Capturing Twitter Negativity Pre- vs. Mid-COVID-19 Pandemic: An LDA Application on London Public Transport System

Abstract

The coronavirus pandemic has affected everyday life to a significant degree. The transport sector is no exception, with mobility restrictions and social distancing affecting the operation of transport systems. This research attempts to examine the effect of the pandemic on the users of the public transport system of London through analyzing tweets before (2019) and during (2020) the outbreak. For the needs of the research, we initially assess the sentiment expressed by users using the SentiStrength tool. In total, almost 250,000 tweets were collected and analyzed, equally distributed between the two years. Afterward, by examining the word clouds of the tweets expressing negative sentiment and by applying the latent Dirichlet allocation method, we investigate the most prevalent topics in both analysis periods. Results indicate an increase in negative sentiment on dates when stricter restrictions against the pandemic were imposed. Furthermore, topic analysis results highlight that although users focused on the operational conditions of the public transport network during the pre-pandemic period, they tend to refer more to the effect of the pandemic on public transport during the outbreak. Additionally, according to correlations between ridership data and the frequency of pandemic-related terms, we found that during 2020, public transport demand was decreased while tweets with negative sentiment were being increased at the same time.

Sentiment Analysis of Insomnia-Related Tweets via a Combination of Transformers Using Dempster-Shafer Theory: Pre– and Peri–COVID-19 Pandemic Retrospective Study

“…Our findings on the significant increase in the number of negative tweets (P<001) during the pandemic is consistent with previously published literature [46]. Politis et al [47] showed an increase in negative sentiment on certain dates by analyzing tweets that were posted before and after the outbreak of the COVID-19 pandemic.…”

Social media insights on public perception and sentiment during and after disasters: The European floods in 2021 as a case study

“…The word cloud is a simple way to make an initial assessment of topics in flood tweets, its results have to be further analyzed, since single keywords cannot easily provide a clear sense of a topic (Politis et al, 2021). The following section describes the topic modeling results of the public perception in detail.…”

“…Predominant keywords in a corpus of tweet tokens can indicate the top topics discussed by users (Politis et al, 2021). In this context, word frequency analysis was used to infer the number of important words in the corpus, which can be also used to identify event hotspots and their changing trends.…”

Public Transport Tweets in London, Madrid and Prague in the COVID-19 Period—Temporal and Spatial Differences in Activity Topics

“…The class names may directly indicate how to deal with customers' satisfaction or dissatisfaction in relevant classes. To discover relevant topics, two approaches were applied-an expert interpretation of the content (bag of words where relevant keywords are defined in advance, followed by a supervised interpretation) and artificial intelligence-based unsupervised analysis (i.e., Latent Dirichlet Allocation LDA), where the data is first processed automatically, followed by an expert interpretation of the emerged topics [24].…”