System for automatic content moderation.
#####Realized features:
- Researched characteristics of content: likability, inverse relevance, correctness.
- Collected and marked data set of 1250 items
- Trained logistic regression model on Matlab
- Implemented сontinuous filtering service on Scala (Vk API, ElasticSearch)
#####Further work
- Discover new characteristics
- Implement online learning (via chrome app and human moderators)
- Make a site for selling
#####Actuality characteristics:
- Likeability - user sympathy
- Inverse relevance - score from elasticsearch
- Correctness - number of punctuation characters to message length, uppercase chars amount, message size
Albert Bikeev.