⏰ WIP

Operational - added ML-worker and keepalive functions (needed for MV3 on Gecko)

Requires new data for tweaking and

Refactoring needed

• moving connection handling into a modular template.
• LIFOQueue to operate like a set, where most recent updates are kept.

---

💡 Purpose

Coming across posts that feature rhetoric on BBC's HYS section, it should be easy to tag these... right?

The comments show poor spelling, weird capitalisation and worst of all, rhetoric (hence the repo name, rhetoric metric).

Border colours fingerprint to keywords. If grayscale, it's only strongly worded.

🗺️ Layout

Runs in Gecko based-browsers. Requires npm since you're building from source.

Analyse each comment:

1. Check for keyword spamming within every sucessive clause. If subclause embedding distance is high, that's spamming.

2. Check for ranting. If clause-to-clause sentiment confidence is high, there's only one meaning from this text.

Future

• See if LLM's keyword-in-context is placed in same locations between different authors/possible sockpuppets, so 'detect LLM accent'

• See if same author/account has unsually wide embedding space layout, would indicate highly knowledgable user using keywords from all topics or possible account sharing.

---

Installation

1. git clone https://github.com/198thread/ricMetric

2. Manually download the models and config files from:

   https://huggingface.co/Xenova/bge-base-en-v1.5

   https://huggingface.co/Xenova/distilbert-base-uncased-finetuned-sst-2-english

   Using git won't work.

It should be in this structure within the clone:

- README.md
- package.json
- package-lock.json
- src
- models
    - Xenova
        - bge-base-en-v1.5
            - onnx
        - distilbert-base-uncased-finetuned-sst-2-english
            - onnx

3. npm install

4. npx webpack

5. Navigate to about:debugging#/runtime/this-firefox ... to Load temporary Add-on ....and add anything in the dist subfolder.

6. Navigate to any bbc.com news article where there are comments.