About Eye-Mitra and B-able

B-able

Basix Academy for Building Lifelong Employbilities, B-ABLE a company focused on private and non-private Projects and help rural , urban and semi-urban people to take part in the projects by becoming an entrepreneur along with training .

Eyemitra

Esselore is world's top 3 spectacles organisation and started project Eyemitra in India with the vision of :
1) Providing opportunity to rural , semi-urban and urban people to start their business shops using eye-mitra.
2) Helping rural , semi-urban and urban people to have right spectacles with them .

Problem team was solving

There are thousands of eyemitra shops in UP, Rajasthan , MP and growing more.
And eyery state has one eyemitra office consisting of few business developers to gather daily information of eyemitra shops . Now since there are tons of shops and therefore they use to travel hundreds of kilometers on bike Therefore it took about a month to cover 200 shops of one region and thereofore the Data Analytics team analyse the deep insights after a month with half data only making delay in taking business decisions.

Solution Planned

Now this is where technology enters and me with my team come up with an solution. We have 2 solutions in our mind.

a) Prepare a simple google form and circulate it everywhere. This consist of no hard work but the major problem is as a user why will I fill a form specially when I am unable to understand english , also taking a lot of time and as an entrepreneur a user don't have so much of time.

b) This is the loveliest part for me where we innovate this problem with the help of Multi Lingual Voice Assistant which will just ask 4 questions in HINDI from user and the user will reply in Hindi , so like this we will have data of all eyemitras in just a minute at the backend making the complete process much easier by consuming a huge amount of time.

c) This was the extended version of Multi Lingual Voice Assistant. Since we are able to collect all the data at backend so team planned to prepare a pipeline of not just getting data but also analysing it along with a prediction model of NVG sales going to happen in future. This will automate the whole process .

How solution is helping and overcoming the problem

I planeed to use Dialoglow , since it the easiest platform of building customised multi-lingual voice assistant and along with that planned to use python flask at the backend for collecting the user data, saving it in database and use for analysis . The data we were getting is obviously in Hindi , therefore used Google apis to convert the data in enlish from Hindi making the process of analysing it easy. I along with my team designed the questions to be asked from Entrepreneurs which were related to:
1) Their name.
2) Father's occupation.
3) Region where the live.
4) Their sales for today of eyemitra as well as non-eyemitra spectacles.
The above questions are in Hindi, also shown in above video whose replies are first getting stored in python and from python to sql. Then this data is use for making prediction model and analysing sales for each day.

What is the functional flow of the voice assistant

The voice assistant is embedded in google so the Entrepreneurs on field start the conversation for giving their data on EOD and their data is recieved at the backend and from backend it gets saved in sql database and then used for analysis and prediction modelling.

Problems Faced

A chatbot sprecially when it is a Voice assistant is just like a small baby which needs to learn a lot and therefore just after the completion of the voice assistant me and team took it in field where it get failed in first test. So, we first trained it for next whole month and then again took it to field and found that bot is now capable of undestanding every assent of Hindi. To understand this concept much better I am attaching a video by Google on "Age of AI".

---

Age of AI (Explaining how a bot learns)

Results

Bot is now well enough trained to understand the words of Hindi in almost every assent.