I think there are many men who are fidgeting as February 14 approaches. This time, I made a system to diagnose compatibility with the person you care about using LINE's talk history!
Details are introduced in the following video https://youtu.be/LNFFCBIHOXY
Easy to make
Follow the steps below to save the talk history
We will format the data of the talk history in text format Defines three data formats: message, turn, and conversation
The data structure is such that a one-line message is a message, a set of messages is a turn, and a set of turns of yourself and the other party is a conversation.
The LINE compatibility diagnosis system has four evaluation axes
The length of the conversation is calculated for each conversation with a perfect score of 10 days.
The reply interval is calculated so that it will be 0 points after 6 hours or more.
As for the quality of the content, this time I made it easy to score when including question sentences or emoji In the future, I would like to add processing such as adding points when words close to "like" are included in word2vec and doc2vec.
In the evaluation of call time, all calls are averaged with a perfect score of 1 hour or more.
The result is as follows
I made it fairly rigorously, so if you have 60 points, you can do it well, straw
This time, I made a system to quantify the compatibility with the other party with a simple algorithm! I'm worried about the data only, but it seems good to publish it on heroku etc. and have it used.
Recommended Posts