Gaurav Jassal:- Good Morning everyone, I am Gaurav from UX Gorilla.
I welcome you all to the webinar on “Conversational Experience” organized by Design Jungle and powered by UX Gorilla.
Today we have a very special guest with us, Monika Khanna. She will lead this webinar and taking us through Conversational Experience basics.
Monika specializes in solving complex problems and creating seamless experiences. During her 9 years in UX, she has worked with many companies like Birlasoft, Bennett Coleman (commonly known as The Times Group), Affle Appstudioz, Oodles technologies and currently she’s associated with Sapient as a Manager UX.
She’s a firm believer in knowledge sharing, she frequently speaks at events and writes articles on User Experience.
Next up, we will have Monika presenting the world of “Conversational Experiences” from her lenses.
Before we go further I want to tell you that throughout the presentation, you can put your quires in the comment section, will try to answer the same.
So let’s get started, over to you Monika.
Monika Khanna:- Hello everyone, thanks for joining for this session. I’m going to talk about “Conversational Experience” the latest trend for human-computer interaction.
In this course of time, I will be covering these topics, overview of the Conversational Design, Chatbots & Voice Recognition Systems, Framework, Conversational Entities, and Quick Tips.
Let’s see how the evolution of interfaces is done in recent times.
Earlier, we were using web browsers like websites were the primary mode of information exchange, and by 2009 smartphones picked up the market and the mobile apps became the preference mode of information access, and these days we have messaging apps, conversational design is an emerging platform growth of chatbots voice recognition systems are becoming smarter with artificial intelligence technologies, and thus creating rich experiences. Web browsers and smartphones were based on GUI (graphical user interface), which comprises of like buttons forms fields textual descriptions, here user does not feel like someone is listening to him someone is talking to him. Whereas messaging platforms give him that kind of Liberty. As per Don Norma, “The real problem with the interface is that it is an interface”.
Next step comes as Conversational Interface, A user interface which is a touch point that enables you to use a language, a human conversation between computer and human to get a task done. It can be via a text message or a chart or voice-based conversation to interact seamlessly. This makes the user feel like someone is talking to him or he is talking to any other human as he is talking in his daily life and thus does not get any sense of hindrance from an interface point of view to convey his intent clearly. So these days we see there are those interfaces used by the Slackbot, Facebook Messenger, Siri, Alexa etc. Now the point comes in play that why conversational design. So let’s see what are the key benefits for this kind of conversational and the latest trend.
Firstly, it feels more natural than apps as I said that the user feels like he is talking to any other human. So and he can talk in his own language so it feels more natural than apps he does not have to you know to learn new skills to interact with an interface like he does not have to learn any kind of gestures, UI elements or navigations etc. This allows personalization because based on his responses your assistant can take the flow ahead. So this sparks an instant connection with the user and the assistant no need to download a separate mobile app, there is no judgment that whether the answer is right or wrong there is no such kind of judgment for the user, and of course it is far more engaging than those static visual counterparts and builds the trust among the end-user(s) through meaningful interaction.
As per Golden Krishna, “Language is the most powerful, useful, and effective communication technology ever, period”, and that why you can see. Starting with chatbots there are two types of interfaces these days, one is chatbot and the second one is voice-based recognition systems. So let’s see what is a chatbot.
A chatbot is a computer program that attempts to simulate conversation via text or voice interaction. A user can ask a chatbot any questions, make a command and the chatbot will respond back and performs the requested action. The evolution of AI technology, they are expected to become more empowered to offer better language processing capabilities.
Golden Krishna rightly said that “The best interface is no interface.” And that’s why we got the evolution of voice recognition systems like Alexa, Siri, Google home etc. Which allows the user to interact with the computer through voice similar to chatbots because chatbots can be voiced based as well as text VUI (voice user interfaces) are hitting the mainstream and we find them in our smartphones, Television and other range of products. According to Google 20 percent of all search queries are on mobile and Android are voice-based and according to Gartner by 2020, 30 percent of web browsing sessions will be done without a screen.
These are all based on the language processing. NLP the concept of NLP (natural language processing) which is an area with interactions between the computer and human (natural) languages, in particular how to program your computers to process and understand that how the users intend for that particular task is, and how to get that done easily for him, and then analyzed those large amount of natural language data that how the responses are coming and how you can make your bot assistant better.
The trend is evolving from NLP to NLU these days which is “Natural Language Understanding.” It has gone one level ahead now, so it is now natural language understanding. Which is a subtopic of natural language processing and artificial intelligence? Which deals with machine understanding and reading those comprehensions, so if AI power chatbots can be done there will be pure magic. Because through machine learning chatbot can learn appropriate answers to specific user queries overtime for which they may no longer require any kind of training and make your chatbots and your assistants much more efficient.
