Tekstikaeve rakendamine isiksusetestidel personali värbamise eesmärgil
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Käesolevas töös rakendatakse tekstikaeve meetodeid ettevõttes Psience OÜ välja töötatud isiksusetesti avatud vastuste ehk enesekohaste tekstide analüüsimiseks. Töö eesmärgiks on analüüsida tekste personali värbamise kontekstis, kasutades tõenäosusliku teemade modelleerimise algoritmi Dirichlet’ peitlahutust (LDA), ja võrrelda tekstianalüüsi tulemusi testi numbriliste skooridega. Teemade modelleerimise tulemusel leiti kolm teemat, mille abil oli võimalik vastajaid nende sõnakasutuse alusel kirjeldada. Samuti leiti võrdleva analüüsi põhjal, et teemade modelleerimise edasiarendatud versioon võiks ettevõttel tulevikus aidata muuta tekstianalüüsi automaatsemaks.
This thesis describes how text mining methods have been applied to analyse self-description texts given in the context of personality tests that are used by Estonian HR-company Psience OÜ. The aim of the thesis is to use probabilistic topic modelling algorithm called latent Dirichlet allocation (LDA) to analyse the texts and to compare the results with the scores of the personality tests. As a result of applying topic modelling techniques, three topics where found that allowed to describe the personality test takers by their word use. Also, the comparative analysis showed that refined version of topic modelling approach could help to make the text analysis more automatic in the future as it is today.
This thesis describes how text mining methods have been applied to analyse self-description texts given in the context of personality tests that are used by Estonian HR-company Psience OÜ. The aim of the thesis is to use probabilistic topic modelling algorithm called latent Dirichlet allocation (LDA) to analyse the texts and to compare the results with the scores of the personality tests. As a result of applying topic modelling techniques, three topics where found that allowed to describe the personality test takers by their word use. Also, the comparative analysis showed that refined version of topic modelling approach could help to make the text analysis more automatic in the future as it is today.