AnantDuhan / Web-Behaviour-Detection

A browser extension that uses sentiment analysis to make your web more positive by reducing negative content visibility.

Web Behaviour Detection:

Use sentiment analysis to reduce negative content visibility to make your web more positive.

What does this do?

A browser extension that finds negative elements on the page and makes them less visible:

• Negativity is calculated using either inin-browserule-based logic or an optional deep learning model running locally in the background (in a docker container).
• The user controls various options (negativity threshold, restyling options) via the options menu:
  • Visibility is reduced by adjusting transparency, or by hiding the element completely.
  • Colours can be added for inspection, or for examining the positivity.
  • Settings can be saved per site or globally.
  • An overall positivity score per page is displayed.

Sentiment "Scoring" options:

1. Use Vader (JS) or AFINN (JS) in the drop-down for browser side, rule-based calculations.

2. For the FlairNLP (Py) option a local python inference server needs to be running: Running in docker (simplest if you have docker)

   Please wait a few seconds after starting to be ready to serve requests.

   Run interactively once (e.g to try it out):

   • Start: docker run -p 8000:8000 -it --rm anantduhan/web-behaviour-detection

   • Stop just Ctrl+C in the terminal where you're running it.

   Run in detached and persistent mode (e.g. for actual usage):

   • Start: docker run -p 8000:8000 -dit --restart unless-stopped --name web-behaviour-detection anantduhan/web-behaviour-detection

   • Stop: docker rm -f web-behaviour-detection.

   Local python (e.g. for development)

   1. Clone repo.
   2. make install to create a local virtual environment and install dependencies in it.
   3. make server to run the server.

Using local extension

Build

Building instructions

• Clone and go to the extension/ folder:
• To install in the local environment: npm install
• Building: npm start for development, npm run build for packaging into a zip file.

Load browser extension into Chrome or Firefox:

Chrome

• Extensions -> Enable "developer mode" -> "Load unpacked extensions" -> Navigate to /extension/dist folder in this project.
• To update (on code changes): go to extension details and press update / reload.
• Docs: Chrome docs

Firefox

• To load for development (will be removed after browser close, but easier to reload on code change):
  • Go to about:debugging -> This Firefox -> "Load Temporary Add-on.." -> Navigate to /extension/dist -> select manifest file.
  • Press "Reload" to update on code changes.
• To load for continuous usage (persistent after closing):
  • Go to about: config and set xpinstall.signatures.required to False to be able to load a local extension.
  • Run npm run-script build to package the extensions into a zip file.
  • Go to about:addons -> "gear" icon -> "Install add-on from file.." -> Navigate to /extension/ folder in this project -> choose web-behaviour-detection-.zip.
  • To update (on code changes): repeat the previous two steps.
• Docs: Firefox docs

Credits and references:

• Python (DL-based) sentiment analysis model and package: flair NLP
• JS sentiment analysis packages:
  • vaderSentiment (called JS-Vader in options menu)
  • AFINNSentiment (called JS-AFINN in options menu)
• Backend API framework: fastapi
• Initial code for browser extension functionality copied from Trump-Filter