/amazon-fake-reviews

Primary LanguageJupyter Notebook

Detection Amazon Fake Reviews with Neural Networks

  • language processing using bert models
  • stitching together reviews in a transformer model to form a context embedding
  • use product information to get a better detection context

Try, if you are better than the AI model:

Working demo hosted at:
http://tadeohepperle.com/amazon-fake-reviews/

The relevant datasets can be downloaded from this Kaggle link.


And the weights for the models can be found in:

Diagram for the Context Embedding model

image

image

Diagram for the 3-BERT model

Miguelmodel drawio2

(This was the final project in a Financial Technology Deep Learning Class at National Taiwan University (NTU), fall semester 2022)

How to run the code

please take a look at the following files, python and node are needed to run the code:

  • scrape.js
  • extract.js
  • preprocessing.ipynb
  • model.ipynb
  • model2.ipynb
  • model2-train.ipynb
  • game (entire folder)
  • final_project_Fintech_3BERT.ipynb
  • context_model_test.ipynb