Chatbots: Reach Should Exceed Grasp
Society has been inundated with promises of revolutionary artificial intelligence that can think, act and respond just like a human. Despite the fanfare, sleek marketing campaigns and stylish launches, most of these claims have fallen flat. Not only were these chatbots not as effective as advertised, they created additional problems and led to frustration and disappointment from businesses that invested in them and customers who interacted with them.
Writing in Forbes magazine, Christopher Elliott presents the findings of CGS, a company which specializes in business applications, learning and outsourcing services. The research carried out by CGS, which forms part of its annual Global Consumer Service Report, clearly demonstrates the limitations of the current crop of chatbots on offer. According to this report, ‘…about half of all respondents said they would turn to a chatbot for a quick customer service need. Another 25 percent they would reach out to a company vie email or social media.” Overall, the results indicate lacklustre enthusiasm for chatbot interaction as “50 percent of U.K. respondents and around 40 percent of U.S. respondents said they’d prefer a person.” To go by Elliott’s reporting, the chatbot is a flash in a pan which has had its fifteen minutes of fame, but now faces a bleak future. Clearly, if chatbots are going to contribute positively to customer engagement and business operations, something must change.
New Development lead to New Horizons
Fortunately, the IT industry is nothing if not adaptable, and this innovative and dynamic world can change with the click of a mouse. Indeed, ground-breaking advancements in machine learning and Natural Language Processing mean that today’s conversational AI can achieve levels of comprehension that had previously only existed in the realms of science fiction.
Essentially, machine learning involves presenting the chatbot with enough information so that it can recognise a pattern. Writing in Towards Data Science, Shivam Kollur uses an analogy of math students. Kollur explains that “the general framework of teaching math is giving students many practice problems along with the answers…Each practice problem encodes pieces of information (kind of like input features) that a student (machine learning algorithm) observes alongside the answer (label). We can call this learning process the training of an algorithm.” Eventually, Kollur adds, “after tons of practice problems, our hypothetical student is expected to have been able to find some sort of pattern to utilize in order to solve the problem.” This is a rather accurate portrayal of how machine learning works.
Teamwork is our winning formula
Close collaboration between back-end developers, content specialists and the chatbot itself is vital, as each party plays a crucial role in the training of the chatbot. First, developers program the algorithms which are the foundation of the bot’s training. Subsequently, these algorithms inform the chatbot which information it needs to pay attention to. Simultaneously, content specialists provide the language that corresponds to the algorithms and reinforces knowledge that chatbot has gained. Ongoing training of the chatbot means that it continually increases its understanding and can intuitively decipher new language and correctly respond to written communication.
Natural Language Processing aims to teach the chatbot the rules of human languages, which are obviously not only difficult, but sometimes abstract. The two main areas NLP focuses on are syntax and semantics, which include parsing, or grammatical analysis, and tagging, which focuses on identifying individual key words in a sentence. By slightly altering language while retaining the overall essence of a sentence, content writers can aid in the chatbot’s education.
Employing machine learning and NLP delivers previously unheard-of capabilities and limitless possibilities. Combined with our vision, skills and knowledge, our multi-talented team has produced a forward-thinking solution to an urgent problem, with the potential to re-shape the way financial institutions manage customer service.