Simplemeet

We all have seen how remote work has come as a rapid flood of change globally. Adapting to this change, we are all dealing with more online meetings on different platforms. We all may have experienced that we must prepare meeting notes during or after it gets finished. Or we may have to relisten to a long record of meetings to note down some specific information out of it. And for this, we must go back and forth playing the record. So, here is acting as real-time speech translation, SimpleMeet application, that will transcribe the audio meeting into notes, and will also provide summary of the meeting once it gets finished.

 

The document becomes handy to access or share with anyone. Simple-Meet works with any meeting application like Google Meet, Skype, Zoom, etc., running parallel to it. To use this service, we simply need to start recording our meeting with the Simple-meet app, and then Simple-meet will provide us with the live transcription of the meeting with around 90%accuracy. We can pause or stop the recording at any time between meetings. Once the meeting is stopped, we will be provided with the whole transcript of that meeting and its summary, which can be saved or shared via E-mail. Simple meet, a Desktop application, can transcribe many other languages rather than just English.

Key Challenges

This innovative project is exciting to develop. And this is not just a simple page with CRUD operations, instead it deals with the real time language data, hardware configuration and technology all put together. And so we faced different situations like while transcribing data, it was difficult to differentiate audio input (microphone audio) and audio-output (Speaker audio), that is to display the transcript data of You: vs Other: Initially we were facing problem getting small chunk of data of certain interval of time, but we have overcome this with solution.

 

A platform for analytics exists that enables us to keep tabs on user behavior, such as the total number of queries, the most recent searches, the most
searched terms, or the percentage of searches that returned no results. Internally, it generates reports based on the terms searched by the user

Our Solutions

To get fresh chunks of data at every interval of time, we have used timeout with time of 5 seconds while recording the video. By doing this, the recorder will give a small chunk of data every 5 seconds. While developers are still working on differentiating audio-input (microphone audio) and audio output (Speaker audio), to avoid transcribing unwanted data.

Tech Stack

Conclusion

Other Case Study