Towards Using Word Embedding Vector Space for Better Cohort Analysis

Abstract

On websites like Reddit, users join communities where they discuss specific topics which cluster them into possible cohorts. The authors within these cohorts have the opportunity to post more openly under the blanket of anonymity, and such openness provides a more accurate signal on the real issues individuals are facing. Some communities contain discussions about mental health struggles such as depression and suicidal ideation. To better understand and analyse these individuals, we propose to exploit properties of word embeddings that group related concepts close to each other in the embeddings space. For the posts from each topically situated sub-community, we build a word embeddings model and use handcrafted lexicons to identify emotions, values and psycholinguistically relevant concepts. We then extract insights into ways users perceive these concepts by measuring distances between them and references made by users either to themselves, others or other things around them. We show how our proposed approach can extract meaningful signals that go beyond the kinds of analyses performed at the individual word level.

What does a Text Classifier Learn about Morality? An Explainable Method for Cross-Domain Comparison of Moral Rhetoric

“…Lexicons are generated manually (Graham et al, 2009;Schwartz, 2012), via semi-automated methods (Wilson et al, 2018;Araque et al, 2020), or expanding a seed list with NLP techniques (Ponizovskiy et al, 2020;Araque et al, 2022). The lexicons are then used to classify morality using text similarity (Bahgat et al, 2020;Pavan et al, 2020). Moral elements have also been described as knowledge graphs to perform zero-shot classification (Asprino et al, 2022).…”

American Politicians Diverge Systematically, Indian Politicians do so Chaotically: Text Embeddings as a Window into Party Polarization

“…Text-based embedding learning has been previously employed to understand phenomenon on social media (Alam, Joty, and Imran 2018;Chen, McKeever, and Delany 2019;Yan et al 2020;Bahgat, Wilson, and Magdy 2020), including politics on social media (Oliveira et al 2018;Hemphill and Schöpke-Gonzalez 2020). Some of the existing methods in the literature, which aim at embedding social media users based on the written content, concatenate all the posts of a user as a single document and then train a document level embedding model Pan 2017, 2018;Benton, Arora, and Dredze 2016) 1 .…”