The Bot 101 [ Part 2 ]
The Bot 101 [ Part 2 ]
Thanks for reading and sharing the feedback on the Bot 101 – Part 1; where we talked about CUI, VUI, bots evolution, types of bots(personal bots, text bots, biz bots, and more), bot platforms (Slack, Facebook, Alexa, Kik, Email), and Bot Anatomy . Now in the Part II series of The Bot 101, we’ll be discussing about Human Involvement, Conversations and AI, Rich Interaction, Context and How to handle the Memory of bots.
Architecting a bot involves a lot of thinking about interfaces and its channels. While designing we’ve to think about the Bot’s branding too, which is all about how our bots will be shipped in the market for customers. In branding we’ve to look in to the aspects of visual branding, logo, stickers (to indicate the intent, state, and context), Images, Naming (Functionality, Brand Name, Trademarks), Personality (Environment, Audience, Jobs to be done, Runtime variations, Locally relevant, Social Relevance, Existing Branding, Values).
- Name: BigDataTeacher
- Environment: Work
- Audience: Adults aged ~20–60 using the bot while reading/writing content in English on Slack
- Task at hand: Find definitions for tools/terminologies in big data space
- Runtime variations: None
- Locally relevant social acceptance: Global safe for work bot
- Service branding: The brand is pleasant and productive
- Values: Focus on the service, be as transparent as possible
- Personality: Simple, getting things done, friendly but succinct, minimalistic, nonintrusive, even dry
We cannot talk about a bot’s brand and personality without exploring how humans help bots. Bots get their personality from human designers and copywriters, bots can fail over to humans when they cannot handle a conversation independently, and bots might require human supervision to make sure they keep providing the service that the brand offers. In some cases, humans and bot personalities work together to provide a great service.
Bots can work 24/7, no days off, no sleep and more. And the bot gets a question and gives an answer. At first every question is sent to a human for an answer, and the bot acts as a simple router. As time passes the bot learns that certain questions have a distinct, repetitive answer. (What is the meaning of life? The answer is always 42.) The bot then moves to a supervision mode, where it offers the answering human a suggested answer. Once the bot becomes confident in the answer (passing a certain threshold of confidence set by the system), it can answer the question directly without human supervision. As time passes, the bot becomes more and more proficient and requires less and less training and supervision.
Bots can add their own slash commands to this interface. Picking the right name for your slash command contributes to its usage. Slash commands need to be linked to your bot together with the functionality the command provides.
As we’ve seen, the designing a reliable personality, giving careful consideration to branding and language, and mapping out the possibilities for human intervention can be critical to the success of a bot.
And the bot’s primary and core principle is the core ability to adapt to change. And AI is underpins bots in many cases, now we’ll look in to the common AI services and how they can help us to build and design a bot. AI today is not a single thing, but a few set of tools that we designers and bot developers can choose to use in order to build a great conversational bot.
NLU (Natural Language Understanding) is one of the core feature from AI, supports in building bots. Example it helps get the intent from multiple phrases like “ I want to buy a cognitive book”, “I wanna get a cognitive book”, “Let’s purchase a book for cognitive study”, and hence the intent is to buy book. Many tools are available to convert text to entities to give an example wit.ai, automat.ai, smarterchild, and more. Extracting the intent and entities are very vital in the aspects of conversation management. And the conversational management is a high level of artificial intelligence, in which AI understands the context of the conversations and very much know to navigate from the conversations contexts to sub conversions. Intent extraction is hard and very complex in Conversation too.
As of now we the human kind had to adjust ourselves to app and software, rather than the other way around. But now our chatbot is new interface that is establishing human interface. With this we can create a connection with users that is the way very much stronger than what web, and mobile apps can ever manage.
Bots Life Cycle ?
Bot Onboarding – It is the very first interaction to users see from the bot; which could be a simple welcome messages sends to the users. And during this phase bot declares its purpose in the context of the conversations. Some bot itself will introduce to the user.
Teach user how to use the bot – In this stage we have to educate users to how to use bots. In this stage we need to depends on lot of information something like what to exposure and many more details.
Configuration – once the above two steps is done, then we have to configure it, some think like grant access to third party software, and more.
Inciting users to get value from bot – In this stage it’s better to incite the users to actually use the bot as part of their journey.
Setting the tone and personality – setting the tone of the bot while on boarding is very much vital to get the context.
Team Environment – Above all with the one on one interactions, but we have to take into considerations of team environment too. The biggest challenges we have when we onboard for team would be is that the team members might not be informed that the bot was invited to the messaging platform. And to be a good practice, DMing all the users in the team without explicit permission from an admin is not advised. And onboarding a bot to a team is very much similar to onboarding a new friendly human team member. And bots in teams act as a team members.
Do we Need AI for Designing Bot?
It’s not mandate to have AI knowledge while designing of Bots, but AI is the technology that underpins bots in several use cases. And AI is not equal to Bots. AI helps with set of tools that designers and bot developers which can be used to build conversational chat bots.
AI Helps Chatbot in the below way,
1 – NLU ( Natural Language Unit ) makes the bot to drive intent, and which is the technology that underpin’s our bots conversations. And in simple NLU helps to translate between user inputs such as – “ I want to buy a Chatbot book”, “I Wana get a Chatbot book”, or “Let’s purchase a Chatbot from the conference”, and the intent is simple to get a book.
2 – Conversation Management is very vital to maintain the multistep flow in with in a Conversations. For Chatbots, Conversation Management is one of the toughest job to pull out the right conversations.
3 – Image Recognition / Computer Vision – Bots can now use an image processing services, recognize images, spot emotions in photos, and extract text from images.
4 – Prediction – AI Usually does a good job of finding patterns and predicting outcomes based on past data.
5 – Sentiment Analysis – One of the unique AI service that is useful for conversation is sentiment analysis.
Rich interaction and controls are a very much great way to simplify, optimize, direct, enrich, and sometimes just replace text-based conversations. In a task-led conversation we might want to over-index on rich controls, in order to work around lengthy conversations. Files, Audio, Videos, Images, Buttons, Templates, Links, Emojis, Events, Menus, Slash commands, Webviews are options to have our chat bot with rich interactions.
Context and Memory:
When possible, bots should not time out context — if a user starts an intent, they should be able to come back to it after a while and the bot should remember the context.
Managing context and memory is probably one of the hardest aspects of designing bots, so we should take little more cautious steps while designing our Chatbot.
Thanks for the reading the Part II of The Bot, and we will be having series Part III and Part IV in due course. We will talk about Bot Discovery, Process Overview, and multiple use cases in the third series of Bot 101, then we’ll have final thoughts around Bot Building Overview, and testing on the fourth series of Bot 101.
And the conclusion is bots are a great new hammer, but not everything’s a nail; so we need to explore which use cases are better for this new and great interface. And with bots any enterprise or business can have their engagements with higher degree and value; hence the right bots are bringing best values.
Ref. Designing Bots, Amir Shevat
interested? questions? feedback? Let us have firstname.lastname@example.org !
Please subscribe to www.dataottam.com to keep yourself trendy on ABCD of Data (Analytics, Big Data, Bots, Cloud Computing, Digital).