The ongoing pandemic has forced us to adapt to online mode of work/education. Most the universities and schools have opted for online classes. In these online classes, there are a lot of problems which are faced by the teachers as well as students. The main aim of our project is to generate a report of the meetings which could pave the way to better meetings!
We start off by collecting data i.e. the class recordings. The next step involves conversion of the recording to text format and retrieving the time stamps of the words using the Google Speech Recognition API. From the word timestamps we calculate the average response time and duration of silience in the meeting. We classify each sentence into question and non-question using SVM (Support Vector Machine). We further check the interest level and difficulty of a question using cosine similarity and KNN. The results of all these modules are combined and sent to the frontend using Anvil.
- Speech to text
- Question detection
- Number of active attendees
- Average response time
- Number of interesting questions
- Number of difficult questions
