The idea is to have an NLU (Natural Language Understanding) of Conversational AI that is fast to train, fast to run, but with a good accuracy.
To do the comparision we use the english and spanish corpus from Amazon Massive dataset. https://www.amazon.science/blog/amazon-releases-51-language-dataset-for-language-understanding
We will compare the speed and accuracy with RASA.
Download this repository.
No need of npm install
as there are no dependencies.
npm start
For the English corpus, these are the results:
- Time for training: 3s 213.6769ms
- Accuracy: 83.66%
- Transactions per second: 181836.99495205944
RASA Accuracy is 81.4%, time to train in RASA is 4517 seconds, Transactions per second in RASA are 84
For the Spanish corpus, these are the results:
- Time for training: 2s 488.0854ms
- Accuracy: 81.91%
- Transactions per second: 141800.23397571384
RASA Accuracy is 80.4%, time to train in RASA is 4712 seconds, Transactions per second in RASA are 82