frederickroman / readgauge

Reading ease scoring web app. It calculates the US school reading level of a text client-side using neural networks.

Readguage

AI-powered reading scoring app.

It calculates the US school reading level of a text.

It uses Neuro-Flesch which is a neural version of the flesch scoring system created for this app.

It's private because all calculations get done in-brower and nothing gets ever sent to a server.

Live website

See Readgauge website.

Main libraries used in this project

• React.js (TS)
• Tensorflow
• • TS - to run the neural network model client-side
• • Node - for model design (research models in python in other repo; rewritten chosen model in node in this repo under /ml)
• MUI
• Chart.js
• moo.js - to tokenize the input text client-side.

Project setup

yarn install

Compiles and hot-reloads for client development

yarn start

Compiles and minifies for production

yarn build

Launches the test runner in the interactive watch mode.

It runs unit tests for the main elements, and integration tests for scoring test cases.

yarn test

Retrains the neural network

It reruns the entire process from data cleaning trough training to evaluation of the model.

node /ml/train