I would like to throw some light on the framework because people generally start thinking about how to manage the user flow. I would like to summarize it in these steps.
The scope than the personality development of your bot or resistant then prioritizing the list the must-have features and then the flow, how your bot or the chatbot or your assistant should follow the flow, and it has to be personalized because depending on what user will be selecting he’ll be taking the flow ahead.
So let’s dig deep into each of these and try to understand that how we can cover and use this kind of a framework, so I would like to define scope, the scope is nothing but basically explains. What the chatbot or your assistant is all about. so making clear what people can expect from your digital assistant defining the exact goal of the bot. what they can expect from your bot what it can be done through your assistant and then considering use cases where chat works best to solve a user query.
Then comes as a very important step in the personality development of the bot or the assistants. so then works on the personality the tonality of the bot like you may see how Alex, Siri or Google home are interacting with users the tone the language the way they respond and talk and the personality behind them so then you need to work on the personality and the tonality of the assistant you are looking for as per the problem you’re solving for the user and then decide, how the bot will talk and act in every situation, define how the bot shall respond to the end users queries.
The third step comes as prioritizing the list of must-have features. List out some of the specific tasks a user can perform with the assistant to decide value proposition. What value you are adding in his life. So must have features without whom they cannot go further or this is the primary pain, which they are trying to solve in there life, and then try to prioritize that list based on his needs and focus on information that any user can find useful through your assistant. so it’s not about perfection, it’s about starting small and then moving fast.
The fourth step as I said it is tasked flow and convert those tasks flow to a chat that how your bot will be talking and how the user will feel like he is talking to any other human. his friends, his relatives, his co-workers or anybody else like he talks in his daily life. Then write a script based on the personality of your assistant, which you have developed in step number two and then start designing for every possible situation so the designing part comes quite later. when you have a complete groundwork done for your assistant.
Moving on with the conversational entities. I have covered the basic entities, which comprise of the welcome message, invocation, requests, intent, context, and the response. So I’ll cover all of them in deep starting with a welcome message.
Welcome messages about setting that first connect with your users. So like in the real world you start your conversations by greeting someone. So similarly your board your assistant should start greeting users and set that first connect, like Siri you see a message what can I help you with, go ahead I’m listening, Google assistant with some animation, Cortana is using here are a few questions you can ask me. so this is the kind of example for the Welcome message.
Invocation is the trigger used to invoke your assistant. what would be that trigger, how you want your assistant to be the trigger and invocate provoked for that application, for example, Siri use “Hey Siri,” Google Assistant “Hey Google” and for Alexa is “Alexa” which is the wake word by default and you can change it also. so similarly you have to work on these for your assistance or the bots or the voice-based systems.
Then comes the request, what is the request? It could be a question. It could be a command or a query asked by the user, through your assistant or the bot, for example, it could be how’s the weather in New York today? what are the nearby restaurants? who was Oscar Wilde? Add them to my cart. so these are four examples of the requests a user can ask the assistant.
Then comes as the concept of intent this is the most important because you need to understand that a user can ask the same question in different ways. so there can be end number of permutations for asking the same question. So work on the user utterances. How user can ask this and make your assistant that intelligent enough to respond based on the level of at utterances and then as I said earlier, that gather the large amount of data which you’re gathering from your user’s responses and work on those are utterances.
So the different ways of calling the same thing, for example,
“I’d like to eat pizza,”
“order a pizza,”
“Do we have restaurants nearby where Pizza is available?”
“Is there Dominos nearby.”
So the user can ask these different ways of the same type of question that he is looking for pizza the action from your assistant could order pizza. So the assistant should be that intelligent that he/she should understand that here the required parameters are people count, time, restaurant, type of pizza etc. So supposing the user said that I would like to eat pizza. The next set of questions from assistant could be like how many people? What time would you like to have? What types of pizza do you prefer Veg or Non-veg, which would you like to have? Just to understand that without having these required parameters people count, time, type of pizza, or the restaurant whether you want to have it from Dominos or somewhere else.
The task cannot be done the action, the final action which is required here to order a pizza cannot be done so building your assistance and bots intelligent enough to understand that what are the required parameters for the task to be done and understand the user utterances to complete their intent.
