/LLM_Social_Media

LLM-based analysis of social media comments

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

LLM-based qualitative and quantitative analysis of social media comments

Overview

A pipeline that retrieves comments of videos to a specified hashtag, translates them into English, and analyses them for topics, needs, and experiences using BERTopic and Llama.

Prompts used with the Llama Model

Prompt Purpose
The main topic is {hashtag}. Please give a not too general but conscience subtheme that unifies these words: {topic_words}. Only state the topic. to give a unifying theme for topic clusters identified by BERTopic
Please identify some needs of people interacting with {hashtag} videos in the following list of comments: {list_of_comments}. to identify needs expressed in comments
Please summarise the summaries in the following text and identify 5 needs: {list_of_needs_per_video}. to summarise the needs across all videos
Please identify some experiences of people interacting with {hashtag} videos in the following list of comments: {list_of_comments}. to identify experiences expressed in comments
Please summarise the summaries in the following text and identify 5 experiences: {list_of_experiences_per_video}. to summarise the experiences across all videos

Workflow

Overview of the Workflow

How to cite

tbc