Three problems with Facebook chatbots

April 12, 2016 was the day when Facebook nailed its blue and white colors to the mast and revealed what it thought would be the next development that will revolutionize the technology industry.

Mark Zuckerberg announced at the company’s F8 conference that their new Messenger bots will be designed so that customers ‘should be able to message a business in the same way that you can message a friend, get a quick response and shouldn’t have to take [their] full attention like a phone call would.’

Since the rollout of more than 11,000 bots, the program has been beset by difficulties. Facebook Messenger’s VP admitted a couple of months into the launch that the bots were overhyped and not good enough. In the last month, an investigation by The Information found the bots were failing to answer 70% of requests without human agents, forcing Facebook to scale back on its ambitions, which means Mr. Zuckerberg will have to wait a bit longer before his bots can be viewed by customers at a philia level.

What has gone wrong exactly?

Messenger bots seem to have fallen at the first hurdle – holding genuine conversations with users.

Many of the bots have difficulty understanding even the most basic forms of communication, such as saying “hi” or answering what the weather will be like. When 83% of Millennials polled say bots either need to be more accurate or have more natural sounding conversations, there is a problem.

 

The solution

It is arguable that Messenger bots have been rolled out far too early without fully utilizing some of the core technologies that allow computers to understand natural language.

  1. Natural language processing

    Forrester’s report from December 2016 outlines how successful chatbots depend on natural language processing, artificial intelligence, and machine learning. These technologies are crucial to being able to understand the meaning behind a customer’s inquiry rather than simply the keywords.Users will become easily irritated when they want to know how a product works and instead they are shown who works at the company or other irrelevant information. The result of utilizing these key technologies is not only more correct answers but also a more natural conversation with the customer.

  2. Know that Messenger bots are a work-in-progress

    Since the launch of the Messenger bots last year, there have been few significant improvements in their ability to match users’ questions with the right answers. Many Messenger bots are failing to correctly utilize their knowledge base and are not discovering new content for their bots.
    The answer to this is employing teams who have spent years refining their semantic capabilities as opposed to using fledgling chatbot creators. Many have developed innovations to recognise when certain types of questions are not being met with satisfactory answers. From this, tweaks and adjustments to the chatbot’s content can be made in order to enrich the customer experience further.

  3. Employ computational linguists

    Finally, Messenger bots could improve their matching success rate by employing native speaking computational linguists to further develop the lexicon. A linguist could act as an additional gap analysis by adding necessary words, semantic relations or disambiguation rules to improve the linguistic resources of the chatbot. The result would be a further increase in matching efficiency for the bot.

Walk before you run:

Given the desire to release the newest innovations in technology as quickly as possible, it appears that Facebook Messenger bots have faced (though perhaps not quite as controversially) the same growing pains experienced by the likes of Google Home and Microsoft’s Tay.

Taking these sensible steps will ensure that the next time a customer starts a conversation with a Messenger bot, it will end with them looking forward to their latest purchase rather than banging their head against the keyboard as the bot pleads its ignorance to basic salutations for the hundredth time.

Inbenta utilizes its patented natural language processing and +11 years of research & development to create interactive chatbots with an industry leading +90% self-service rate.

Companies around the world including Ticketmaster UK utilize the InbentaBot to maintain a personal service for their customers while reducing support tickets.

Interested in finding out more? Our team of experts are at your service to design a custom proposal for you.

Let’s get in touch

Share to Twitter
by Inbenta Team