Chatbots Get Smarter


Improving computing capabilities and the continued maturing of artificial intelligence and machine learning has led to a renaissance in chatbot capabilities since their introduction in the mid-1990s.

More firms have started introducing sentiment analysis and recommendation capabilities via their chatbot conversational interfaces, Usama Noman, founder an CEO of chatbot-provider Botsify, told IntelAlley.

When users search for an item from a company’s e-comm store and engage the chatbot, the chatbot can recommend products. If users open details regarding the recommended product or put it in their carts, the chatbot registers the preferences and use them to build future product recommendations.

However, the most significant area of growth Noman has seen within the chatbot field is the development of multi-lingual support.

Over Botsify’s three-year history, the company has witnessed tremendous advancement in the technical dictionaries used to develop and operate chatbots in many languages.

English, French, Spanish, and German dictionaries lend themselves to natural language processing due to the large datasets that are available in their respective languages, according to Noman.

“When you talk about languages like Arabic, Chinese, or Urdu, there has been very little work that has been done in natural language processing in these languages,” he said. “It is tough for anyone to answer using your chatbot in these languages automatically.”
Firms like Botsify often use an intermediate translation step to support languages that do not have datasets large enough to use natural language processing natively.

“When our users come in to ask a question in a language in which you have not trained your chatbot, we convert the question into English or whatever language the chatbot was trained using the Google Talk API,” said Noman. “We then find out the right answer for the customer, translate it back to the original language in which the users asked the question, and reply in that language.”

He noted the process is not as polished as it could be and that there are more accurate translation engines, but they only support only 20 to 30 languages compared to the more than 200 languages Google Translate supports.

Besides broadening their support for additional written languages, chatbots are also increasing support for voice-based transactions via Amazon’s Alexa and Google’s Google Home platforms.

“These conversational interfaces will be everyday means of consumer interaction,” he said. “It eventually will be in every room in every house, just like mobile. With an average family of five, there are seven mobile phones in particular homes. They will have a similar number of voice interfaces in the next three or four years.”

Botsify already supports the Amazon and Google platforms but does not support Apple’s Siri platform.

“There has been a lack of demand by our clients, so we haven’t looked at it,” said Noman.