Application of Language Models for the Segmentation of Website Visitors

Website visitor segmentation is crucial for effective web presence management and online marketing. We explore methods for grouping website visitors based on user behaviour and incorporating their interests, using the GPT3 language model to analyze the text content of viewed pages content and build user profiles. In our method, the language model GPT3 is used to summarize the content a user visited to build a meaningful user profile, and to answer queries concerning the interests of the users. We then segment the users via text based topic modelling. Our findings indicate that our method of classifying user interests through text analysis and direct language model queries offers high transparency and versatility, surpassing traditional segmentation techniques. Users are characterized in clear profiles, and queries can be tailored to specific interests.