Figure 5 - Securitization of Disinformation in NATO’s Lexicon: A Computational Text Analysis

Topic count graph demonstrating the optimization rationale for our stm algorithm's choice of 50 topic models. The 'K-value' shows the optimum number of 'structural topic models' the algorithm has to go through the text to find the optimum semantic coherence. In other words, the K number designates the optimum number of structural topic models in texts that have the highest statistical coherence coefficients. Often, K values are assigned by the programmer and an optimum number gets eyeballed after several trial and error runs. K-value optimization uses machine learning to iterate through the text multiple times to find the optimum K-value by statistical clustering of frequently collocated word combinations.

Go to figure page

Reference

Securitization of Disinformation in NATO’s Lexicon: A Computational Text Analysis - Scientific Figure on ResearchGate. Available from: https://www.researchgate.net/figure/Topic-count-graph-demonstrating-the-optimization-rationale-for-our-stm-algorithms-choice_fig4_362348156 [accessed 8 Dec, 2023]

Caption

Figure 5: Topic count graph demonstrating the optimization rationale for our stm algorithm's choice of 50 topic models. The 'K-value' shows the optimum number of 'structural topic models' the algorithm has to go through the text to find the optimum semantic coherence. In other words, the K number designates the optimum number of structural topic models in texts that have the highest statistical coherence coefficients. Often, K values are assigned by the programmer and an optimum number gets eyeballed after several trial and error runs. K-value optimization uses machine learning to iterate through the text multiple times to find the optimum K-value by statistical clustering of frequently collocated word combinations.

Embed code

<a href="https://www.researchgate.net/figure/Topic-count-graph-demonstrating-the-optimization-rationale-for-our-stm-algorithms-choice_fig4_362348156"><img src="https://www.researchgate.net/profile/Hamid-Akin-Uenver/publication/362348156/figure/fig4/AS:1183447863312384@1659167366023/Topic-count-graph-demonstrating-the-optimization-rationale-for-our-stm-algorithms-choice.png" alt="Topic count graph demonstrating the optimization rationale for our stm algorithm's choice of 50 topic models. The 'K-value' shows the optimum number of 'structural topic models' the algorithm has to go through the text to find the optimum semantic coherence. In other words, the K number designates the optimum number of structural topic models in texts that have the highest statistical coherence coefficients. Often, K values are assigned by the programmer and an optimum number gets eyeballed after several trial and error runs. K-value optimization uses machine learning to iterate through the text multiple times to find the optimum K-value by statistical clustering of frequently collocated word combinations."/></a>