So recently I bought this product from amazon which stopped working in 10-12 days of usage. Now the sole reason for buying that particular product out of all similar products offered was that its reviews and ratings looked very positive. It had many positive reviews which were mostly found fake when I tried researching about them deeper. So after that bad purchase me and my team thought of an idea to save users from just manipulative reviews of products since reviews are the only way we can judge the quality of product when buying online.
What it does
So Fake Review Detector is basically a web app which takes an amazon product url as input and tells user if it is recommended for buying or not, the number of fake reviews associated with it and percentage of total fake reviews. If percentage of fake reviews are more than 65%, it will warn user not to purchase that item and user becomes aware before ordering it.
How we built it
It is built using node, express for backend and machine learning model which has CNN, LSTM, Doc2Vec, TFDIF and html, css, bootstrap for frontend .
Challenges we ran into
The major challenge we ran into was to scrap reviews from amazon website using python but since python was taking too much time we switched to node js for scrapping purpose then we had challenge of combining node app with machine learning model fir which we hosted the ml model and used its api in our node application.
Accomplishments that we're proud of
Using ml model inside node application with JS is the major accomplishment we are proud of, also the precision of ml model and hosting of ml model and then whole web application are another things we are proud of
What we learned
We have learned a lot from this project like how to scrape web for information using node application, how to use ML model in node application, how to host ml model online and create its API and how to host web application on heroku.
What's next for Fake review detector
We are looking forward to add more features in this web application, like Price Tracker which will track the price of product for specific interval of time and notify the user as soon price will be dropped so that user can buy the product at discounted prices. We have already implemented this feature in our app but due to time constraint we were unable to solve the errors it was giving hence it still has to be made working.