Businesses today are interested in developing bots that are able not only to understand a person but also to communicate with him on the same level, which will reduce the routine in working hours of call-center employees, doctors, consultants of various kinds. To understand if this can be done in reality or still an area of fiction stories, let's take a look at the possibilities of top chatbots nowadays.
Much effort expended in the development of artificial intelligence and machine learning of neural networks while creating a high-quality chatbot. At the same time, the core on which the “smart” chatbot is built is also important.
- NLP or Natural Language Processing is a general term that describes the ability of a machine to process what was said, understand its meaning, determine the necessary response action and respond in a language that the user understands. NLP – the process of turning a computer text into structured data.
- NLU or Natural Language Understanding is a subset of NLP processes, which is the basis of any chatbot. The process has a much narrower, but extremely significant spectrum of action. It means the computer's ability to choose the best way to process unstructured input data and convert them into a structure understandable for the machine. Such data processing is crucial for such unpredictable data as abbreviations, colloquial speech, words written with errors that a person understands, but a machine does not.
- NLG or Natural Language Generation is the process of turning structured data into text. In essence, this is the process when bot creates a text to communicate with a person in his language.
Is There Any Difference in How Robots and People Use the Language?
Human speech is directly related to his thinking. To understand whether a chatbot can communicate like a person, you must first answer the question: “Сan chatbot think like an ordinary person”?
How different are computer and human thinking? Let us explain with a simple example: your wife asked to buy bread. Of course, people are different, but suppose that you are the kind of person who loves accuracy in everything. Then the algorithm of your actions will be as follows:
- First, you ask her what kind of bread is needed. Let’s suppose she replies: “I want a buckwheat baguette”;
- Then you specify how many baguettes she needs: a loaf, half a loaf, a quartet? You find out that your wife wants one loaf;
- Then you check the places you know where you can find the right one. You know that ordinary baguettes are in many stores, but the buckwheat can be found only in a few specific supermarkets;
- You find a baguette in two stores near your home. In the first store, it costs 0.5 USD in the second 0,75 USD You ask your wife which one to choose;
- Having received the answer about price, you buy the right option. Moreover, if irritation was heard in your wife’s voice, then in addition to the baguette, you take also a chocolate bar - for better effect.
Chatbot performs the same process. In fact, the main difference between chatbots and humans is the lack of intuitive understanding. Chatbots are not able to read between the lines and understand the shades of mood. So if you bought a bot via chat, your wife would hardly get a chocolate bar. However, she may get an offer to buy a product that fits the best to what she chose.
Chatbots with Artificial Intelligence Those Communicate Humanlike
Here is how chatting with artificial intelligence looked like back in 2011.
However, today, progress breakthrough in science and technology has gone much further, and chatbots are capable of much more “human" communication. A vivid example of this is the voice chatbot Mitsuku, four-time winner of the Löbner prize for passing the Turing test in 2013, 2016, 2017 and 2018.
In the next video, you can see a comparative analysis of the answers of Mitsuku and another well-known chatbot, created by Amazon, named Alexa for the same questions. Also, the difference between them is really impressive, although it still does not hold out to real human communication.
Online Psychotherapist Bot – Woebot
Particular attention should be given to this chatbot, since the sphere of human feelings and emotions is complex and subtle and, as it is believed, not yet cogitable to chatbots. In 2016, Alison Darcy, the creator, and CEO of Woebot, created a robot to help people in depression. Being a Psychologist at Stanford University, she knew that cognitive-behavioral therapy is a powerful tool to help people suffering from depression or increased anxiety. This is a form of treatment, whilst the therapist teaches the patient to use simple techniques to help overcome negative patterns of thinking.
It is easy to learn, but the treatment is more effective if regularly shown to the doctor, which is not available to many for a busy schedule, financial, or other reasons. Darcy knew that two-thirds of people in the United States alone never visited a specialist, not to mention the rest of the world. In some countries, people do not have access to primary medicine to say nothing of supporting mental health.
The chatbot was created in cooperation with a team of psychologists and Andrew Ng, an artificial intelligence pioneer, and programmed for a variety of conversation options in the style of behavioral therapy. The chatbot helps users more accurately describe their states and overcome distorted thinking or panic attacks using simple messages.
Such a bot therapist, available 24 hours a day, can help many people around the globe. Сhatbot’s avatar is not human, but a robot picture, what was made on purpose – to make it easier for users to trust him. However, his personality, the ability to use suitable Emoji and GIFs while conversation, makes communication with him very similar to chatting with a person.
In the first week after Woebot launch, 50 thousand people took advantage of it. This amount is more than any psychotherapist can take in a lifetime. Darcy believes that very soon, artificial intelligence will be able to recognize the feelings of people better than ourselves.
Chatbot Friend in Insomnia
Let us examine one more small example. We all know the feeling of not being able to fall asleep. That is a time when we are actively looking for friends online to chat. The announced problem solution is the chatbot Insomno Bot 3000. If you are in insomnia, you can write to a specific number – and a pleasant, caring, tireless dialog partner is provided.
Alice Virtual Dialog Partner
ALICE (Artificial Linguistic Internet Computer Entity) – virtual chatbot, for several years running included to the world top-3 chatbot, is able to carry on a conversation in almost natural human language. The system matches the user's phrases with samples in its knowledge database using the heuristic method.
So Is It Possible to Replace Employees with Chatbots?
As we see bots already and to a great extent can communicate the way people would do. However, the main question is if you can replace the seller in your store or any other employee with a bot?
Unfortunately, for now, this is not possible. Even the latest developments fall short compared to people’s abilities in thinking and speaking. Chatbots are still not able to understand slang, unintelligible speech, abbreviations, modified words, metaphors. They do not respond to hints and still do not recognize the tone of the conversation and the mood of the person they are talking to.
Benefits of chatbots:
- bots are available 24/7, without sick leaves and distemper
- instantly respond to the client's request, without making him wait
- find out what customer wants via standard questions and recommends precisely the product that is needed
- provide links to direct purchase of the desired product
- search for answers to questions in the database using natural speech recognition algorithms (NLU)
- if necessary, transfer the client to a live manager
Disadvantages of chatbots:
- can’t respond to non-standard requests
- don’t consult or give pieces of advice in the full sense of these words (“Will this fabric get wet?”, “Is it suitable for a 3-year-old child?”, “Please, advise something for a grandmother”)
- don’t resolve issues in non-integrated systems automatically (issue a paper bill, call another department, etc.)
Therefore, for now it is possible to use chatbot as more auxiliary tools to shorten the sales process, improve support processes and reduce the number of employees, but not yet as an alternative to a person.To understand whether you need a chatbot, answer the question: which category of goods do you sell? If it is an inexpensive commodity, customers easily accept the bot consultant. If you sell exclusive, expensive goods or services, they certainly want an individual approach, which so far can only provide people.
Do you think your business needs a chatbot right now? Write to us, we can create one or more for your needs.