, payment issues), tagging named entities (dates, locations, amounts), and marking sentiment polarity. I implemented quality assurance measures such as double‑review cycles, inter‑annotator agreement checks, and automated validation scripts. The dataset was later used to train and fine‑tune a multilingual chatbot that improved response accuracy by 30%.