This project was made at Pixels Camp in under 48 hours, we took on a challenge proposed by the Gulbenkian foundation, Hack For Good.
A growing body of evidence indicates that children are accessing the internet at increasingly younger ages. Smartphones are fuelling a ‘bedroom culture’, with online access for many children becoming more personal, more private and less supervised. At the same time, digital technologies are bringing opportunities for learning and education to children, especially in remote regions and during humanitarian crises.
Technological solutions to prevent and intervene in cyberbullying incidents.
Which turned into a chat bot, that should be seen as a kind buddy to prevent cyberbullying and social harassment in network acting in between the bully and his victim.
We designed a custom Machine Learning webservice, using Azure Cognitive Services, processing the text messages as n-grams and trained a model for the probability of being a message with a "bully" feeling to it. We published this as a web service for consumption from our backend.
This custom service took an input from the Microsoft Text Analytics API aswell to optimize the output.
This was intendend to be used for chat rooms where the main audience are children, that's how the product was designed.
Kind
Read all about our product here.
Given the 48 hour time constraint we tried to use the technologies we knew better as a team and that provided out of the box solutions.
- Real time chat - Graphcool
- Client - React
- Kind API - NodeJS with Express
- Sentimental Analysis API - Microsoft Azure Cognitive Services
- Insult Analysis API - Microsoft Machine Learning Studio
We only have one endpoint and that's it.
Returns an analysis of the message, the sentiment, kindness and determines if its kind or not, when not kind we suppose it as an insult.
-
URL
/is
-
Method:
POST -
Data Params
Required:
message=[string] -
Success Response:
-
Code: 200
Content:Message:
Hey{ "success": "true", "data": { "isKind": true, "kindness": 1, "sentiment": 0.9869865775108337, "insult": "0", "message": "kind" } }Message:
You are dumb{ "success": "true", "data": { "isKind": false, "kindness": 0, "sentiment": 0.011714428663253784, "insult": "1", "message": "insult" } }
-
Run in the development environment
npm run dev
Clean up the code
npm run eslint
This project was setup to deploy with the Global Serverless Deployments of Now.
Make sure your src/config.js file is setup with your API keys, find an example at src/config.example.js
Build the project with
npm run build
Start the production server
npm run start
Deploy, install Now CLI
now