We are delighted to announce a recent chapter from our research team, which was a contribution to the monography entitled “Innovations, Trends, and Socio-Economic Transformations – Interdisciplinary Studies” (original title in Polish: Innowacje, trendy i transformacje społeczno-gospodarcze – badania interdyscyplinarne).
In this monography, our team contributed a chapter entitled: “Comparative Analysis of Classical Questionnaires and Metrics Obtained from Social Media: A Case Study of the War in Ukraine” (original title in Polish: Analiza porównawcza klasycznych kwestionariuszy i metryk pozyskanych z mediów społecznościowych na przykładzie wojny w Ukrainie). Authors of this chapter are: Hubert Plisiecki, Maria Flakus, Piotr Koc, & Artur Pokropek. Congratulations to all Authors for contributing to this interdisciplinary research!
Below, we provide abstract and keywords for this publication. Also, the full text of the chapter (in Polish) is provided at the end of this note.
Abstract: This study enriches scientific literature by comparing classic questionnaires (KK) and metrics extracted from social media (DMS). It compares trends in a 9-item survey of fear of war in the context of Russia’s attack on Ukraine, with metrics extracted from Polish Twitter. The Twitter data was collected between March and August 2022. During this time, questionnaire data was collected in uneven segments. The latent semantic scaling (LSS) method, and affective word norms, in particular valence and arousal norms were used as metrics. LSS involves creating a lexicon of words related to and opposed to a given concept using the word embeddings technique and calculating the metric of a given concept in each text by counting the words occurring in it. The affective word norms, in turn, are ready-made lexicons of words, evaluated in terms of various components of emotion. Similar to the LSS, the score for a given text is calculated by averaging the ratings of the words occurring in it. The analysis showed significant differences between the two data sources. The study was preregistered, which makes the results valuable despite their low statistical power.
Keywords: NLP, SMD, computational social science, sentiment analysis, metrics validation