Running the code

Trsl

  $ git clone https://github.com/iisc-sa-open/trsl
  $ cd trsl

Dependencies for Trsl

• Python 2.7
• Pip
• Other packages required by trsl are given in requirements.txt

  $ sudo pip install -r requirements.txt

Generating the Tree model

Generates a decision tree model, using one of the following options

• A text Corpus and a config file. The contents of the config file are as follows:

     [Trsl]
     ngram_window_size   =  < number of predictor variables >
     samples             =  < number of samples at each leaf node >
     reduction_threshold =  < reduction threshold at which tree growth should be stopped >
     set_filename        =    None
  
     [Set]
     no_of_clusters      =  < number of clusters to be formed from the word vectors >
     no_of_words         =  < number of words to be utilized for word vectorization >
     word2vec_model_path =  < word2vec google binary news vector>
  

  Example

     $ cd ..                 # change directory outside the parent directory of trsl
     $ python -m trsl.main --config ./trsl/examples/example.config --text ./trsl/data/corpus/last_question
  

• A Text Corpus and pre-computed sets

    $ main.py [-h] [-v] [-s] [-m MODEL] [-c CONFIG] [-t CORPUS] [-g GROUP]
  
    Script used to generate models
  
    optional arguments:
    -h,        --help           show this help message and exit
    -v,        --verbose        increase output verbosity
    -s,        --silent         silence all logging
    -m MODEL,  --model MODEL    pre-computed model file path
    -c CONFIG, --config CONFIG  config file for the model generation
    -t CORPUS, --text CORPUS    text corpus for model generation
    -g GROUP,  --group GROUP    groups of words, pre-clustered words based on vectors [ sets ]
  

  Example

     $ cd ..                 # change directory outside the parent directory of trsl
     $ python -m trsl.main --group ./trsl/data/word-sets/Inaugural-speeches/Inaugural-speeches-Kmeans-8851words-50clusters.json --text ./trsl/data/corpus/last_question
  

Predict the Next word

• The user provides a sequence of words (number of words being that of the ngram window size), and
• the program predicts the next probable word (a.k.a. target-word)

$ predict.py [-h] [-v] [-s] -m MODEL

Example script for target word prediction based on the input predictor variables of ngram window size

optional arguments:
 -h, --help               show this help message and exit
 -v, --verbose            increase output verbosity
 -s, --silent             silence all logging
 -m MODEL, --model MODEL  pre-computed model file path

Example

  $ cd ..                 # change directory outside the parent directory of trsl
  $ python -m trsl.examples.predict --model ./trsl/data/model/last_question/model

Random Walk

• The user provides a sequence of words (number of words being that of the ngram window size) and also the amount of text to be generated
• The program then generates the required amount of text, using the initial sequence of words as a seed.

$ tree_walk.py [-h] [-v] [-s] -m MODEL

Example script for random tree walk based on the input predictor variables 
of ngram window size and the number of words to be generated

optional arguments:
 -h, --help               show this help message and exit
 -v, --verbose            increase output verbosity
 -s, --silent             silence all logging
 -m MODEL, --model MODEL  pre-computed model file path

Example

  $ cd ..                 # change directory outside the parent directory of trsl
  $ python -m trsl.examples.predict --model ./trsl/data/model/last_question/model