Next entity which I would like to cover is the context which is also quite important because it represents the current context of a user’s request to the system should remember the previous conversational turns that what is going on? what he is trying to cover? so, for example, asking a Google assistant that who was Oscar Wilde? gives the user some result and if the user asks can you tell me more, so the more should reflect back to Oscar Wilde where he is not mentioning the name of the author but still the assistant understands that they are referring to Oscar Wilde and it shows the relevant results. So this is the context, that he understands from the previous conversations and this is very essential for making a conversation sound natural rather than robotic because this is how we conversate in our daily life.
Next, how your assistant should response it could be in the form of acknowledgment a confirmation informational statement, so it is the response created by the assistant to serve the user requests the representations can be in many forms one could be a simple response text only, audio only or text and audio both. for example, the question was how’s the weather in New York tomorrow? and the response in the text which can be shown on the screen it’s going to be a pleasant day tomorrow and show whether or the response and voice could be it’s going to be a pleasant day tomorrow. so that’s the simple response representation.
The second representation of the response could be a visual card. which can display an image, title, subtitle, with some text and a link in the example, where it is shown when a rapper will come to the city and the response is in text and in the visual card shown on the screen on Tuesday, May 15. I see if there are tickets, should I see if there are tickets available? and then a visual card. so these are the representations in which the assistant can show his response.
Another way of representation for response could be a suggestion chip. Chips is nothing but like giving hints at responses to continue the conversation it helps user refine those topics because sometimes its faster and easier for users to tap a chip, then it is to say or type that response to continue the conversation for example what event are you interested in and then showing him some suggestion chips. shows this week, this weekend, popular ones, so that he can just tap on this chip and show you his intent. since the user has loved high expectations from AI chatbots and the visual of sorry voice-based systems they have low tolerance about error rate in chatbots. so we can easily solve this problem by limiting users input to just a few options like instead of asking openly what he wants to try to proactively give him some suggestions and from the wide range of questions try to narrow it down. As in this example shown what kind of bouquet would you like and the user can response from a wide range of queries like a sympathy bouquet, it’s for birthday etc. And then narrow it out to what kind of flowers would you like in your bouquet. so now the user has a limited range of options where you can select just the kind of flowers roses, sunflowers etc and then further narrowing it leads to what color carnations do you want? red, pink, yellow, and white. So now he just has to choose from a single category. so in this way you are helping him to get his task done by narrowing it the questions and giving him the few options to limit his input, and then the final question could be are you ready to place your order? again user has to choose from just a few options either yes or no.
Now I would like to give you some quick tips in the end that your bot our assistant needs a voice. so try if not only going by chatbot and the chats try to include the voice also. so that it makes more natural for the user be diligent of how the representation of your assistant affects the user experience. As I said earlier that how it will be representing how the tonality and the personality should be given a lot of importance. In this kind of conversational design copies, the new design is the ultimate mantra for designing a magical assistant because you have to work on the user utterances you have to understand how naturally user can talk, speak or write his queries or give you a command.
Give users time to read and see just so that he feels more comfortable. he gets your trust and leads him to more engaging in the content which you are trying to show him. Make sure buttons are functional at all points of the conversation. Right quick replies from a first-person perspective that is as a user would respond. so try doing some mock-ups like talking to someone and then creating your script in the fourth step which I showed you earlier. And then think that how your user would respond to this. Also, learn to say no tactfully don’t be rude but again if this is out of the limit or the scope we define in the first step of your assistant maybe you have to say no or if you don’t understand the assistant does not understand something he can just simply say apologies. I don’t understand that humble yet form.
Then the next step which I can give you for this kind of interface is to add emotions to your conversation because emotions are the thing we all need and since we are humans. we Deal in emotions, so if you can include those emotions in your conversation. it’s going to hit your user’s heart.
To summarize I can say that conversation design by a bot assistant or voice-based systems is better as it understands customer needs his wants and his surroundings much better than the interfaces we have used till date. Thank you so much, everyone, over to you Gaurav.
Gaurav Jassal:- Thank you so much, Monika, for sharing your knowledge on conversational experience with us and thank you so much people for joining this webinar. We had a lot of great interaction.
On behalf of UX Gorilla, thank you for joining us today. We hope you found this information helpful. If you missed anything no worries still you can watch this recorded webinar on a youtube channel.
Thank you so much, guys.
All right, guys, we have received few questions. Monika let’s quickly see some questions right.
We have received our first question Garima, she would like to ask:-
Q1) How we will set the time to read or see this actually may vary user to user also depends on age, eyesight, listening power. what is your thought on this?
