Published on | Prototyping: From UX to Front End — Medium Mrinalini Sardar

The rise of Chatbots & Conversational UI & the way it impacts Search in Content Management Systems (CMS).

Conversations — The Way We Communicate

When was the last time you had a meaningful conversation?

A conversation with someone really leads to some answers, hopefully talking to someone who makes sense and can understand the question. We often speak to many people with whom we have had difficulty making conversations, either because they were not good listeners or because their perception of the question or context thrown by you were completely off target. The interesting part is, if humans don’t understand something, they might mask the non-understanding with a wrong perception of the context but machines have no ego and they will declare their inability to understand the problem. Very rarely do people want to understand the core context and make the effort to find out. But those who do will lead to being the real problem solvers.

Conversations through messaging are on the rise

In today’s age we tend to message more than we speak. Conversations through messaging tools are on the rise. A new generation of tools are rising that enable brands to communicate with customers in powerful new ways at reduced cost. Retailers and technology firms are experimenting with chatbots, powered by a combination of machine learning, natural language processing, and live operators, to provide customer service, sales support, and other commerce-related functions.

Chris Messina of Uber recently coined the term “conversational commerce” to describe this movement, which he defines as:

…utilizing chat, messaging, or other natural language interfaces (i.e. voice) to interact with people, brands, or services and bots that heretofore have had no real place in the bidirectional, asynchronous messaging context. The net result is that you and I will be talking to brands and companies over Facebook Messenger, WhatsApp, Telegram, Slack, and elsewhere before year’s end, and will find it normal.

Chatbots are our new best friend!

We can now have conversations with machines so easily. The rise of conversational UI and chatbots have given us friends who try to understand our questions and give us answers. How awesome is that. Yes, the accuracy of systems are questionable. Of course our machine friends are learning all the time, just like humans are, while making conversations. And as humans & machines learn conversations over time, the differences will become blurry and cease to exist.

Chatbots: Also known as an interactive agent, it is a computer program which conducts a conversation via auditory or textual methods. Such programs are often designed to convincingly simulate how a human would behave as a conversational partner, thereby passing the Turing test. Some of them use sophisticated natural language processing systems, but many simpler systems scan for keywords within the input, then pull a reply with the most matching keywords, or the most similar wording pattern, from a database.

Common examples are Apple’s Siri, Microsoft’s Cortana, Google’s Search Assistant, Facebook’s Messenger and many more brands out there.

Common chatbots in the market

“I don’t know anyone who likes calling a business. And no one wants to have to install a new app for every business or service that they interact with. We think you should be able to message a business, in the same way you would message a friend.” — Mark Zuckerberg at F8 in 2016.

Conversations with Machines

How did we get here? A lot of technology and the understanding of human behavior. To begin understanding conversations with our new friends, we need to understand some of these technologies better. Just to skim through some of them quickly and how it will affect building accurate chatbots.

Artificial Intelligence, Machine Learning & Deep Learning Intricacies

Natural Language Processing (NLP): When the machines began to understand and fruitfully process large natural language data then the true interaction between machines and human (natural) languages started.

Lots of challenges in NLP to achieve accuracy but Machine Learning helped in leveraging it by helping it to learn with large amounts of data and unlearn the aberrations in it.

Machine Learning(ML): A type of Artificial Intelligence which helps machines to learn things without explicitly programming them. The idea here is to train your chatbot with a lot data that is meaningful.

Deep Learning: Deep Learning focuses even more narrowly on a subset of ML tools and techniques, and applies them to solving just about any problem which requires “thought” — human or artificial.

To understand the technological intricacies: Read here

The Evolution of Search — Forrester Report 2017

Search — As basic bread & butter

The way we conduct search is usually typing in a ‘search engine’ exposing filters which are switched on & off to give more refined results. As an example, lets think of Dropbox, that we all can commonly relate to.

The scenario is to look for backpack images that are large sized pngs, and have been added recently on dropbox.

Dropbox has limited search functionality and the way we conduct search is through relevance query mapping and sorting functionality to refine the results. See the video below to understand a typical dropbox search workflow for the web.

<a href="https://medium.com/media/baf63ad9e3418561b57853a9f00086a4/href">https://medium.com/media/baf63ad9e3418561b57853a9f00086a4/href</a>

Now, imagine if you could just message or talk to dropbox to find those backpack images quickly.

Search — In Enterprise Content Management Systems

The search capabilities for Enterprise Content Management Systems (CMS) is much refined than content repositories like Dropbox or Box because there are tons of assets that need quick access and with multiple parameters.

For example, if a marketeer wanted to quickly find a backpack image which is a png, it would be easily possibly. But moreover, if a backpack image of a particular file size, or file type, or modified, or published on a particular date needed to be found, that would be possible as well through an interface of filters and checks placed in the Enterprise Search UI.

Old keyword-based enterprise search engines are soon going to become obsolete. Cognitive search is the new generation of enterprise search that uses artificial intelligence (AI) to return results that are more relevant to the user or embedded in an application issuing the search query. Forrester defines cognitive search and knowledge discovery solutions as

A new generation of enterprise search solutions that employ AI technologies such as natural language processing and machine learning to ingest, understand, organize, and query digital content from multiple data sources.

IBM Watson Conversation Tool — Build a Chatbot without Code

Search — As Conversations

I tried re-imagining this search workflow as conversational UI and built a basic chatbot through IBM Watson Conversation tool. Here is a brief demo.

