ConvoKit

Why we included this project

ConvoKit is a Python toolkit for turning raw conversation logs into measurable social signals, aimed at computational social scientists and NLP researchers who would rather not write a new parser for every chat corpus they touch. It wraps the messy work behind a scikit-learn-style interface, so the same calls work whether you are examining a single exchange or a whole conversation thread. The feature set goes well beyond word counts. It can measure linguistic coordination and power dynamics between speakers, flag politeness and impoliteness strategies, and track how a person's voice shifts over time. It can even predict where a conversation is heading as it happens. Several large conversational datasets ship with the package, along with example notebooks that walk through the intended workflow. If you work with Reddit threads, parliamentary records, support tickets, or any dialogue data and want structural and social analysis rather than bag-of-words counts, this is a documented, ready-to-use starting point.