chatbot workshop introduction.#digitized16
TRANSCRIPT
Voice, Natural Language, Notifications,API, and Conversation
How we build and designproducts in the future.
There are more thanthree million appscombined in all app stores
An average smartphone userhas 42 apps on his devicebut spends 90% of his timeon only 9 or 10 of them.
Present problemwith UI
25% of apps areonly used onceand 75% of users leavewithin the firstthree months.
The growth of mobileand its app-centric worldhas been the opposite of the web,and there is no analog of PageRankfor mobile apps.
Messaging is the mostwidely used applicationon the mobile platform,with chat-based apps
Already, more than15% search queriesare made on Baiduusing Voice Input.
HISTORICAL EVOLUTION
The Story of New Platforms
PC
Each platform requires recreationof the application layer, built and customizedto grow and scale the new platform.
Web Mobile
Time Mid-80’s
Clients Websites Mobile apps
Mid-90’s Mid-00’s
Applications
gestation phase growth phase
PC era
PCDesktop
WebBrowser
MobileiOS, Android
Internet era
Mobile era
Key Product& Companies
Paradigm
Platform
HYPER GROWTH PHASE
The shift is set to happen again
PAST PRESENT FUTUREhuman to human human to machine machine to machine
Google MindBoston Dynamics-Atlasvirtual reality & dronesare just a glimpseof things to come.
The primary interfacefor interacting with appsmight not be the app itself.
The age of appsas service layers
How does it work?
MACHINE LEARNING
MACHINE LEARNING
Supervised Unsupervised Adaptive
A model is prepared through a training processwhere it is required to make predictionsand is corrected when those predictionsare wrong. The training process continuesuntil the model achieves a desired levelof accuracy on the training data.
Supervised machine learning
Classification
Clustering
Learns patterns in input data when no specific output values are given.
Unsupervised machine learning
Learns by an indication of correctness at the end of some reasoning.
Environment
observation
Reward
Action
Adaptive/Reinforcement machine learning
A branch of machine learning based on a set of algorithmsthat attempt to model high level abstractions in data by usinga deep graph with multiple processing layers, composed ofmultiple linear and non-linear transformations.
Neural Networks & Deep Learning
The experiment started with a machine learning algorithmand a database of over 10,000 songs from more than 100 rap artists.
The machine produces rap lyrics that rivalhuman-generated ones for their complexity of rhyme.
The words must first be converted into phonemes.
Finding rhymes is then simply a question of scanning the phonemeslooking for similar vowels sounds
How machine Mines Rap Lyricsand Writes Its Own
For a chance at romance I would love to enhance
But everything I love has turned to a tedious task
One day we gonna have to leave our love in the past
I love my fans but no one ever puts a grasp
I love you momma I love my momma – I love you momma
And I would love to have a thing like you on my team you take care
I love it when it’s sunny Sonny girl you could be my Cher
I’m in a love affair I can’t share it ain’t fair
Haha I’m just playin’ ladies you know I love you.
I know my love is true and I know you love me too
Girl I’m down for whatever cause my love is true
This one goes to my man old dirty one love we be swigging brew
My brother I love you Be encouraged man And just know
When you done let me know cause my love make you be like WHOA
If I can’t do it for the love then do it I won’t
All I know is I love you too much to walk away though
Lyrics
CONVERSATIONAL UI
The rise of hybrid Interfaces
Command-lineThe command line was the originalconversational interface.
You’d input a textual command,hit enter, the computer would executethe command and print the answer.
IRCIRC already supported bots,massive group chat quizzes,polls and other typesof conversational applications
Each messagebecomes a mini appBlended interfaces, bringing the best of the command lineand GUI paradigms together.
OPERATORCompanies like Operatorare leading the way,designing rich experiencestheir clients can interact withdirectly, not by replyingsimply with text.
First impressions & Expectations
“Take me to the moon, Bot!”
Introduce yourselfYou only get one or two lines,so keep it short and to the point.
Having no visible interface means:
This thing can do whatever I ask him,so I’m going to ask him to make mea sandwich.
I have no idea what I’m supposedto do now, so I’m just going to freezeand stare at the screen.
Meeting synced! Did you know I can also findand book a conference room?
Ping! There is a meeting coming up in one hour.Would you like me to order lunch for 3 people?
Once the first interactionsare successful, the robot can beless verbose and more efficient.
Proactively suggest things to do.
Great! Find us some sushi!
Ping! There is a meeting coming up.Would you like me to order lunch for 3?
Validating input_Give hints _Aknowledge
Small please! Thanks. Great! Find us some sushi!
Ping! There is a meeting coming up.Would you like me to order lunch for 3?
What size t-shirt are you?We have small, medium, large.
Got it. Size small.And what color would you like?
Explain what went wrong...
brlbrbl
gray
I’m sorry, “brbrbl”? Is that a color?We have white, gray, brown.What color would you like?
Cool! So a large gray t-shirt!
Awaiting critical inputSometimes you need a piece of informationthat you absolutely cannot proceed without.
Schedule a new meeting tomorrow?
Am I busy tomorrow?
To do my job, I need accessto your schedule.Follow this link to connectyour calendar.
Seriously, you need to connectyour calendar here to enjoymy scheduling superpowers.
I can’t wait to start workingon your schedule!Please connect your calendarso I can do my magic.
What you need to consider
BUSINESS CASES FOR BOTS
(Most valuable bot}
Start with looking at the overarchingbusiness objectives.
_What are the business priorities this year?_Acquisition?_Retention?_What KPIs can we identify to measure value?_What point in the customer journey does this fit?
YOUR MVB
_Brand guidelines_Tone of voice _Consumers have different degrees of tolerance_Script out anticipated conversations and review
_Decide how much of your UI will be conversation driven rather than manual_Quality NLP and some degree of AI or machine learning._Lean on the universal UI_Remember: GUIs > DOS
Conversational DesignVs User Interface Design
Personality
_Store data securely and inline with any privacy agreements_Don’t ever break Facebook or any other platforms’ guidelines_Brand reputation can be fatal.
_Pick a solution that can scale up or down quickly.
_Allow reprioritisation and user testing to happen throughout the build_Establish a group of testers_Market your bot.
Security, Rules & Data
Scale
Development, Testing & Promotion