<a href="https://medium.com/media/a8afe9d3526a798065c0282a20fb7af3/href">https://medium.com/media/a8afe9d3526a798065c0282a20fb7af3/href</a>

Conversations without Code

The IBM Watson Conversation tool is a boon for designers as well as users who want to build chatbots without code.

What I defined were intents, entities & dialog and I was good to go!

Here is a video that help you understand the terms quickly.

<a href="https://medium.com/media/db575cd162d0a10b7f2616acaa4e3df7/href">https://medium.com/media/db575cd162d0a10b7f2616acaa4e3df7/href</a>

To get into the depth of the definitions & specificity to the search context, here are some examples.

Intents: It is a category that defines the users goal or purpose. It is symbolized in the tool as #. Keep the intents short and support it with user examples.

This video explains Intents (#) in detail.

<a href="https://medium.com/media/f421063e32d030018d43cc5beff0604f/href">https://medium.com/media/f421063e32d030018d43cc5beff0604f/href</a>

So for my scenario, the Intent was #Search. User Examples I supplied for the Search Intent was <I need backpack images> or <I want all images uploaded recently> or <I want all images to show up>.

Screenshot of my defined Intent as #Search from Watson Conversation Tool

Entities: These represent a class of object or a data type that is relevant to a user’s purpose. By recognizing the entities that are mentioned in the user’s input, the Conversation service can choose the specific actions to take to fulfill an intent. It is symbolized in the tool as @.

This video explains Entities (@) in detail.

<a href="https://medium.com/media/335035bcd182fc7ac05cba90183a9317/href">https://medium.com/media/335035bcd182fc7ac05cba90183a9317/href</a>

So for my Search scenario, the typical Entities were @Added, @Modified, @FileSize, @FileType etc. because I wanted a user to be able to search on these parameters.

My defined Entities from Watson Conversation Tool

Entity values that I supplied for @Added were <Last Few Hours> <Month> and <Recently>. You can also support each entity value with synonyms & patterns to make the conversation more rich. I supplied for Entity @Added, Value <Last Few Hours>, Patterns <added in the last few hours> or <Added in the last few hours> or <last few hrs>.

My defined Entity Values for the Entity ‘Added’ from Watson Conversation Tool

Dialog: The dialog uses the intents and entities that are identified in the user’s input, plus context from the application, to interact with the user and ultimately provide a useful response.

Watson Conversation builds the dialog as a tree structure with dialog nodes. Each dialog node has a trigger and a response. When a trigger which is a condition matches a users response, then the response is generated. An important thing to remember was that node sequence matters.

This video explains Dialog in detail.

<a href="https://medium.com/media/9d2f5ffadfbdc3c19269f43c8fdd940f/href">https://medium.com/media/9d2f5ffadfbdc3c19269f43c8fdd940f/href</a>

My dialog construction from Watson Conversation Tool looked like this:

So for my Search scenario, a dialog would be broken into 4 triggers:

Welcome: If the bot recognizes the trigger ‘Welcome’ then it responds as <Hello. I am Jeeves. Your Search Assistant! How can I help you?>

Dialog for Trigger Section 01: Welcome

2. Intent #Search: If the bot recognizes the trigger ‘#Search’ then it responds as <Sure. Would be glad to help. Here are all your backpack images. Is there a particular file type you are looking at? PNG/ JPEGS or generally all images?>

Dialog for Trigger Section 02: Intents

3. Entities: If the bot recognizes the trigger as entity ‘@Added’ with different values then it would respond accordingly. As an example, trigger ‘@Added(Recently)’ would invoke a response as <Relax, while I hunt your recently added images> or a trigger like ‘@Added(LastMonth) would enable a response <Relax, while I hunt your images added in last month>

Dialog for Trigger Section 03: Entities

4. Anything Else: For all triggers the chatbot does not understand, the responses need to be added to get more clarity on the trigger.

Dialog for Trigger Section 04: Anything-Else

Conversation Tones Matter

Just like when we talk to a friend or parents or a client, our conversation tones or words used changes. Similarly for chatbots, emotions matter too. Adding emotional responses in terms of response variations matter. This is a part I am yet to touch upon and will leave it for another time. But if you want to dig into emotional quotient of conversations, read up here on Emotional Chatbots.

Continuing the Conversations

I wondered why I built a chatbot to solve this search issue in a Content Management System. It’s because Conversational UI is the next big thing.

We learn language to communicate from the moment we are born. Our parents teach us how to talk, and that’s how we learn to interact with the world around us.

Language is the most natural interface humans understand, and that’s the interface that bots use. Instead of needing to constantly learn visual interfaces, bots will enable us to naturally use language, the first interface we were ever taught.

Talking to a bot in the future will be talking to a really intelligent person who has instant access to entire databases of information and it will remember and bring it up in the right context. They will instantly be able to find things for us when we try to search anything.

But, what will be interesting to see is whether the bots can truly answer when we try to search for the meaning of our existence.

Read more to Chat more!

Resources & References

11 Examples of Conversational Commerce and Chatbots

How Bots Will Completely Kill Websites and Mobile Apps

Cognitive Search Is The AI Version Of Enterprise Search

The monumental rise of conversational interfaces

Emotional Design for Chatbots: How to make humans love your bot

Check out more of my experiments @mrinalinisardar.com

<a href="https://medium.com/media/b85dfbb5286d8a25cf2e754b9462cf45/href">https://medium.com/media/b85dfbb5286d8a25cf2e754b9462cf45/href</a>

Search Seamlessly with Conversations was originally published in Prototypr on Medium, where people are continuing the conversation by highlighting and responding to this story.

Search Seamlessly with Conversations

Cool stuff, every week

More Articles