Browse: meal_type categories near-empty (Lunch, Dinner, Snack, Beverage, Side Dish) #122

New issue

Closed

opened 2026-04-26 19:02:38 -07:00 by pyr0ball · 0 comments

pyr0ball commented

2026-04-26 19:02:38 -07:00

Owner

Problem

The recipe browser meal_type domain has very uneven coverage. Breakfast has substantial results, but Lunch, Dinner, Snack, Beverage, and Side Dish return very few or no recipes.

Root cause: the keyword lists in browser_domains.py do not align with the actual category and keywords values in the food.com corpus.

Investigation needed

Run the following against the corpus DB to understand the actual distribution:

SELECT category, count(*) FROM recipes
GROUP BY category ORDER BY count(*) DESC LIMIT 100;

Also check which keywords actually appear in the keywords JSON column:

SELECT keywords, count(*) FROM recipes
WHERE keywords IS NOT NULL
GROUP BY keywords ORDER BY count(*) DESC LIMIT 50;

Fix

Update app/services/recipe/browser_domains.py meal_type keyword lists to include the corpus-specific category values that map to each meal type. For example:

Dinner may need to include corpus categories like "chicken dishes", "beef dishes", "pork", "seafood", "vegetables" etc.
Lunch may need to include "salads", "sandwiches", "soups" as corpus category values
Snack may need to include "appetizers", "snacks", "dips" etc.
Beverage and Side Dish similarly

Acceptance criteria

Each meal_type category returns at least 100 recipes from the corpus
Category counts are visible in the browser domain list endpoint (/recipes/browse/meal_type)
Verified with SELECT category, count(*) FROM recipes GROUP BY category cross-referenced against updated keyword lists

See also: domain keyword enrichment audit ticket (sister issue).

## Problem The recipe browser `meal_type` domain has very uneven coverage. Breakfast has substantial results, but Lunch, Dinner, Snack, Beverage, and Side Dish return very few or no recipes. Root cause: the keyword lists in `browser_domains.py` do not align with the actual `category` and `keywords` values in the food.com corpus. ## Investigation needed Run the following against the corpus DB to understand the actual distribution: ```sql SELECT category, count(*) FROM recipes GROUP BY category ORDER BY count(*) DESC LIMIT 100; ``` Also check which keywords actually appear in the `keywords` JSON column: ```sql SELECT keywords, count(*) FROM recipes WHERE keywords IS NOT NULL GROUP BY keywords ORDER BY count(*) DESC LIMIT 50; ``` ## Fix Update `app/services/recipe/browser_domains.py` `meal_type` keyword lists to include the corpus-specific category values that map to each meal type. For example: - **Dinner** may need to include corpus categories like `"chicken dishes"`, `"beef dishes"`, `"pork"`, `"seafood"`, `"vegetables"` etc. - **Lunch** may need to include `"salads"`, `"sandwiches"`, `"soups"` as corpus category values - **Snack** may need to include `"appetizers"`, `"snacks"`, `"dips"` etc. - **Beverage** and **Side Dish** similarly ## Acceptance criteria - Each meal_type category returns at least 100 recipes from the corpus - Category counts are visible in the browser domain list endpoint (`/recipes/browse/meal_type`) - Verified with `SELECT category, count(*) FROM recipes GROUP BY category` cross-referenced against updated keyword lists ## Related See also: domain keyword enrichment audit ticket (sister issue).