Monika Khanna:- Yeah! thanks, Gaurav and thanks Garima for the question. Yes, that is why you need to see that how users are responding and obviously as I earlier said that start talking to someone and then start building your script and see that now how the users are responding and do your testing well that based on the age, eyesight. you can clearly said but then you have to also understand that the conversational design is more about getting less into the interface and then write some direct content. which user can relate to as for his conversation and then do the testing that how users are responding. whether they are able to read there of course of itself. It should be in some chats kind of format. where you just write two lines of online something like that and then see how much time users are taking to read it and then you know iterate it accordingly.
Gaurav Jassal:- Great! so, Monika, we have received our next question from Sumeet Sharma:-
Q2) How do we decided which platform Facebook or slack etc. We should focus on while building our boat? What is your thought on this?
Monika Khanna:- Thanks Sumeet for the question. Yeah, so like Facebook and slag bots these are all examples which you can refer to but as I have given you some tips according to your bot, according to your business scope. You need to understand that what is the difference between these bots and what kind of scope and the business goal. Which you are trying to achieve through your assistant otherwise you can follow these quick tips. which I have tried to given you because these are generic. so you can use it and then try that how you can set those connection with your users through your assistant or your bot. so these are very good examples. which you can refer to and there is no choosing between either one of them because all of them are good they are serving their business goals well. so you need to understand that as for your business goal. how you can match your user’s requirement better.
Gaurav Jassal:– Thank you one more question, Monika. This question comes from Anish Kumar:-
Q3) How to approach in order to get maximum of them covered?
Monika Khanna:- Thanks Anish for the question. As I earlier said that in this third step which I tried to give it as per the framework I am proposing that prioritize the must-have features. You don’t have to show everything or cover the maximum, as I said that you don’t have to be perfect in doing it right at the first time you can start with the small chunks and then build it moving forward and see that how your users are responding and covering the must-have features first and then later covering to the rest. so try not to do everything at once but try to see. if you can cover and prioritize the must-have features for the first time and then try to build over it once you are getting your users interaction. you winning their hearts and the conversation is going smoothly between the assistant and the users.
Gaurav Jassal:- Another question we have received from Amit Manchanda. He wants to ask you like if you can share another example of adding emotions to your conversations.
Monika Khanna:– Yeah sure, thanks Amit. As I said that the response from the assistant could be in the form of acknowledgment confirmation etc. so during that suppose someone has answered the question well as per your expectation and he is performing the steps very well so you know you can acknowledge the user by saying, hey great going or well done you have reached to this level just two more to go or something like that where you know someone feels that wow I am doing well. I am about to go for it just the way we greet someone, just the way I near dear ones of are doing something good. we know to acknowledge them, that’s good news you know so just like that adding some emotions and that’s why I recommend it in the very first tip that if you can include voice because without the voice the chat does not show the emotions right. so if you could include the emotions and if someone is not responding in the way it is expected you cannot be that harsh that and in a robotic manner your assistant can say oh looks like you know this is not working this seems some unexpected response. we can try in this way hey how about trying it in this way right. so like you know you are talking to someone who is younger to you and instead of making him feel that oh you are lost or you know for example your younger sister or brother is preparing for his/her exams and you feel like you know trying to motivate her in a way that fine. if you didn’t get or you didn’t score well and this term maybe you can do it better in the next term. so in the same way you have to get that emotion first through your script in your boats personality, that he is trying to help you to get your tasks done and that’s why as I keep saying not in the robotic fashion but sounding more natural and responding him with emotion these kinds of emotions as I’ve given you the examples. so I hope it helps.
Gaurav Jassal:- Thank You, Monika, so Anish he’s asking:-
Q4) How much important is the design look and feel of a chatbot, also what level of coding will be efficient for a creating a chatbot?
Monika Khanna:- Thanks, Anish for the question. Yes, if we talk about chat bot then look and feel is important but as I said earlier the copy is the new design for this kind of conversational design because firstly, what users are going to see and respond is the most important mantra for your conversational design. The look and feel come next and it could be the second thing which you need to focus while designing any chatbot and but of course for the chatbot. It comes in the look and feels also that try giving a face to your bot right rather than just showing a dummy user icon or something like that try to build that as for his personality that so that user actually feels like, we are doing chatting through WhatsApp, Skype, Google Messenger Hangouts. We see someones face and then we feel that connection so try to give a face to your bot or assistant and that can be a part of your look and feel also.
Gaurav Jassal:- Alright! Thank you so much guys for joining this webinar. we had a lot of great interaction with you guys. Thank you so much for commenting and joining us.
On the behalf of UX Gorilla, we hope you found this information helpful if you missed anything no worries still you can watch this recorded webinar on our YouTube channel.
So signing off with Monika Khanna bye bye thank you.