The age of touch could soon come to an end. From smartphones and smartwatches, to home devices, in-car systems, touch is no longer the primary user interface. (source: Design News)
During this talk, Christophe, Principal Consultant at GraphAware will walk you through the design of building Conversational Bots. To this end, he used Amazon Alexa and combined it with a Natural Language Processing stack backed by a Neo4j Graph Database.
You will discover the basics of an Amazon Alexa skill and how the user experience with voice devices can be enhanced with graph based algorithms such as recommendations.
19. Chatbots history
• Eliza (1966) mimicked human conversations by matching user
prompts to scripted responses. It was able, at least for a time, to
pass the Turing test
20. Chatbots history
• Parry (1972) – Inexplicably simulated a person with paranoid
schyzophrenia. Parry was more advanced than Eliza.
21. Chatbots history
• 1988 – Jabberwacky
• 1992 - Dr. Sbaitso
• 1995 – A.L.I.C.E
• 2001 – Smarterchild
• 2006 – IBM’s Watson
• 2010 – SIRI
• 2012 – Google NOW
• 2016 – Bots for Messenger, TAY
• And Alexa ? Ask her !
23. Our conference assistant
• Tell me how many sessions ?
• Tell me how many sessions about a specific topic ?
• Recommend me a session based on my topic preferences
• Find me a session about a specific topic with NLP
Alexa should be able to :
25. Voice command
• Give a name to our Alexa skill -> CONFERENCE ASSISTANT
• Define our first intent name (code) -> talksCount
• Define our first utterance
26. ALEXA, ASK CONFERENCE ASSISTANT
(skill invocation)
HOW MANY SESSIONS TODAY ?
(skill utterance)
INTENT DETECTION
SPEECH TO TEXT
TEXT TO SPEECH
LAMBDA FUNCTION
OR
YOUR API SERVER
ALEXA VOICE SERVICE
27. Utterances
• How many sessions ?
• How much sessions ?
• How many talks at the conference ?
• How many talks today ?
28. Intent detection
• 2 main types of detection :
-> Pattern matching (how | count)* many (session | talks) today?
-> Classification machine learning, neural networks, word2vec, ..
33. ALEXA, ASK CONFERENCE ASSISTANT
(skill invocation)
HOW MANY SESSIONS TODAY ?
(skill utterance)
INTENT DETECTION
SPEECH TO TEXT
TEXT TO SPEECH
LAMBDA FUNCTION
OR
YOUR API SERVER
ALEXA VOICE SERVICE
34. Slots
• Variables in the utterances
• Should be filled with possible values
• Slots are sent in the intent request to your API
37. Adding personalization
• If you have informations about the attendees and their topic
preferences, you can easily get the sessions in accordance with it.
41. GraphAware NLP
• Graph based Natural Language Processing framework
• Information Extraction, Sentiment Analysis, Data enrichment
(ontologies and concepts), similiraty computation, knowledge
enrichment, …
• Beta program, first public open-source version to be released very
soon
45. Tackling the last intent
• Question / response game
• Treat question as text and process NLP
• Relate question text with Session abstracts texts
• Special type of utterance with only one slot {text}