William González-Baquero

Dr. William González-Baquero

Sociología y Comunicación, Universidad de Salamanca, Salamanca, Castille-Leon, Spain
orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 May 2023 in Religions
Verified
Journal Article
Open access
PDF file
WGWilliam González-Baquero
William González-Baquero
Javier J. AmoresJavier J. Amores
Javier J. Amores
Carlos Arcila CalderónCarlos Arcila Calderón
Carlos Arcila Calderón
Social media, especially Twitter, has become a platform where hate, toxic, intolerant, and discriminatory speech is increasingly spread. These messages are aimed at different vulnerable social groups, due to some of their differentiating characteristics with respect to the dominant one, whether they are phenotypic, religious, cultural, gender, sexual, etc. Of all these minorities, one of the most affected is the Muslim community, especially since the beginning of the Mediterranean refugee crisis, during which migration from the Middle East and North Africa increased considerably. Spain does not escape this reality as, given its proximity to Morocco, it is one of the main destinations for migrants from North Africa. In this context, there are already several studies focused on specifically investigating Islamophobic speech disseminated on social platforms, normally focused on specific cases. However, there are still no studies focused on analyzing the entire conversation around Islam and the Muslim community that takes place on Twitter and in a southern European country such as Spain, aiming to identify the latent sentiments and the main underlying topics and their characteristics, which would help to relativize and dimension the relevance of Islamophobic messages, as well as to analyze them from a more solid base. The main objective of the present study is to identify the most frequent words, the main underlying topics, and the latent sentiments that predominate in the general conversation about Islam and the Muslim community on Twitter in Spain and in Spanish during the last 8 years. To do this, 190,320 messages that included keywords related to Muslim culture and religion were collected and analyzed using computational techniques. The findings show that the most frequent words in these messages were mostly descriptive and not derogatory, and the predominant latent topics were mostly neutral and informative, although two of them could be considered reliable indicators of Islamophobic rejection. Similarly, while the overall average sentiment in this conversation trended negatively, neutral and positive messages were more prevalent. However, in the negative messages, the sentiment was considerably more pronounced.

1 January 2025 in Cuadernos.info
Verified
Journal Article
Open access
MAMartín Oller Alonso
Martín Oller Alonso
Carlos Arcila CalderónCarlos Arcila Calderón
Carlos Arcila Calderón
Javier J. AmoresJavier J. Amores
Javier J. Amores
Patricia Sánchez-HolgadoPatricia Sánchez-Holgado
Patricia Sánchez-Holgado
MVMaximiliano Frías Vásquez
Maximiliano Frías Vásquez
MBMarcos Paulo Gomes Barbosa
Marcos Paulo Gomes Barbosa
Social media platforms have become critical spaces for the dissemination of negative, toxic, intolerant, and discriminatory sentiments and messages targeting stigmatized minority groups defined by their ethnicity, religion, culture, origin, gender identity, or sexual orientation. This article presents a sentiment analysis applied to Spanish-language messages on X, Instagram, and Facebook about migrants, Muslims, Jews, the LGBTIQ+ community, and the Roma population in Spain. The SentiStrength tool analyzed 523,374 messages containing keywords related to these groups. The results show that Instagram and Facebook promote a more favorable environment, with slightly positive latent sentiment averages towards these stigmatized communities. In contrast, X is characterized by a higher prevalence of negative sentiments in the messages analyzed overall. In terms of comparisons between groups, the most negative average sentiment was found in messages related to Jews and migrants, particularly on X. Conversely, the messages in which the most positive latent sentiment average was found related to the Roma community and LGBTIQ+ groups.

