Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Assessing post-disaster recovery using sentiment and topic analysis on social media posts: L'Aquila, Italy

Contreras Mojica, Diana, Forino, Giuseppe, Pezzica, Camilla ORCID: https://orcid.org/0000-0002-0512-7591, Liberatore, Federico ORCID: https://orcid.org/0000-0001-9900-5108 and Veliu, Enes 2024. Assessing post-disaster recovery using sentiment and topic analysis on social media posts: L'Aquila, Italy. Presented at: 18th World Conference on Earthquake Engineering - 18WCEE 2024, Milan, Italy, 30 June - 05 July 2024. World Conference on Earthquake Engineering Online Proceedings. IAEE,
Item availability restricted.

[thumbnail of Contreras et al_2024_Assessing post-disaster recovery using SA and TA_LAquila_240131.pdf] PDF - Published Version
Restricted to Repository staff only until 8 August 2025 due to copyright restrictions.

Download (1MB)

Abstract

In this conference paper, we assess the progress in post-disaster recovery by analysing 4349 tweets posted between 4th and 10th April 2019 that we collected around the 10th anniversary of the earthquake in L’Aquila. Text data collected from social media is unstructured; therefore, we need to use natural language processing techniques such as topic and sentiment analysis to extract meaningful information to assess recovery. Sentiment Analysis (SA), or opinion mining, classifies people's opinions, expressed in written text, into a specific polarity, i.e. positive, negative or neutral. Topic Analysis (TA) classifies data into recurrent themes or topics users address in their posts. These analyses can be done at the tweet or sentence level, and the classification into polarities and topics can be done manually or automated. We analysed a sample of 10% of tweets covering 24 hours daily. The SA and TA were performed at both levels, and the classification was done manually. The SA at the tweet level indicates that most of the posts were classified into neutral polarity, followed closely by positive and negative. Similar results were obtained from the SA analysis at the sentence level, only with variation in percentages of the sentence classified into each polarity. The TA at the tweet and sentence levels indicate that the most frequently addressed topics by users at both levels were commemoration actions, restoration, reconstruction, governance, and distress. The SA per topics at the tweet level indicates that the topics with neutral polarity are critical infrastructure, commemoration actions, and seismic information. Topics with positive polarity are cultural heritage, early recovery, emergency response, lifelines, preparedness, restoration, solidarity messages/actions and urban facilities, which are considered successful aspects of the recovery process in this methodology. Topics with negative polarity are building damages, construction practices, depopulation and displacement, distress, governance, injuries and casualties, intensity, prevention, and reconstruction, considered the failures of the process. We also conducted a two-tailed Pearson correlation analysis between polarity and topics of tweets for each day, which confirmed, in most cases, the results of the SA for each topic at the tweet level. According to the methodology applied, we can conclude that the perception of the recovery of L’Aquila by the 10th anniversary is mainly neutral.

Item Type: Conference or Workshop Item (Paper)
Date Type: Published Online
Status: Published
Schools: Schools > Computer Science & Informatics
Schools > Architecture
Schools > Earth and Environmental Sciences
Subjects: G Geography. Anthropology. Recreation > G Geography (General)
H Social Sciences > H Social Sciences (General)
H Social Sciences > HA Statistics
T Technology > TA Engineering (General). Civil engineering (General)
Publisher: IAEE
Funders: Cardiff University
Date of First Compliant Deposit: 1 July 2025
Date of Acceptance: 5 February 2024
Last Modified: 22 Jul 2025 14:53
URI: https://orca.cardiff.ac.uk/id/eprint/179451

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics