Analysis of Breast Cancer Information on Facebook Using Neural Network–Based Topic Modeling and Metadata Analysis of English and Spanish Content: Comparative Study
Notice bibliographique
Résumé
BACKGROUND: Breast cancer is the most common cancer diagnosis among women, with approximately 2.3 million new cases annually. When faced with a cancer diagnosis, individuals often turn to the internet for information or reassurance, despite the risk of encountering low-quality or incorrect information. While this observation is well documented in English, limited work has been done to understand the quality of breast cancer information in Spanish, the second most commonly spoken language in the United States. OBJECTIVE: This study uses natural language processing methods and quantitative modeling to analyze English and Spanish breast cancer posts from Facebook, a vital source of health-related information for 40% of English-speaking and 60% of Spanish-speaking adults in the United States. METHODS: Using the CrowdTangle application programming interface, we collected and processed 243,029 English-language and 96,334 Spanish-language Facebook posts. We applied BERTopic with the all-MiniLM-L6 model and k-means clustering to infer thematic structures and used coherence scores to determine the optimal number of topics for each language. Descriptive statistics compared metadata differences across languages. We calculated descriptive statistics and ran inferential tests for likes, comments, and shares. Finally, we examined the top 1% (n=2430 English and n=963 Spanish) of the most engaged content to analyze differences in poster characteristics across languages. RESULTS: Coherence scores indicated an optimal topic solution of k=40 (coherence=0.58) for English and k=30 (coherence=0.52) for Spanish. Thematically, we observed similar content in both languages, with topics spanning mammography, breast cancer events, pink ribbon month, and personal narratives. However, Spanish posts included local and municipal breast cancer events not present in English. Additionally, Spanish posts were more likely to mention at-home breast exams, which are no longer recommended in the United States. Engagement behavior showed statistically significant differences by language across likes, comments, and shares. English posts exhibited more consistent liking and sharing behavior, while Spanish posts showed more consistency in commenting. The top 1% (n=2430) of engaged content in English came from leading breast cancer nonprofits, whereas in Spanish (n=963, 1%), it originated from local governments or food and beverage companies. CONCLUSIONS: Facebook breast cancer content is generally consistent across languages. However, differences in engagement behavior suggest that English- and Spanish-speaking populations engage with content differently, highlighting cultural variability that warrants further exploration. Notably, leading cancer authorities may not have a strong presence in Spanish, indicating that the most accurate and up-to-date information may not be reaching a population particularly prone to worse breast cancer prognoses.
Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.
Comment cette classification a été obtenuedéplier
Prédiction machine sur la base complète
Imitation des enseignantsNi prévalence calibrée, ni vérité terrain. Validation humaine à venir. Le volet Gemma est une étiquette directe du modèle pour chaque travail de la base, lue sur la notice réduite au titre. Le volet Codex est un classifieur appris des 10 348 étiquettes directes de Codex et calibré sur les taux pondérés de l'échantillon; les champs sans appui suffisant ne portent aucun appel Codex. Le mode candidate est l'union des deux volets; le consensus est leur intersection. Ces sorties portent le statut machine_predicted_unvalidated et ne sont pas des étiquettes humaines.
Scores du classifieur distillé par catégorie (deux têtes)
| Catégorie | Codex | Gemma |
|---|---|---|
| Métarecherche | 0,002 | 0,007 |
| Méta-épidémiologie (sens strict) | 0,000 | 0,000 |
| Méta-épidémiologie (sens large) | 0,000 | 0,001 |
| Bibliométrie | 0,003 | 0,002 |
| Études des sciences et des technologies | 0,000 | 0,000 |
| Communication savante | 0,001 | 0,001 |
| Science ouverte | 0,000 | 0,001 |
| Intégrité de la recherche | 0,000 | 0,000 |
| Charge utile insuffisante (le modèle a refusé de juger) | 0,001 | 0,000 |
Scores machine (provisoires)
Les deux têtes enseignantes du modèle étudiant, lues sur ce travail. Un score ordonne la base pour la relecture; il n'affirme jamais une catégorie, et le statut de validation accompagne chaque rangée tel quel.
Scores de référence d'un modèle non mature (critères de maturité non atteints, 7 itérations). Un score ordonne; il n'affirme jamais une catégorie.
score_only:v0-immature-baseline · tel quel depuis la passe de notation : score_only signifie que le nombre peut ordonner les travaux, et qu'aucune étiquette de catégorie n'en découleClassification
machine, non validéePrédiction automatique; un appel candidat d’une seule source (Gemma direct ou Codex distillé), pas un consensus.
Le détail, modèle par modèle et score par score, se trouve en fin de page sous « Comment cette classification a été obtenue ».