31 July 2026 in Revista Austral de Ciencias Sociales
Verified
Journal Article
Open access
PDF file
MOMartín Oller-Alonso
Martín Oller-Alonso
Carlos Arcila CalderónCarlos Arcila Calderón
Carlos Arcila Calderón
MFMaximiliano Frías-Vázquez
Maximiliano Frías-Vázquez
WGWilliam González-Baquero
William González-Baquero
En este artículo desarrollamos un marco teórico y conceptual que distingue la judeofobia religiosa del antisemitismo moderno y permite separar el odio antijudío de la crítica política legítima a Israel. Para ello, articulamos tres definiciones operativas a partir de la Holocaust Remembrance Alliance (IHRA), la Nexus Task Force (Nexus) y la Jerusalem Declaration on Antisemitism (JDA), con las que operacionalizamos el antisemitismo (conspiraciones sobre “poder judío”, culpabilización colectiva, doble lealtad, negación o banalización de la Shoá, negación del derecho a existir y dobles raseros extremos) y delimitamos aquello que no lo es per se (crítica basada en hechos a las políticas de Israel, apoyo a derechos palestinos o BDS no violento sin estereotipos). Sobre esta base analizamos el antisemitismo digital y su entrelazamiento con el (anti)sionismo y la islamofobia en X, Facebook, Instagram y TikTok. Partimos de un corpus de 798.475 mensajes en español geolocalizados en España (abril–agosto de 2024). En nuestro método, combinamos la detección automatizada de toxicidad (API Perspective) con modelos de categorización temática de odio dirigido a cinco grupos vulnerables: antisemitismo, antigitanismo, islamofobia/racismo antiárabe, xenofobia y LGBTIQfobia. Los resultados muestran (i) picos de “neoantisemitismo” ligados a coyunturas relacionadas con la escalada bélica en Israel–Gaza, donde el antisionismo funciona como vehículo de tropos clásicos; (ii) transferencia cruzada de hostilidad entre comunidades judías y musulmanas y solapamientos con otras formas de racismo; (iii) así como diferencias significativas entre plataformas en conspiracionismo y negacionismo, siendo las analogías nazismo–Israel y los dobles raseros extremos marcadores de alto riesgo. Finalmente, mostramos cómo las lógicas algorítmicas de “enragement” y la difusión de contenidos tóxicos convierten estas redes (en especial X) en entornos proclives a la intensificación del (discurso de) odio, más que a su mitigación. Como posibles contraposiciones a esta violencia, proponemos trabajar en la alfabetización conceptual, los protocolos de moderación basados en IHRA–Nexus–JDA, la monitorización de crisis y las medidas de protección para comunidades judías y musulmanas, alineadas con los marcos de la Organización de Naciones Unidas (ONU), la Organización para la Seguridad y la Cooperación en Europa (OSCE) y la Unión Europea (UE), a fin de mitigar el odio (online) sin restringir el disenso y la crítica legítima.

17 August 2026 in Central European Journal of Communication
Verified
Journal Article
Open access
PDF file
WGWilliam González-Baquero
William González-Baquero
Carlos Arcila CalderónCarlos Arcila Calderón
Carlos Arcila Calderón
Javier J. AmoresJavier J. Amores
Javier J. Amores
RZRocio Zamora
Rocio Zamora
PCPilar Garrido Clemente
Pilar Garrido Clemente
PRPedro Rojo
Pedro Rojo
Islamophobic hate speech on social media includes overt toxicity and contextual frames that associate Muslims with invasion, criminality, cultural threat, or racialized exclusion. This study examines 21,326 Spanish-language tweets about Islam and Muslims posted in Spain between 2010 and 2022. We combine human annotation, exploratory discursive coding, Perspective API scores, and a multilingual BERT classifier. Human coders classified 28.2% of tweets as Islamophobic hate speech. Explicit hostility was the most frequent pattern, followed by invasion, cultural threat, criminality, race/ethnicity, and socioeconomic burden. Although hate-labeled tweets received significantly higher toxicity-related scores than non-hate tweets, many remained below a 0.70 high-probability threshold. The BERT classifier achieved 0.89 accuracy, 0.97 precision, 0.85 recall, and 0.91 F1 for the hate class. The findings show that generic toxicity measures capture overt aggression more effectively than contextual hostility, supporting target-specific, context-sensitive detection.