Mivar system of text conversion translates text into a universal semantic graph.
At first, we mean a transition from attempts to understand the meaning within one sentence to complete meaning understanding at levels “paragraph”, “text”, “chapter”, “book”, since only by defining the context different cases of polysemy and ambiguity can be eliminated.
The system consists of modules performing syntactic analysis, semantic conversion, graph building, context ambiguity elimination. Thus, there is gradual comprehension of the text performed by the system and the complete natural language understanding is achieved by means of artificial intelligence.
Operation description
The main limitation of artificial intelligence used to be the inability to work with context and take into account polysemy of natural language. Artificial intelligence is not always capable of understanding and considering language environment, communicative situation, existence of different meanings of one word, situational word usage.All these factors put serious obstacles to recognition of the meaning conveyed in the text by the author. Application of mivar technologies allowed us to overcome principal limitations of natural Russian language understanding.
Qualitative transition was implemented due to multi-dimensional knowledge representation in the form of semantic mivar graph VSO (“Thing-Property-Relation”) and building of a knowledge base consisting of concepts and their definitions, system relations and contexts.
The tasks it solves
Mivar system of text conversion allows us to make texts of any complexity level and subject matter as well as spoken language maximum comprehensible for program complexes and accessible for further processing.
The functions of the system:
- building models on the basis of downloaded texts;
- managing models (editing, combining, adding, deleting);
- exploitation – consulting users about any subject area;
- defining and recognizing the text.
Practical application
On the basis of mivar technologies specialists of the company have developed the platform Tel!Mi. It is designed for creating several classes of products – “Virtual consultants” (for enterprises and projects) and “Personal assistants” (for sites as well as web-portals combining different services).
"Virtual consultant"
“Virtual consultant” is an intelligent assistant designed for semantic search of answers to the questions in any subject area. The main advantage of the product is the capability to accumulate knowledge. This characteristic feature allows us to teach intelligent assistant evolutionary.
"Personal assistant"
“Personal assistant” is an intelligent assistant designed for the users of websites and large web-portals. As well as “Virtual Consultant” it is implemented on the website in the form of pop-up window, in which the user can ask questions concerning web-site information, website usage, necessary information search and services provided. In addition, it provides information about solving problems that arise when working with the website.
The main functions of “Virtual Consultant” and “Personal Assistant”:
- it assists in searching for information on the website;
- it assists in solving problems associated with the website usage;
- it guides users through the website;
- it answers the questions on subject area.