Code for Short Text Topic Modeling with Topic Distribution Quantization and Negative Sampling Decoder

PDF

Usage

0. Prepare environment

For exact package version, please refer to the environment.yml file

requirements:

python==3.6
tensorflow-gpu==1.13.1
scipy==1.5.2
scikit-learn==0.23.2 

1. Prepare data

Note: the data should be preprocessed with tokenization, filtering non-Latin characters, etc before.

   function is in utils/input_preprocessing.py

python utils/preprocess.py --path data/stackoverflow --output_dir input/stackoverflow

2. Run the modeli

In the main file run_NQTM, there are a set of parameters can be tuned, like number of topics, model hidden layer dimension. 
Two main function to be called:
1) print_tpic_words: this function will select topic and associated words for that topic for each of the document. two parameter:
	A) n_topic_selection=2  number of topic selected 
	B) n_top_words=2        in each of the topic how many words to be kept
2) print_top_words, print all of the selected topic words in the model 


two data sets can be used, one is stackoverflow, the other is Calculus 


python run_NQTM.py --data_dir input/stackoverflow --output_dir output

3. Evaluation

topic coherence: coherence score.

topic diversity:

python utils/TU.py --data_path output/top_words_T15_K50_1th