Introduction to NLP : How to improve accessibility with Machine Learning

WWCode CONNECT REIMAGINE 2021

Transcript

2. Victoria Ubaldo @Vikyale Introduction to NLP : How to improve accessibility with Machine Learning CONNECT REIMAGINE 2021
3. Hello! Computer and System Engineer. Peruvian Data Analyst @Interbank Msc Candidate Computer Science, PUCP. Enjoy mentoring, illustration and dancing!
4. Agenda • Machine Learning and Accessibility • Natural Language Processing • Tensorflow for NLP • Next steps
8. Machine Learning is programming with data
11. Natural Language Process (NLP)
13. Where find NLP ? Chatbots Predictive text (autocomplete) Transcription Language translation Others: Text Analysis Email filters Search Result Sentiment Analysis
14. Challenges in NLP
16. Challenges in NLP : Textual Accessibility Visual Auditory Speech
17. Accessible = Comprehensible
18. Challenges in NLP : Textual Accessibility Transcription Speech recognition Language Translation
19. Case : Recognize sentiment in Text
21. Amazon and Yelp reviews 1 : Good 0: Bad
22. Pre-processing data
23. Text Processing
24. Tokenization
25. Out Of Vocabulary (OOV) EXTRA WORDS unknown tokens (UNK)
26. Stop Words
27. Stemming
28. Lemmatization
29. Padding
30. Tools! • Python 3.+ • Jupyter Notebook or Google Colab • and using a framework for ML
31. ¿Why Tensorflow? Whether you’re an expert or a beginner, TensorFlow is an end-to-end platform that makes it easy for you to build and deploy ML models.
32. Trying in Google Colab bit.ly/2TU9MJz
33. Challenges in NLP : Textual Accessibility Transcription Speech recognition
35. Challenges in NLP : Textual Accessibility Language Translation
37. Next steps Try new tools : - Recurrent Neural Network (RNN) - Long short-term memory (LSTM)
38. Next steps Increase your corpus with diverse data. - PDFMiner - Tweepy API
39. Next steps Datasets with native or indigenous language or code mixing
40. “We can build a much brighter future where humans are relieved of menial work using AI capabilities” - Andrew Ng
41. Thanks! @womenwhocode #WWCode Keep in touch! linkedin.com/in/victoriaubaldo @vikyale