Testing process

Running the tests

A pre-trained model can be utilised for the same or the model can be generated for a corpus using main.py

Usage Details

$ python2 tests/trsl_tests.py --help

usage: trsl_tests.py [-h] -m MODEL

Test script for trsl.py

optional arguments:
  -h, --help                 show this help message and exit
  -m MODEL, --model MODEL    Model file path

Example

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

Tests verified against the given pre-trained model

• There is always reduction in a node from the parent

• Correct probability distribution from the parent node between the children

• Validate len_data_fragment of every node is less than input sample size threshold

• Validate every leaf node contains a word probability The tree is built using main.py from the corpus.

• External files used for generating sets

  • Vocabulary file
  • Vector file

• Future Testing strategies

  • A language modeling challenge set of almost complete sentences from Microsoft