Add TechGraph on Google
Chatbots may have become ubiquitous in recent times, assisting businesses in everything from bill payments to customer services, but these automated algorithms still struggle to grasp the nuances of human language and context of conversations.
In order to improve user-experience, researchers at IBM India are using artificial intelligence (AI) to make chatbots more emphatic and personalised so that they can meet a client’s specific needs and provide business value at a much higher scale.
According to an estimate, chatbots are predicted to tackle a massive 85 per cent of customer service interactions by 2020.
India is a key player in the chatbot market today where such services have been introduced by many banking and insurance companies, said Gargi B Dasgupta, Director, IBM Research-India and CTO, IBM India/South Asia.
Banks and insurance providers, she explained, were the early adopters of chatbots in the country, helping customers for bill payments, mobile recharges, booking travel, so on and so forth.
However, when it comes to making people feel comfortable, AI has a lot to learn, Dasgupta noted.
A lot of available systems make building the first chatbot easy, but do not scale-up when you increase the scope of the chatbot or need to build several hundred more, she said.
This, she said, is because of two main factors: overt dependence on the manual effort for creating chatbots which makes them both time consuming and brittle, and lack of an end-to-end fall-back plan with continuous learning and improvement.
“We at IBM Research are working towards training these applications with data by infusing AI and related technologies to help organisations get business value at a much higher scale,” Dasgupta told .
The tech giant’s IBM Watson Assistant can be trained to represent a company’s specific brand voice and values using that firm’s customer and business data.
“Conversational agents are going to be a primary way people communicate and the AI that powers it needs to be very strong,” Dasgupta said.
“We work with several enterprise customers to enable their conversation agents to be more usable and affordable,” she said.
“There are a couple of AI projects we at IBM Research India are doing on conversational AI,” Dasgupta said.
The technology, she noted, empowers more realistic conversational AI which is able to handle nuances of the user language and that can handle domain and context understanding.
The enterprise chatbot evolution in India started with everyone being impressed by the fact that they can have a 24/7 channel of communication with their customers.
In addition, they can delight their customers with an instantaneous response to their queries.
Further, many are looking at offering more services like register a claim, get a quote and help customers understand their various policies better.
“Conversation is the most natural way for humans to interact with an AI system and the predictions are by 2024 this will be double-digit billion markets,” said Sachindra Joshi, Manager – Conversational Platforms, IBM Research-India.
“Since human conversation is complex, contains forth and back references, is contextual, it poses a very interesting research problem to solve,” Joshi told .
“We have built chatbots using the research technologies for multiple external and internal clients. They have been running successfully now for 1-2 years,” he said.
The researchers use deep learning techniques on conversations between end-users and human agents that provide support in traditional call centres.
Their analysis helps identify dominant problems or reasons for which end-users contact call centres as well gives insights on how experts agents handle those problems.
“These inputs are extremely valuable to build a chatbot. They reduce the manual effort and the initial build time as well as provide concrete ways to continuously improve the chatbot as it is being used,” Dasgupta said.
“We have focused on continuous learning of conversation agents, scalable bootstrapping from documents and logs and innovative and expressive ways of dialogue design,” she said.
“As these technologies mature, users of chatbots will see an improved experience with the bot being able to learn from past conversations and get more effective.
“This should increase the proportion of successful conversations. Also bot designers will be able to use our tools to create conversation agents faster and make them more effective,” she said.
Joshi noted that one big challenge for chatbots in India is the support for local languages.
Ninety per cent of the new internet users are expected to be non-English speakers in the next few years.
Researchers, he said, are trying to work with linguists alongside technologists developing verticalised vernacular models for handling local languages.
“We have also developed deep learning techniques that help in creating chatbots in a data driven manner and have brought down the initial build time of chatbots from several weeks to only a few days,” noted Joshi.
Researchers said while designing chatbots, it is crucial to keep in mind the affordability of creating chatbots and their usability.
Chatbots are only useful when they help a particular business function to improve, they said.