CHATBOTS: TECHNICAL REVOLUTION OR OLD HAT?
Chatbots? Never heard of them before? Then it’s high time. These little linguistic helpers not only relieve the workload at companies, but actively support their customers as well. Whether in customer service, on websites, in online shops or social media platforms, chatbots are used almost everywhere, beginning with the same question: “How can I help you?”
We introduce: Eliza (born between 1964 and 1966)
The first chatbot was programmed between 1964 and 1966 by Joseph Weizenbaum. Looking at the year, one first rubs one’s eyes and wonders how this was accomplished with the technology of that time. The chatbot, then named Eliza, was a kind of psychotherapist who listened to the concerns and problems of the test subjects and analyzed their issues.
Nowadays, chatbots are usually used for other purposes, although somehow this still involves the worries, problems and wishes of users. Anyone who was used to picking up the phone or waiting in a long line just to find out how to solve a technical problem with a router, or whether there was a malfunction, or where they could view their bill in the online customer portal, can now simply ask such questions via chat and get answers in no time at all.
But what exactly are these chatbots, and how do they work?
Chatbots can be compared to language assistants like Siri, Cortana or Alexa. The difference is that you don’t talk to a chatbot – you actually chat. Chatbots are text-based dialogue systems with natural language capabilities. However, in order for the bots to be able to answer your question in a targeted way, they must first be fed a lot of information. This information is stored in a knowledge database. The user asks a question and the bot delivers the answer because it understands the intention behind the user’s question and can therefore search for the corresponding entry in the knowledge database.
We talked to specialists in this field, the BOTfriends from Würzburg, to satisfy our curiosity about how chatbots could respond to users’ questions and wishes. In fact, the current best practice systems include intents . This allows the bot to recognize a user’s intention from a series of question variations, which always target the same answer, and to react accordingly. Sounds complicated? Here is a small example:
“How will the weather be tomorrow?”
“Do I need an umbrella tomorrow?”
“Will there be bathing weather in Würzburg tomorrow?”
“Temperatures in Würzburg tomorrow will be between 18 and 25 degrees, mostly sunshine.”
After it is received, the question is assigned to one or more existing intentions and does not even have to be stored in an exact form in the knowledge database. With the help of probabilities, the bot recognizes the intention behind the question and can answer accordingly. “Quite often, the chatbot even understands entire sentences better than individual keywords,” says Daniel Rösch, CEO of BOTfriends in Würzburg. “But in order for the chatbot to be able to answer your questions correctly, you should start with at least 20 question variations per intention,” Daniel continues.
What about international companies? Chatbots and multilingualism – how do they work?
As a language service provider for EnssnerZeitgeist Translations, we are particularly interested in the translatability of chatbots. Machine translation is a good way to develop multilingual chatbots. Nevertheless, machine translations that are not proofread run the risk of not matching, thus affecting the company’s brand. In addition, machine translation systems may not be able to recognize simple phrases or routine formulas 100% and may not correctly reproduce them in the foreign language. The system cannot correctly understand the question or intention of the user and thus provides the wrong answer. As a language service provider from Schweinfurt, Germany, we are of course committed to correct translations: With the help of the many translations already in our glossaries and TMs, such problems could sooner or later disappear.
This all sounds as if chatbots can’t help with or be used for more complex topics at all. Is that so?
Far from it! Bots are also suitable for more complex topics such as technical documentation. The BOTfriends already have experience developing this kind of complex application. The quality of the knowledge base is extremely important. At this point, we come back into play as a language service provider. We have an immense database of glossaries and language-specific features from our customers that could be used to develop complex bots.
Chatbots can do much more than just answer the question “How will the weather be today?” Complex topics and questions can be processed just like translation and transferred into other languages. The right content in the knowledge database and its translation are the be-all and end-all. Chatbots are definitely not old hat, but advancing more and more, and will revolutionize communication between companies and their customers in the medium and long term. We at EnssnerZeitgeist Translations have also recognized the potential of multilingual chatbots and are actively engaged in discussions with our customers.