Research on methods of anomaly detection for chatbots which was presented in a poster format on Russian Summer School of Information Retrieval 2017 (RuSSIR'17). The .pdf version of the poster could be found here: https://yadi.sk/i/UEi2jwU43MFWVi
Dependencies:
- numpy
- pandas
- gensim
- scikit_learn
- tensorflow
- keras
- pymorphy2
- matplotlib
- seaborn
Code of the experiments is proposed at scikit_learn_anomaly.ipynb
You should create a directory models and put there Google News model (rename it to google_news.bin) and Dvach model. Then download datasets 2ch-topics and reddit-topics and put them in the root of the project.
Then you will be able to reproduce the experiments.
If you have any questions or comments on the paper please feel free to contact me:
amir{my username on GitHub}@gmail.com