NLU Services - foxymoron/test GitHub Wiki

Services to choose from

  1. api.ai (Google)
  • Launched: 2010 (acquired by Google in 2016)
  • Pricing model: free
  • Interface: HTTP REST API
  1. wit.ai (Facebook)
  • Launched: 2013 (acquired by Facebook in 2015)
  • Pricing model: free
  • Interface: HTTP REST API
  1. Watson (IBM)
  • Launched: 2016
  • Pricing model: Pay as you go
  • Interface: HTTP REST API
  1. LUIS (Microsoft)
  • Launched: 2015
  • Pricing model: Pay as you go
  • Interface: HTTP REST API
  1. Lex (Amazon)
  • Launched: 2016
  • Pricing model: Pay as you go
  • Interface: HTTP REST API
  1. RECAST.AI
  • Launched: 2016
  • Pricing model: Free Tier + Contact Sales
  • Interface: HTTP REST API
  1. snips
  • Launched: 2017
  • Pricing model: Free to test + Pay per device
  • Interface: On Device

Benchmark Performance: (2,3,4 very close)

  1. IBM Watson, Microsoft LUIS, Google api.ai
  2. Facebook Wit.ai, snips
  3. recast.ai
  4. Amazon Lex

Benchmark Price:

  1. Google api.ai, Facebook wit.ai
  2. Microsoft LUIS, Amazon Lex
  3. IBM Watson

Summary of superiority:

  1. Google api.ai at price (free)
  2. Microsoft LUIS at speed (50% faster response)
  3. IBM Watson at performance (esp. at smaller datasets)
  4. Facebook wit.ai at extreme language coverage

References

  1. https://www.slideshare.net/KonstantinSavenkov/nlu-intent-detection-benchmark-by-intento-august-2017