Investigating the Healthiness of Internet-Sourced Recipes Implications for Meal Planning and Recommender Systems Dr. Christoph Trattner Department of New Media Technology
Investigating the Healthiness of Internet-Sourced Recipes
Implications for Meal Planning and Recommender SystemsDr. Christoph Trattner Department of New Media Technology
Context
Trattner, C. and Elsweiler, D. Investigating the Healthiness of Internet-Sourced Recipes: Implications for Meal Planning and Recommender Systems. In Proc. WWW’17.
@ctrattner @delsweil
Why is research on (online) food useful?
Food is one the main concepts that shapes how good we feel and how healthy we are.
According to the WHO, within the last three decades overweight and obesity in the EU population rised dramatically > 30% (especially for the younger generation
Resulting in a cost of approx. € 81 billion a year to help people with chronic diseases
…A government’s response to increasing incidence of lifestyle-related illnesses, such as obesity, has been to
encourage people to cook for themselves…
Cooking inspiration:
Questions RQ1: How healthy are Internet-Sourced recipes with
respect to recognized standards? RQ2: How do user interactions such as ratings, comments
or social bookmarks people apply to recipes relate to the healthiness of recipe content?
RQ3: How healthy are the recipes recommended by standard recommendation algorithms when applied to the food recommendation problem?
RQ4: Can we improve standard recommender algorithms in terms of making the recommendations they offer more healthy?
Question?
Dataset
Nutrition Facts
Basic statistics:
60,983 recipes 125,762 users rating 155,769 users bookmarking 1,032,226 ratings 17,190,534 bookmarks
Dataset: Allrecipes.com
Allrecipes.com popularity
According to Alexa.com
Questions RQ1: How healthy are Internet-Sourced recipes with
respect to recognized standards? RQ2: How do user interactions such as ratings, comments or
social bookmarks people apply to recipes relate to the healthiness of recipe content?
RQ3: How healthy are the recipes recommended by standard recommendation algorithms when applied to the food recommendation problem?
RQ4: Can we improve standard recommender algorithms in terms of making the recommendations they offer more healthy?
Question?
Determining the healthiness
FSA food health criteria
Usda. cook more often at home. available at http://www.choosemyplate.gov/weight-management-calories/weight-management/better-choices/cook-home.html. Last accessed on 20.6.2016. 2011.
WHO
food
hea
lth c
riter
ia
J. Who and F. E. Consultation. Diet, nutrition and the prevention of chronic diseases. World Health Organ TechRep Ser, 916(i-viii), 2003.
Determining the healthiness
Internet recipe healthiness
Internet recipes are quite unhealthy!
User perception
Results when asking users (n=32) how healthy categories are on Allrecipes.com
(Kappa κ = .165, z = 42, p < .001)
Internet recipesvs
TV chef recipesvs
Supermarket ready meals
Comparing recipes
Internet recipe health comparison
Trattner, C. Elsweiler, D. and Simon, H. Estimating the Healthiness of Internet Recipes: A Cross-Sectional Study. Frontiers in Public Health, 2017.
WHO
crite
riaFS
A cr
iteriaHealthiness over time
There is not much change over time!
Questions RQ1: How healthy are Internet-Sourced recipes with respect
to recognized standards? RQ2: How do user interactions such as ratings, comments
or social bookmarks people apply to recipes relate to the healthiness of recipe content?
RQ3: How healthy are the recipes recommended by standard recommendation algorithms when applied to the food recommendation problem?
RQ4: Can we improve standard recommender algorithms in terms of making the recommendations they offer more healthy?
Question?
Spearman rank correlation,* p < .05, ** p < .01, *** p < .001
Results: Interaction Correlation
Questions RQ1: How healthy are Internet-Sourced recipes with respect
to recognized standards? RQ2: How do user interactions such as ratings, comments or
social bookmarks people apply to recipes relate to the healthiness of recipe content?
RQ3: How healthy are the recipes recommended by standard recommendation algorithms when applied to the food recommendation problem?
RQ4: Can we improve standard recommender algorithms in terms of making the recommendations they offer more healthy?
Question?
*** p < .001
Results: Recommender
Libray: LibRecEval: 10 fold-cross validation
∆=𝑡𝑟𝑎𝑖𝑛−𝑝𝑟𝑒𝑑
Std. recommender produce unhealthy
suggestions!
Questions RQ1: How healthy are Internet-Sourced recipes with respect
to recognized standards? RQ2: How do user interactions such as ratings, comments or
social bookmarks people apply to recipes relate to the healthiness of recipe content?
RQ3: How healthy are the recipes recommended by standard recommendation algorithms when applied to the food recommendation problem?
RQ4: Can we improve standard recommender algorithms in terms of making the recommendations they offer more healthy?
Question?
Pearson correlations (= rho) between MAP and nDCG and FSA and WHO health scores (on user level) for individual algorithms.
Sign. high correlation
Correlation
Post-Filter scoring functions
Linear combinations as discussed in Elsweiler et al. (2015) did not work
Re-ranking for health
D. Elsweiler, M. Harvey, B. Ludwig, and A. Said. Bringing the "healthy" into food recommenders. In Proc. ofDRMS’15., pages 33–36.
Results: Recommender (2)
Note: similar results with bookmarks
Std. recommender can be enhanced
to produce healthy suggestions!
OK, that’s it!
for now ;)
Take home:
Internet-sourced recipes are not per
se healthy