Topic detection model

This component identifies the main (sub-)topics from Twitter posts in Italian about immigration. The core of the component is a BERTopic model fine-tuned on a comprehensive set of tweets representing the information provided by both political entities and news sources on the immigration subject during the 5-year period 2018-2022. The model includes a pointer towards the model to be loaded in with SentenceTransformers, and returns as output the distribution(s) of the input text(s) over the 36 topics identified through the fine-tuning process. We further relied on OpenAI gpt-3.5-turbo to generate topic labels based on their best representing documents.