firsttosay's Introduction

First to Say

                                                            wmjg

A webapp playing with document similarity

Say something original or be shamed as a copycat

A docker-compose is included to allow you to deploy it all the same way I do.

If you want to start fresh, you'll need to download some data from kaggle (sources in helper_funcs.py)

Then, in command line run the command:

    | docker-compose up --build

Built Utilizing:

flask and sqlalchemy
sqlite
elasticsearch
spacy language detection
nginx and gunicorn
docker

Example of how it works:

            1. you register as a User so you can be the author
            2. input a sentence you think is original
            3. return your Quote and the most similar quotes in our db
            4. You either are original -> now your quote is part of the db
                * Or you're not, and we show you who got there first

Under the hood:

            1. registered as an author in a sqlitedb
            2. normalize your input then spacy checks for grammaticality
                * Using language detection, treshhold of recognition
            3. query elasticsearch for MOST similar quotes
                * top elasticsearch results get cosin similarity scored
            4. below a certain similarity threshhold
                * your Quote is generated as legitimate, entered into db
                * else, show you who got there first

Every User (authors included) has an accessible profile page of all their quotes

And all Quote can be liked by a User

Future implementations:

* Gamification: 3 strikes and you're out, track longest streak
* "Followers": track who you're similar to, display in profile
* Topic Model: display in profile your most common topics
* cookies and tracking: can't see your entry result more than once; refresh or back loses that similarity results page

Recommend Projects