Skip to content

Evaluate Agent

Ayush Bhardwaj edited this page Jul 9, 2020 · 1 revision

We have customized a set of test data and created an evaluator to test any agent.

If you want to test our existing agent or you’ve developed a new one, just evaluate it on our algorithm to compare it with others.

File: evaluator.py

Evaluation is based on two factors:

  1. Time
  2. Accuracy

To test

cd into atarashi/atarashi/evaluator/ directory

  • DLD agent

    python evaluator.py -a DLD

  • wordFrequencySimilarity agent

    python evaluator.py -a wordFrequencySimilarity

  • tfidf agent

    python evaluator.py -a tfidf

  • tfidf with Cosine similarity

    python evaluator.py -a tfidf -s CosineSim

  • tfidf with Score similarity

    python evaluator.py -a tfidf -s ScoreSim

  • Ngram agent

    python evaluator.py -a Ngram

  • Ngram With Cosine similarity

    python evaluator.py -a Ngram -s CosineSim

  • Ngram With Dice similarity

    python evaluator.py -a Ngram -s DiceSim

  • Ngram With Bigram Cosine similarity

    python evaluator.py -a Ngram -s BigramCosineSim

To check out the latest evaluation report please refer to Atarashi Evaluation Report