What you'll learn
- Understand the concepts of NLP, NLU, NLG, NER and how they are applied
- Write your own tokenizer and use existing tokenizers from Natural and NLPjs
- Measure the precision of the different NLPs using a custom corpus and SIGDIAL22
- Understand what is stemming, apply stemmers to the classifiers that you built (Brain, Tensorflow, NLPjs neural, Microsoft LUIS and your own classifier), measure the precision and compare with non-stemmed versions
- Understand Measures in Machine Learning: Confusion Matrix, Accuracy, Precision, Recall, F1-Score
- Different KPIs for NLU: Machine Learning Measures, Confidence, Clarity, TPS, Language Support
- Named Entity Recognition: enum entities, regular expressions, builtins using Microsoft Recognizers and Duckling (extract emails, URLs, dates, numbers...)
- Useful NLP Algorithms: utterance generation from patterns, N-Grams, Language Prediction, Predictive Text, Levenshtein distance, get best similar substring, emoji replacement
- How to build your own chatbot, benchmark it, do it multi-language, publish into console, add a web for talking with the chatbot
You'll learn what is Natural Language Understanding, how to build classifiers using different technologies and how to measure them.
Also Named Entity Recognition, to extract entities from the sentences of the user to complement this NLU.
Natural Language Generation, to generate correctly the answers of the Conversational AI, and how to have answers based on context variables to give a contextualized experience to the users.
Also some util algorithms, like N-grams, and how to use them for real problems like guessing the language of a sentence or predicting the next word of a sentence.
Finally we will build a chatbot, multi-language in english and spanish, able to guess the language from the sentence, and using this Conversational AI techniques that we have learn during the course, and finally we will connect this chatbot to a web exposing an API.
Who this course is for: