Filtering insights from multi-channel feedback

The problem Voicesift solves

Our project aims to harness the power of AI and data analytics to extract actionable insights from customer reviews across various channels . By utilizing advanced natural language processing (NLP) models and sophisticated algorithms, we strive to provide businesses with valuable insights to improve their products, services, and customer satisfaction.


Multi-Channel Data Ingestion: Ingests reviews from the App Store, Play Store, and Twitter using DevRev APIs in real-time, ensuring a comprehensive view of customer feedback.

Four level noise filtration: Initially we identify and eliminate duplicate reviews. Then distinguish between AI-generated or human-generated reviews using GPT Zero. Next we determine spam and NSFW content using Llama Guard 7B.

Review Segmentation: Segment reviews into feedback, feature requests, bugs, or questions using Llama Mixtral 8X 7B.

Sentiment Analysis: Perform sentiment analysis on feedback to categorize them into positive, neutral, and negative classes using Rapid API.

Feature Importance Analysis: Analyze feature requests to determine the importance of each feature and assess its business impact.

Bug Severity Analysis: Identify the severity of bugs, their potential benefits if resolved, and provide suggestions for fixing them.

Customer Knowledge Gap Analysis: Analyze questions to identify areas where there is a customer knowledge gap.

User Engagement Metric: Calculate metrics like reach based on public metrics such as retweet count, reply count, like count, etc., specifically for the Twitter channel.

Customization and Scalability: While we primarily utilize Llama Mixtral 8X 7B for most tasks, our platform is flexible, allowing businesses to tailor AI models to their specific needs and requirements.

Actionable Insights: We use simple logics and counters across the pipeline to generate insights like average sentiment scores, importance and severity scores etc.

Challenges we ran into

Challenges we faced are:

  1. We did not have any previous experience with typescript or any javascript based languages, But it was very fun learning along the way.
  2. It's the first time using Devrev platform. Same for a lot of us :)
  3. Some of the API we had to get by ourselves, which was a hassle but we got it done.