The problem M2CRM solves

Our application that Converts a simple smartphone device into a Smart CRM system and helps in lead generation from Calls and WhatsApp messages, through storing conversation in textual format, it aims at helping new struggling entrepreneurs those are not very prosperous financially in the initial stages of their start-ups to afford heavy subscriptions and compliances based products present out there in the market.

The application solves their various problems like:
In the initial days, for Marketing, they need Mobile Numbers and Emails of the customers who have shown interest in their product or services(Manual Lead generation).And that can be easily stored as we provide the data in the form of spreadsheet.
Also, they need daily analytics and performance to examine outcomes of the marketing campaigns and experiments they are trying, to keep their schemes updated and effective. Also how a particular employee is performing.
All verbal communications(Calls) are lengthy to analyse, and consumes much time. So, the application gives the sentiments tracked properly, and also stores the call data for further evidence if any issue or misunderstanding arises.
It's difficult for them to separate all Verbal conversations manually into different categories like query or complaint or feedback. But using the application, gives them textual format that is easier to analyse.
As customer experience is the most important so, They can't track customer sentiments, like how they are changing while being on call when new things are being introduced to them by Start-up Executive(all the changing sentiments as the call goes on).The application helps them to track the changing expressions of the customer as the call continues, and it also helps them to know whether the query brought at the start was resolved by the end or not, or whether the customer was satisfied with the solution for the complaint filed etc

Challenges we ran into

Difficulties

1. Changes were not being reflected in Excel Spreadsheet directly so, we had to find another way for updating it apart from google cloud API so, we used appscript as a backend for the spreadsheet
2. Converting of speech to text was bit easier using Google Cloud Speech API but we had to convert call recordings after call so, we had to implement special different model for this
3. Classification into categories was bit easier but for Sentiment’s analysis we had to build special model and in that, Passing the same audio input into both was a challenging task
4. Wav2TXT2 was creating huge file which was not able to host it on Heroku so, we find 2 alternatives, one is to decrease some dependencies and other by using cloud
5. As size of model is huge so, API delays was one concern but we tried to minimize the time