The seminar "Mathematical Economics"

On Tuesday, March 30, the National Research University Higher School of Economics hosted the seminar "Mathematical Economics".

Theme: «Method of taxonomic content-analysis of thematic text collections»
Speaker: B.G. Mirkin (FKN, MCAIR NRU HSE)

Seminar leaders:
Doctor of Physical and Mathematical Sciences V.I.Danilov
Academician V.M. Polterovich

Annotation :
The initial data for the method is a collection of texts related to a certain subject area and the taxonomy of this root tree of area concepts (the closer to the root, the more common). The leaf concepts of taxonomy are the elementary units of meaning. The method includes three stages:

(1) construction of a matrix of relevance scores "texts - sheet concepts";
(2) the formation of fuzzy clusters of leaf topics (so to speak, relevant to the same texts);
(3) better lifting of fuzzy clusters in taxonomy into so-called "head topics".

The quality of the interpretation of the lifting results determines the degree of success of the method application. All three stages are carried out by their own methods, so to speak, by their own efforts. Each serious approbation took place in a fairly environment (methods of text analysis and methods of fuzzy cluster analysis).

Now the criterion for the optimality of stage (3) is the maximum saving of the total penalty for the introduction of new elements of meaning: "head concepts", "gaps" and "outliers". However, we managed to fulfill the criterion and method of the likelihood problem for this task - we are currently implementing it.

The method was applied to the analysis of two collections of scientific articles in the field of data, as well as to collections of: (a) reviews of restaurants and cafes in Moscow. Moscow, (b) cars sold on the Internet, (c) all articles published in the magazine "Classification Journal" (Springer) in 1984-2018, etc.

S. Nascimento (Lisbon, Portugal), T. Fenner (London, Great Britain), D. Frolov (FKN and MCAIR NRU HSE), as well as students of the FKN NRU HSE A. Vlasov, A. Ushakova, D. Babin, A. Guzharina, A. Sitnikov, A. Denisenko, J. Hayrapetyan. This work was supported by the grant for the implementation of the research project "Development of methods for structuring and conceptualizing textual data based on the taxonomy of the subject area" No. 19-04-019 in the competition of research projects of research groups of the HSE Science Foundation (HSE) Program in 2019-2020 biennium