Malaya

1. Sentiment and Toxicity analysis now will use `naive_stemmer` to classify.
2. Toxicity analysis now supported `['bahdanau', 'hierarchical', 'luong', 'fast-text', 'entity-network']`
3. No longer depends on Keras.
4. No longer have any CNN based model due to CuDNN unstable.
5. Added `entity-network` for sentiment and toxicity analysis.
6. Added `bert` for sentiment analysis.
7. Generated readthedocs documentation, https://malaya.readthedocs.io/en/latest/
8. House keeping.

1. Added Deep learning summarization, skip thought vector, simply call by `malaya.summarize_deep_learning`.
2. Added TF-IDF string matching for Topics and Influencers Analysis, simply call `malaya.fast_get_topics`, `malaya.fast_get_influencers`
3. Added Deep learning string matching, skip thought vector, for Topics and Influencers Analysis, simply call `malaya.deep_get_topics`, `malaya.deep_get_influencers`
4. Major housekeeping for `text_functions`.
5. Deep learning Part-of-Speech case sensitive.
6. Retrain malaya word2vec
7. Normalizer now ignores Proper Noun.
8. Spelling correction and Normalizer will ignore location.

1. Added toxicity analysis
2. Added spelling correction
3. Improve normalizer
4. refactor some variable names

1. Added deep learning stemmer
2. Added sastrawi stemmer
3. Improve topic-modelling with stemming
4. Improve summarization with stemming
5. Added more wiki
6. Added more sentiment data
7. Retrained sentiment models
8. Added fast-text model for sentiment analysis

1. Added calculator parser for word2vec
2. Added deep learning models for entities recognition
3. Added deep learning models foe POS recognition
4. Improve Bahasa stopwords
5. Improve summarization
6. Improve topic modeling
7. Improve crawler