https://github.com/meghaniankov/news-toxicity-analyzer

📰 Analyze news sites for text toxicity using TensorFlowJS
https://github.com/meghaniankov/news-toxicity-analyzer

javascript personal-project tensorflow

Last synced: 21 days ago
JSON representation

📰 Analyze news sites for text toxicity using TensorFlowJS

• Host: GitHub
• URL: https://github.com/meghaniankov/news-toxicity-analyzer
• Owner: meghaniankov
• Created: 2020-02-22T15:53:47.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2022-12-10T19:09:03.000Z (about 2 years ago)
• Last Synced: 2024-11-03T16:16:12.466Z (2 months ago)
• Topics: javascript, personal-project, tensorflow
• Language: JavaScript
• Homepage:
• Size: 5.7 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 7
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# News Toxicity Analyzer

A tool that analyzes text toxicity (using Tensorflow's toxicy model) of news articles.

## Built With

* Javascript

* Tensorflow.js (Toxicty Model)

* Firebase (databse)

## Running The Program

Install dependencies and prepare the build directory:

```$ yarn```

Run the server:

```$ yarn watch```

View in browser:

Visit ```http://localhost:1234/```

## How It Works

1. User inputs text and clicks 'classify'

2. ```predict()``` is run

    1. Loads TF toxicity model

    2. Saves user input to ```var text```

    3. Runs ```classify()``` on ```text```

    4. Runs ```sendData()```

    5. Saves prediction to ```var allPredictions```

3. ```classify()```

    1. Runs ```model.classify()``` (model that was loaded in earlier step, as TF classify function) on user input

    2. Sets the output of that function to ```var results```

       - output is an array of objects, one object for each toxicty type/label & it's probability 

    3. Creates ```const obj``` that has a property of ```text``` set equal to the user input

    4. Iterates over ```results``` and adds each toxicty type/label & it's probability to ```obj```

    5. Returns ```obj```

4. ```sendData()```

    1. Saves ```obj``` data into database

    2. Returns error if it did not save correctly.

    3. Returns Firebase key if data successfully saved.

5. User clicks 'Display Predictions'

6. ```displayPredictions()``` is called

    1. Iterates over each prediction in ```allPredictions```

    2. console.logs key/value pair of each prediction (toxicity type/lable => true/false)