Sometimes all you need is to talk to someone. Someone who can cheer you up in their own way, someone who is so full of life and chatty that you forget all your problems in life. Someone that amuses you by coming better than your expectations. Everyone is not so comfortable about talking to other ‘humans’ about things, but there are some curious people who do talk to AI. Here, Ruuh comes to the picture.
Ruuh is capable of listening to one’s question, detect their emotions, learn about the user’s background and make appropriate replies and more. This enhances their bonding and the relationship they share with the user. It directly implies to more valuable and sensible chats between the chatbot and the user.
Ruuh is good at making conversations
Without the involvement of emotions, the existence of chatbots is useless. Just being able to reply without any personal connection makes the chat formal and many times uninteresting. A chatbot is interesting only if they are able to make conversations on the foundation of emotions being involved with it. About this, Microsoft says,
Building a conversational layer in Ruuh helps her develop relationships so users can be more open, more casual and more engaged. This leads to better, more honest and natural conversations that ultimately lead to added value and a better experience for users.
Aim of building Ruuh
Microsoft’s main aim behind building this AI-powered chatbot was to make it for the young, tech-savvy early adopters in India. It was already meant to be similar to Microsoft’s Chinese Chatbot named Xiaoice. Ruuh is more of a digital friend rather than just a digital assistant. Ruuh is a software that is not just a piece of code; it is your friend.
How deep learning works.
Ruuh is a fictional character, we all know that. But her character is modeled after a young, urban Indian girl who is about 18-24 year old. She seems to be interested in Pop culture and is great at the usage of fluent urban slangs used in India.
The first step in creating Ruuh was to collect data. She was meant to by affable as well as witty. The source for this personality for Ruuh was real-time conversations, Social Media conversations, forums, social platforms and messaging services where the data is collected to improve user experience anonymously.
Next, they had to refine the useful data that they collected. This step took 70% of total data collected as useless and was removed. Microsoft made sure that there are no offensive comments for people in the US, UK and Australia and any sexist or political comments.
Now, this refined and useful data was to be applied in the selected model. This model was the cDSSM or Convolutional Deep Structured Semantic Model. This is a newer model and helps in more better and deeper human-like behavior in AI.
How cDSSM results in better AI
Query Identification is the first step in making AI more like Humans. An algorithm takes the input query and looks in the database for similar questions. This is also referred to as Information Retrieval or IR.
For Example: if the query is, “how do I make chicken pasta?”, Ruuh analyzes the data and finds multiple samples of similar questions.
Here, the algorithm sorts out the responses based on how relevant the samples are. This is how the most relevant data is given as an output.
Now, it might be pointless if the chatbot forgets what the user is talking about.
For Example: Question: “Do you like ice cream, Ruuh?”
Ruuh: “Yes, I like it.”
Question: “which flavors do you like?”
Ruuh: “Chocolate and Vanilla.”
Now, Ruuh knew that the second question was regarding ice creams and hence, the reply was appropriate.
To be so good at her functionality, Ruuh’s algorithm constantly looks up for data in the previous queries from the user and understands the context about what the user is talking about.
Detection and response to emotional cues
Now, more human-like means detection of emotions. This is so because humans have emotional mindsets. So, in order to detect users’ emotions, Ruuh looks up for patterns in chat messages received by her and the type of emojis used in the chat. So, when you are talking to her, she knows if you are happy, sad, excited or upset.
Ruuh is powerful and a great way to show the power of what AI can do today to behave like a human being. With the power of cDSSM, Ruuh is much smarter.
To summarize, the model combined with deep learning integrates context and the user’s message to extract the appropriate response. The model extracts the context from the message, retrieves previous messages, creates a group of appropriate responses, ranks them according to relevance, and generates the final output.
Let’s understand this better with an example. If a user asked Ruuh, “Which pizza toppings are most popular?”, Ruuh would identify the query as about ‘pizza toppings’ and retrieve the most relevant answers based on this query. Ruuh would rank similar answers from the database based on relevance to generate the most appropriate response. With contextual awareness, Ruuh can easily answer follow-on questions such as, “Which ones do you like?” by replying “I love mushroom and pineapple”.
Ruuh is now one year old, and I must say that the future of AI is bright because of this rate at which we are seeing more and more advanced AI emerging, we are about to see smarter things around us very soon. We wish the team at Microsoft, a very best of luck and I hope they will keep surprising us in the future with these great products.