Active Semi-supervised Variational Autoencoders under Representativeness Constraints

This project applies semi-supervised VAEs on pool-based uncertainty active learning. To respect the underlying class distribution different representativeness strategies can be added. To find the original published code of the thesis: git checkout thesis_release

Installatiion

active learning code requieres python 3.6+ and tensorflow-1.12+

Project Structure

|
|--data: already preprocessed data of the ADR-Dataset (http://diego.asu.edu/downloads/twitter_annotated_corpus/),
|         US Airline (https://www.kaggle.com/crowdflower/twitter-airline-sentiment) Sentiment Dataset, 
|         Large Movie Review dataset (http://ai.stanford.edu/~amaas/data/sentiment/) and
|         side effect synonyms    
|--src
    |--java: contains tweet preprocessing (tweet tagger http://www.cs.cmu.edu/~ark/TweetNLP/#%23parser_down and 
    |        Lucene are requiered)
    |--python: 
          |--evaluation: contains proposed active learning methods with differnt models + QBC Naive Bayes with EM
          |              to run the proposed method use active_learning.py
          |       |--asiddhant: adapted code of https://github.com/asiddhant/Active-NLP
          |--pre_processing: Further preprocession files