AnalyzeAI
Unveiling Insights, Empowering Decisions: Analyze AI - Your Key to Understanding User Feedback
Created on 23rd February 2024
•
AnalyzeAI
Unveiling Insights, Empowering Decisions: Analyze AI - Your Key to Understanding User Feedback
The problem AnalyzeAI solves
Analyze AI
Analyze AI is a tool designed to analyze and process data gathered from multi-data sources like Twitter, Instagram, and Reddit. It aims to help businesses gain better insights and information from user feedback using a user-friendly interface. This tool classifies user reviews/feedback into various classes highlighting their characteristics. It utilizes a LLM model for data classification into various labels, which are named as tags, aiding in effective categorization, filtering, and analysis.
Three-Fold Chain Structure
Our LLM model uses a three-fold chain structure for labeling:
- Sentiment Analysis: Categorizes tweets into positive, negative, or neutral sentiments, providing insights into the overall mood of the user.
- Category Classification: Assigns tweets into categories such as feature, bug, question, or answer, aiding in understanding the nature of discussions.
- Intent Classification: Helps identify the urgency level of user feedback, distinguishing between medium, urgent, and low priority inquiries or concerns.
Features
Moreover, Analyze AI offers:
- Insights Extraction: Allows users to retrieve meaningful insights from user reviews based on specified tags or categories.
- Graph-Based Representation: Integration of a graph-based representation to enhance visualization and better understandability of the information.
- Summarize Feature: Provides a summarize feature in case the insight generated is too long to read.
- Trend Analysis: Enables businesses to identify emerging topics or discussions regarding their product for informed decision-making and strategic planning.
- Keyword-Based Filtering: Allows targeted searches to gather information about specific topics, such as new features or product launches.
Dashboard
We have also developed a user-friendly dashboard that visualizes various analyses performed on the data, including sentiment distribution, category breakdown, and insight summaries.
Challenges we ran into
Finding the right data to import proved to be a crucial first step in our hackathon experience. The work of classifying the ingested data into multiple unique categories took a lot of time and effort from our team. We worked with the GPT-3.5 API and looked into a variety of open-source models in our search for a solution. Our experimental phase included the idea of creating a Langchain agent, but in the end, we decided to use the Mixtral-7B model to create a simpler chained architecture.
But there were some challenges with this choice. We ran into several issues with the devrevSDK, which forced us to make a calculated decision to use the API instead of the devrevSDK to create tickets. We were able to go beyond the challenges we were having with the SDK thanks to this change in strategy.
Our project's integration of Snapkits was designed to improve user interaction. But this project also brought with it a new set of challenges, mostly because of how detailed the Snapkits documentation is. Despite these difficulties, we were able to successfully incorporate Snapkits into our project thanks to the devrev mentors' tremendous advice and assistance.