If someone needs specific information, voice assistants have a decisive advantage over other information sources: I can ask a question directly with my words. This is as simple as it is ingenious, because no one wants to learn the structure or concept of an operating manual for a knowledge management system, a catalogue or any other source of information in order to find the answers to one or two questions.
How does a voice assistant work?
Even though the tools are always slightly different, voice assistants work on a similar principle:
The user starts the voice wizard and says for example "Hello" or asks a question. Speech recognition now attempts to transcribe the statement or question, which can be difficult depending on the ambient sounds, dialect, etc. By the way, this is the first obvious demarcation to a chatbot that is allowed to work directly with the written language (which most voice assistants can do). Once the statement has finally been translated so that the artificial intelligence can "read" it, the voice assistant tries to find out what the user wants. This is also called Natural Language Processing (NLP) or Natural Language Understanding.
If in the next step the intention of the user is known – one also speaks of an Intent (short for "intention", i.e: "Intention" or "Plan") –, an appropriate reaction is sought. In our example, we might have taught the voice assistant that "Hello" is a greeting. Now, we may have told the voice assistant that in case of a greeting, it should first find out the name of the user. Since the user has not yet given his or her name, the voice assistant asks: "Hello, what's your name?"
If the user replies with "Christopher", the voice assistant now knows the name and greets the user with "Hi Christopher, nice to meet you" (see Figure 1). In the answer of the voice assistant, the second obvious demarcation to a chatbot is shown: The text is spoken, thanks to a text-to-speech algorithm and not just displayed on a screen.
The way to your own voice assistant
The paths to your own voice assistant can look very different. In order to outline a possible way forward, we have developed the following three phases, which can be used to design voice assistants for individual use cases. The use cases are crucial and the technology plays a downstream role.
Design the Bot
Voice assistants are strong when they have a specific area of responsibility and should not have an answer for everything. So before the voice assistant is given life, it must first be designed. To do this, we have to find your exact key use cases. Maybe it already helps that the voice assistant can name the correct E-plan or the correct parameters. But perhaps it should also lead through complex decisions, like a fault tree or when a product is configured.
Build the Bot
In accordance with your requirements and key use cases, we jointly select the appropriate software for the creation and publication of the voice assistant. No matter which way you choose, it will always be about creating dialogues between voice assistants and users. Whether the voice assistant then knows the answer directly or addresses another data source depends on your individual concept.
Beat the Bot
There is a question in the room to start the voice wizard: "What do users really ask?" An intensive learning phase should be planned before (and during) the actual start. In this way, the voice assistant can be trained for the "real" language of the users.
We follow your lead and are happy to start by working with you to identify the key use cases in your company. Depending on your requirements, we can also advise you on designing and setting up a suitable production process. We are also happy to take care of the content for you, create it directly or design interfaces to other source systems.