Oswald provides you with all the tools you need to train your chatbots NLP. In the background, it uses Oswald NLP, Watson Assistant or Microsoft LUIS. It's up to you!
You can train your chatbot to understand certain intents in the input of the user. To train the NLP with intents, just add some example sentences. The chatbot will learn how to interpret new sentences by recognising the right intent with a certain confidence score. You can even export and import example sentences for easy translation.
Entities are recognised, extracted and named from the user input. You can add them by defining an entity type, some entity values and even synonyms. Entities are extracted even if some minor typoes are made by the user. Furthermore, you can upload entities from a file. Finally, the chatbot can detect multiple entities in the user input.
Use the Oswald UI to test different input strings. See how the chatbot NLP reacts. Which intent is detected? Is the confidence high enough? Which entities are extracted? Can you add some typoes? Based on the result, you can improve the chatbot. If the wrong intent is recognised, select the right one and validate. The chatbot will learn and relearn.
The default NLP service in Oswald is Oswald NLP. However, you can swap the NLP to another service, like IBM Watson Assistant or Microsoft LUIS. That way, you can combine all Oswald features with the power of external NLP services. Everything else in the platform stays the same. Additional fees for IBM of Microsoft may apply.
The Oswald confusion matrix is a technical feature to take your chatbot's NLP to the next level! The idea is quite simple. If the squares on the diagonal of the matrix have high scores and the other squares have low scores, your chatbot is trained very well. If not, you should consider retraining your chatbot.