~/talk

Mindmeld: An Open-source, Deep-domain Conversational AI Toolkit

links:none on file
slides:none on file

Mindmeld is an open-source, Python toolkit (licensed Apache 2) that builds production-grade voice and chat assistants by combining pattern rules, ML models and IR instead of end-to-end black boxes. It powers Cisco WebEx assistant, uses Elasticsearch plus TensorFlow, scikit‑learn, numpy and SciPy, and provides a dialog manager, entity resolver and QA layer. Typical training needs are 500–1,000 queries for domain/intent and 1,000–10,000 for open entities; deployments hit 95%+ entity accuracy when tuned and can run cloud or on‑prem.

AI-written summary · about

  1. 0:00Intro
  2. 1:05Conversational interfaces are the next frontier
  3. 2:12Building production-quality conversational assistants is one of today's hardest Al challenges
  4. 3:58Today's machine learning tools are insufficient Cloud NLP Services
  5. 6:46The Mind Meld Conversational Al Platform
  6. 7:33Github
  7. 7:53Natural Language Processor understands the user query
  8. 8:48Domain Classifier assigns the user query to a pre- defined domain
  9. 9:21Intent Classifier determines the user intent
  10. 10:16Entity Recognizer detects all relevant entities in the user query
  11. 11:27Role Classifier assigns role labels to extracted entities
  12. 12:05Entity Resolver transforms each extracted entity into its canonical form
  13. 13:09Language Parser clusters extracted entities into meaningful entity groups
  14. 15:22Dialogue Manager directs the flow of the conversation
  15. 16:46Application Manager is the platform orchestrator
  16. 17:51A sample interaction with an enterprise voice assistant
  17. 18:16Domain Classification
  18. 18:38Intent Classification
  19. 18:54Entity Recognition
  20. 20:14Role Classification
  21. 21:04Language Parsing
  22. 21:34Dialogue Management & Question Answering
  23. 28:33Entity Resolution
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.