Blabber
Your AI-powered meeting companion that captures, tracks, and summarizes discussions, generating insightful reports for seamless collaboration.
Created on 30th September 2024
•
Blabber
Your AI-powered meeting companion that captures, tracks, and summarizes discussions, generating insightful reports for seamless collaboration.
The problem Blabber solves
Inaccurate Record-Keeping: Traditional methods of taking notes during meetings can lead to missed information or inaccuracies, causing miscommunication and confusion. Blabber solves this by providing real-time transcription capabilities, ensuring that all discussions are accurately captured.
Inefficient Report Generation: Manually compiling meeting notes into comprehensive reports is time-consuming and prone to errors. Blabber streamlines this process by allowing users to generate customizable reports in multiple formats (PDF and DOCX) with just a few clicks, saving time and effort.
Lack of Contextual Information: Often, reports lack the necessary context to understand key points. By enabling users to capture screenshots during meetings, Blabber enhances reports with visual elements that provide clarity and context to discussions.
Difficulty Analyzing Discussions: Understanding the dynamics of conversations can be challenging without proper analysis tools. Blabber offers various reporting customizations, including speaker-based, interval-based, and sentiment-based reports. This allows users to gain insights into individual contributions, time management, and emotional tones, facilitating better decision-making and conflict resolution.
Communication Gaps: Sharing meeting outcomes with team members or stakeholders can be cumbersome. Blabber simplifies this by allowing users to email reports directly to relevant parties, ensuring everyone is on the same page and promoting transparency.
Challenges I ran into
Obtaining Transcripts
We integrated high-quality speech-to-text APIs for accurate real-time meeting transcripts. To improve accuracy, we applied audio pre-processing to reduce background noise and allowed manual corrections, enhancing the overall quality and providing user feedback.
Capturing Screenshots
We developed a lightweight screenshot tool within the Chrome extension, enabling users to capture images unobtrusively using a keyboard shortcut. Screenshots are automatically linked to transcripts and can be annotated for added context.
Sentiment Analysis
Utilizing NLP techniques and diverse sentiment analysis libraries, we assessed emotional content in discussions. A scoring system was created to help users quickly identify prevailing sentiments and address team dynamics effectively.
Machine Learning Integration
We established a modular machine learning architecture for transcription, sentiment analysis, and report generation. Focused on data quality and continuous performance monitoring, we employed cloud solutions for scalable model training and improvements based on user feedback.
Effective Summarization of Transcripts
We implemented a multi-step process for summarizing lengthy transcripts by breaking them into segments based on speaker turns. Extractive summarization techniques highlight key sentences, allowing users to choose between brief summaries and detailed reports for efficient information digestion.
Tracks Applied (2)