Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Elastic Meetup: Estrarre business insights da file audio con AWS Transcribe, AWS Comprehend, Elasticsearch e Kibana

Elastic Meetup: Estrarre business insights da file audio con AWS Transcribe, AWS Comprehend, Elasticsearch e Kibana

Matteo Zuccon

April 20, 2018
Tweet

More Decks by Matteo Zuccon

Other Decks in Programming

Transcript

  1. Estrarre business insights da file audio con AWS Transcribe, AWS

    Comprehend, Elasticsearch e Kibana 20/04/2018 Florence
  2. Elastic 20th April 2018 About me MATTEO ZUCCON @matteo_zuccon [email protected]

    https://whiletrue.run Solution Developer | #ElasticMeetup #PyconNove #PyconIT
  3. Elastic 20th April 2018 • Speech to Text and Natural

    Language Processing • AWS Transcribe and AWS Comprehend • Use Case: Architecture and Implementation Agenda
  4. Elastic 20th April 2018 Speech to Text “Speech to text

    is the process of converting speech input into digital text based on speech recognition”
  5. Elastic 20th April 2018 Speech recognition Speech recognition is the

    ability of a machine to identify words and phrases in spoken language and convert them to a machine-readable format
  6. Elastic 20th April 2018 Speech recognition - Algorithms • Hidden

    Markov models • Dynamic time warping (DTW) • Neural Network
  7. Elastic 20th April 2018 Natural Language Processing “Natural Language Processing

    (NLP) is a branch of artificial intelligence that helps computers understand, interpretand manipulate human language”
  8. Elastic 20th April 2018 Natural Language Processing - Tasks •

    Language Detection • Key Phrases Extraction • Entities Detection • Sentiment Analysis
  9. Elastic 20th April 2018 NLP - Algorithms • Support Vector

    Machines • Naïve Bayes • Neural Network • Latent Semantic Indexing (LSI) and Latent Dirichlet Allocation (LDA)
  10. Elastic 20th April 2018 Algorithms – Hard work Pays off

    • So many Algorithms and Theory • How to set parameters? • Which library? Who trains (and continue training) the model? • Should I focus on the library and its usage or my application?
  11. Elastic 20th April 2018 Algorithms – Hard work Pays off

    • Know what’s going on • Build your own model or train it • Improve results for your use case • Keep control
  12. Elastic 20th April 2018 AWS Transcribe and AWS Comprehend “AWS

    Transcribe is an automatic speech recognition service. Analyze audio files stored in Amazon S3 and have the service return a text file of the transcribed speech” 4th of April: Amazon Transcribe Now Generally Available
  13. Elastic 20th April 2018 AWS Transcribe and AWS Comprehend “AWS

    Comprehend is a NLP service to find insights and relationships in text. It identifies the language of the text; extracts key phrases, entities; sentiment; and organizes a collection of files by topic”
  14. Elastic 20th April 2018 Alternatives – Cloud Providers • Watson

    • Text Analytics API and Bing Speech API • Cloud Speech to Text and Natural Language AWS vs Azure - Text Analytics: https://bit.ly/2IKWJzA
  15. Elastic 20th April 2018 Use Case Convert audio and voice

    (e.g. customer service calls) into written text and extract useful information for quick understanding of content, index information in Elasticsearch 6.2.3 for fast search and visualize the data with Kibana 6.2.3
  16. Elastic 20th April 2018 Customer service calls - Questions •

    How many positive recordings do I have? • How many customers are complaining (negative feedback) about my products? • Which is the sentiment about my products? • Which entities/key phrases are the most common in my recordings? Nobody wants logs everyone wants answers Elasticsearch and doubling down on “open”: https://changelog.com/podcast/292
  17. Elastic 20th April 2018 Conclusion • Speech to text –

    as a service – AWS Transcribe • NLP - as a service – AWS Comprehend • Search and aggregate – Elasticsearch • Visualize and distribute - Kibana
  18. Elastic 20th April 2018 Elastic – AWS Transcribe – AWS

    Comprehend Thank you Questions? @matteo_zuccon [email protected] https://whiletrue.run #ElasticMeetup #PyconNove #PyconIT https://bit.ly/2J7jde8