Research indicates that how individuals utilise language to express themselves reflects individual-level differences regarding psychosocial characteristics, including perceived Quality of Life (QoL). In this study, we apply a language modelling technique to the natural user-generated language from Facebook to examine associations between language expressed on Facebook and self-reported QoL. Specifically, we collected the user-generated language from a sample of 603 Facebook users (76.3% females), mined emerging text corpora using the LIWC closed-vocabulary approach, and examined associations between LIWC features and self-reported domain-specific QoL (Physical, Psychological, Social), and General QoL. In line with previous research, we found use of pronouns, negative emotions, death and sleep words, and use of profanity to be significantly associated with QoL. Next, we used the Random Forest algorithm to test the predictability of QoL dimensions based on LIWC features and posting activity statistics. The models achieved moderate predictive power (r ranging from.22 to.33), the Psychological and General QoL dimensions showing the highest accuracy. An alternative approach combining LIWC features, posting activity, and predicted scores for domain-specific QoL components showed increased accuracy when predicting General QoL (r =.43). Findings are discussed in light of previous literature. Suggestions for improving models in future studies are provided.

Mining Facebook data for Quality of Life assessment

Marengo D.;Longobardi C.;Settanni M.
2021-01-01

Abstract

Research indicates that how individuals utilise language to express themselves reflects individual-level differences regarding psychosocial characteristics, including perceived Quality of Life (QoL). In this study, we apply a language modelling technique to the natural user-generated language from Facebook to examine associations between language expressed on Facebook and self-reported QoL. Specifically, we collected the user-generated language from a sample of 603 Facebook users (76.3% females), mined emerging text corpora using the LIWC closed-vocabulary approach, and examined associations between LIWC features and self-reported domain-specific QoL (Physical, Psychological, Social), and General QoL. In line with previous research, we found use of pronouns, negative emotions, death and sleep words, and use of profanity to be significantly associated with QoL. Next, we used the Random Forest algorithm to test the predictability of QoL dimensions based on LIWC features and posting activity statistics. The models achieved moderate predictive power (r ranging from.22 to.33), the Psychological and General QoL dimensions showing the highest accuracy. An alternative approach combining LIWC features, posting activity, and predicted scores for domain-specific QoL components showed increased accuracy when predicting General QoL (r =.43). Findings are discussed in light of previous literature. Suggestions for improving models in future studies are provided.
2021
40
6
597
607
data mining; digital footprints; Facebook; LIWC; Quality of Life; text-analysis
Marengo D.; Azucar D.; Longobardi C.; Settanni M.
File in questo prodotto:
File Dimensione Formato  
Mining Facebook data for Quality of Life assessment.pdf

Accesso riservato

Tipo di file: PDF EDITORIALE
Dimensione 1.4 MB
Formato Adobe PDF
1.4 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2318/1753944
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 12
  • ???jsp.display-item.citation.isi??? 9
social impact