Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

You Tweet What You Eat: Studying Food Consumption Through Twitter (1412.4361v2)

Published 14 Dec 2014 in cs.CY, cs.SI, and physics.soc-ph

Abstract: Food is an integral part of our lives, cultures, and well-being, and is of major interest to public health. The collection of daily nutritional data involves keeping detailed diaries or periodic surveys and is limited in scope and reach. Alternatively, social media is infamous for allowing its users to update the world on the minutiae of their daily lives, including their eating habits. In this work we examine the potential of Twitter to provide insight into US-wide dietary choices by linking the tweeted dining experiences of 210K users to their interests, demographics, and social networks. We validate our approach by relating the caloric values of the foods mentioned in the tweets to the state-wide obesity rates, achieving a Pearson correlation of 0.77 across the 50 US states and the District of Columbia. We then build a model to predict county-wide obesity and diabetes statistics based on a combination of demographic variables and food names mentioned on Twitter. Our results show significant improvement over previous CHI research (Culotta'14). We further link this data to societal and economic factors, such as education and income, illustrating that, for example, areas with higher education levels tweet about food that is significantly less caloric. Finally, we address the somewhat controversial issue of the social nature of obesity (first raised by Christakis & Fowler in 2007) by inducing two social networks using mentions and reciprocal following relationships.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (3)
  1. Sofiane Abbar (18 papers)
  2. Yelena Mejova (41 papers)
  3. Ingmar Weber (66 papers)
Citations (244)

Summary

An Essay on "You Tweet What You Eat: Studying Food Consumption Through Twitter"

The paper, "You Tweet What You Eat: Studying Food Consumption Through Twitter," presents a meticulous investigation into the potential use of Twitter as a tool for analyzing dietary patterns on a large scale within the United States. The research is centered on exploiting the extensive data available on Twitter, particularly focusing on food-related content to infer dietary habits and correlate them with public health issues, such as obesity and diabetes.

Methodological Approach

The paper utilized data from 210,000 Twitter users, aggregating approximately 502 million tweets. The authors deployed a Naive Bayes classifier to identify tweets related to food using an enriched lexicon. This lexicon included terms derived from manual curation and nutritional information from online sources. The paper correlates food mentions with caloric intake estimates and further validates these findings by comparing predicted obesity and diabetes statistics across U.S. regions using actual health data. Pearson correlations of 0.77 for obesity and 0.66 for diabetes statewide are notable, showcasing the model's efficacy in reflecting public health trends through social media dialogue.

Statistical Analysis and Model Comparisons

Significantly, the research extends beyond mere correlation by constructing regression models to predict county-level health statistics. It compares models based on LIWC lexicons with those using food mentions, revealing substantial predictive power in food mention-based models, particularly when combined with demographic data. The food-demographic model outperforms traditional approaches, delineating its potential in public health monitoring. Highlighting this correlation offers pragmatic implications for real-time health surveillance and dietary studies.

Sociocultural Insights and Demographic Analysis

Delving deeper, the authors explore demographics and user interests through Twitter profiles and networks, demonstrating how gender, income, education, and geographical setting (rural vs. urban) influence dietary mentions. They observed that urban users tweet less about caloric foods compared to their rural counterparts, aligning societal and cultural observations with empirical data analytics.

The paper also examines the association between users' social networks and their dietary habits. Employing networks based on Twitter mentions and reciprocal followership, the paper finds "homophily" in dietary expressions among connected users, further extending the paradigms suggested by Christakis and Fowler regarding societal factors influencing obesity. An increased probability of food-related tweet similarity corresponds with higher network connectivity, reinforcing the imperative of network effects in dietary research.

Implications and Future Work

The paper presents pivotal implications for utilizing social media as a comprehensive data source for dietary and nutritional research. It proposes the potential for targeted public health interventions, leveraging the synchronous nature of Twitter data to dynamically inform public health strategies. However, it recognizes limitations in sampling bias and the sporadic nature of food-related tweets which can misrepresent regular dietary habits.

Future directions pointed out include enhancing classification accuracy, especially for detecting the nuances of food consumption and incorporating real-time analytics for dynamically monitoring dietary patterns. Additionally, integrating more nuanced demographic factors and advancing computational models can substantiate the current methodological framework, allowing for broader applicability across diverse populations.

In conclusion, the paper systematically investigates Twitter's viability as a tool for dietary health monitoring, articulating compelling quantitative and qualitative insights. By bridging data science with public health, it elucidates pathways for ephemeral yet impactful health interventions and socio-cultural analysis, while presenting a robust commentary on the role of digital footprints in profiling societal health trends.