The variational principle for weights characterizing the relevance

TL;DR

This paper introduces a variational principle-based method for constructing optimal word weights to assess text relevance, resulting in sparse key word sets and improved accuracy in thematic classification.

Contribution

It presents a novel variational approach that generates sparse, optimal weights for text relevance, outperforming traditional frequency-based methods.

Findings

Optimal weights have small support, usually under 10% of words.

The method effectively identifies key thematic words.

High efficiency and performance in relevance determination.

Abstract

The classical method of the thematic classification of texts is based on using the frequency weight on the list of words occurring in texts from the text corpus that determines the theme. In this method , the weight of each word is defined as its normalized frequency in the texts of the corpus. The frequency weight is applied for determining the relevance of the tested text to the theme given via a text corpus: the relevance of the tested text is defined as the value of its frequency weight (see [1]-[3]). In the present work we propose a method of constructing some optimal weights generated via certain variational principles leading to LP (linear programming)problems. A noteworthy feature of those optimal weights is a relatively small number of words belonging to their supports: in all the examples we considered that number did not exceed 10 percent of the quantity of different words (lemmas) occurring in texts of the corpus, and besides the majority of those words were apparent candidates to be selected as key words of the corpus. The application of our method to the determination of the relevance of texts to a given collection of thematic text corpuses demonstrated its high efficiency and performance.