Igor Ashmanov and Natalia Kasperskaya
Stanislav Ashmanov
'Nanosemantics' is a Skolkovo foundation resident and the National Technological Initiative of Russia NeuroNet cluster participant
“Elephants were the first topic that we trained our inf to know. Mikhail Volovich wanted to create your ordinary, unsophisticated gal, so that in a conversation, it would be possible to play with her character and communication style. I started with two infs: a hedgehog seller and Petrovich. Petrovich was supposed to be just one of us guys, the one who can talk to you heart-to-heart about the real stuff. The hedgehog seller is still popular with the audience, people engage with him regularly! It is a promoter inf who offers its interlocutor to buy a hedgehog, has a great grip of the conversation focus, always comes back to the purpose of the conversation and actively promotes it.”
“The idea was for people to start creating chatbots that are interesting for them to keep a conversation with. Say you are interested in cultivating guppy fish; then your chatbot will know absolutely everything about guppy including jokes and anecdotes. This is how one person can accumulate their knowledge in quite a tangible form. It is one thing to create text, but we take it to a whole new level by offering to create an informed chatterbox. Further on, if a user has a website or a lifejournal, the chatbot can be incorporated there as well. Meaning that if the robot owner is offline, it is still possible to check out the resource and ask a few questions about guppy fish. From the very beginning we intended that the bots interact not just with people, but with one another. Such conversations would allow them to identify who is more knowledgeable in a certain area of expertise.”
“Linguists see our technology as some sort of a toy. Take ABBYY that works on machine translation. Their experts have been trying to comprehensively analyze the text structure, build its syntactic representation, followed by the semantic representation. They have been working on this for over 20 years, and it is only now that they are ready to showcase some Alpha versions. At the same time, our infs recognize text on the heuristic level, foregoing the in-depth analysis. It is the language that we have developed that allows us to trust various heuristic methods. It accounts for the context without spending tremendous power supplies on its attempts to calculate all possible options and converge or imitate the human possibilities.”
We have developed the first Vietnamese-speaking chatbot. Specially invited expert linguists from Vietnam take part in its development. This virtual consultant can be found on the www.wada.vn search engine. The chatbot that employs such a complex, from the linguistic point of view, language has been created to go through the ordeal by fire and prove that our technology is universal and autonomous as it does not rely on the core technology language of creating virtual assistants.