The application we am making a dating app for my final project where AI will give people matches. The matches will be based on the users interests. When the user signs up, they will add up to 10 interests in their profile. These interests of each user are saved in the database as Tags. eg: Football, cricket, Manchester United etc.. All the messages the user sends will be recorded and exported from the database.
We need you to create the following models:
Model 1) TF-IDF + SpaCy Take input of all the messages sent by each user and create new tags from that. we will add these tags to the user's profile in database. when the tag is created by AI it will be called hidden tag. the user can't see or edit this tag.
The Ai Tags will be 3 categories:
1) Likes
2) Dislikes
3) Neutral
So for example maybe the user sends message to another user saying 'I like fishing' the AI will analyse this text and put a new tag in the users profile, under likes: Fishing.
Model 2) TF-IF + Word2vec/Doc2vec to find answers to questions that we don't directly ask the user but they might share in their messages with other users. The questions are pre-set questions. eg: where were you born or what school did you go to. more examples of the questions are in the excel file we sent you.
After running these two algorithms, we will write the data back in the database using a rest API.
Model 3) Doc2Vec + K-means/KNN Rest API will deliver input data from my database. We will look at the total number of tags (hidden: likes, dislikes, neutral and Visible: always likes) and run the next model. It is based on K-means. We get the full list of updated tags from the users profile in the database and run k-means to separate the all the users into into clusters. Finally, we write the users cluster number back in the database using rest API.
Then in the app we will write a function that will make matches based on the K value of each user.
You need to create the models so that they take input data in CSV format and export data in CSV format
You need to install the model on my PC via remote or Zoom etc so that we can import CSV on it and run the models with GPU. The output should be in CSV too.
We don't have any data so you will have to gather and label the data from internet too.
About the recuiterMember since May 20, 2018 Jonathan Gillha
from Echternach, Luxembourg