Analysis of topics discussed in online employee reviews
Author
Jeppesen, Markus Hegermann
Term
4. term
Education
Publication year
2022
Submitted on
2022-05-30
Pages
11
Abstract
The growing availability of public data on social platforms offers new ways to understand what is discussed at work. In this study, we analyze 40,000 employee reviews on Glassdoor from 20 large technology companies to see what topics employees talk about. We first use Latent Dirichlet Allocation (LDA) topic modeling—a statistical method that groups words that often occur together—to identify nine recurring corporate topics. We then examine how these topics appear in each company using correspondence analysis (to map relationships between companies and topics) and sentiment analysis (to assess whether the tone is positive, negative, or neutral). The results show what kinds of insights employee reviews can provide and how companies can use them to better understand working life and employee satisfaction.
Den stigende mængde offentligt tilgængelige data på sociale platforme giver nye muligheder for at forstå, hvad der bliver diskuteret på arbejdspladser. I dette studie analyserer vi 40.000 medarbejderanmeldelser på Glassdoor fra 20 store teknologivirksomheder for at få indblik i, hvilke emner ansatte taler om. Først anvender vi LDA-emnemodellering (Latent Dirichlet Allocation) – en statistisk metode, der grupperer ord, som ofte forekommer sammen – til at identificere ni gennemgående virksomhedstemaer. Dernæst undersøger vi, hvordan disse temaer viser sig i den enkelte virksomhed ved hjælp af korrespondanceanalyse (kortlægger relationer mellem virksomheder og temaer) og sentimentanalyse (vurderer om tonen er positiv, negativ eller neutral). Resultaterne viser, hvilken viden der kan udledes af medarbejderanmeldelser, og hvordan virksomheder kan bruge den til bedre at forstå arbejdsliv og medarbejdertilfredshed.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords