A platform for connecting social media data to domain-specific topics using large language models: an application to student mental health.

ABSTRACT:

Objectives

To design a novel artificial intelligence-based software platform that allows users to analyze text data by identifying various coherent topics and parts of the data related to a specific research theme-of-interest (TOI).

Materials and methods

Our platform uses state-of-the-art unsupervised natural language processing methods, building on top of a large language model, to analyze social media text data. At the center of the platform's functionality is BERTopic, which clusters social media posts, forming collections of words representing distinct topics. A key feature of our platform is its ability to identify whole sentences corresponding to topic words, vastly improving the platform's ability to perform downstream similarity operations with respect to a user-defi

SUBMITTER: Ruocco L 

PROVIDER: S-EPMC10799551 | biostudies-literature | 2024 Apr

REPOSITORIES: biostudies-literature