Question Answering for Food Sustainability

This example outlines a collection of resources that could be used to answer complex questions about sustainability, while providing context, and tracing the sources of different data, and suggesting possible opportunities for synthesis.

• "you don't always know what question you're trying to ask until you see some answers"

  • an overlay should expose multiple possible routes to answering a question, allowing the asker to narrow down the query if necessary. An example of this is google translate:

  • one interlay might support and enumerate a diversity of overlays, and draw together the locations of the required resources (or identify where they don’t exist)

1. Context:

• An example overlay is voice interface for an automated assistant: part of a broader suite that acts as a cooking aid

  • this tool will work alongside + potentially interface to question-answering interfaces that provide information about:

    • ingredient substitutions

    • recipes

    • cooking processes

    • flavour combinations

  • What queries might this overlay require?

  • What kind of interlay might support this and other sources?

• Question-answering about food sustainability requires a complex set of intersecting datasets, as questions themselves are open to a number of interpretations.

  • The sustainability of different foods is strongly determined by your location, budget, and local food options.

  • 'Sustainability' could apply to greenhouse gas emissions, the seasonality of ingredients, water consumption, animal welfare, the likelihood of pesticide use, the effects on local ecosystems, the logistical practices of food distributors, the maintenance of soil quality and the sustainment of agricultural communities through fair trade and food sovereignty.

2. Data Sources:

• general

  • global impacts of food production database. Analyses environmental impact of 40 products representing ~90% of global protein and calorie consumption. Methods given in Poore + Nemecek, Reducing Food’s Environmental Impacts through Producers and Consumers. Covers five environmental impact indicators:

    • land use

    • freshwater withdrawals weighted by local water scarcity

    • GHG emissions

    • acidifying emissions

    • andeutrophying emissions

  • Chaudry, Gustafson + Mathys: Multi-indicator Sustainability Assessment of Global Food Systems (paper + accompanying data). Comparative analysis of different global food systems, assessed against UN Sustainable Development Goals. draws together multiple data sources to cover:

    • Food Nutrient Adequacy

    • Ecosystem Stability

    • Affordability and Availability

    • Sociocultural Wellbeing

    • Resilience

    • Food Safety

  • water footprints of crops and derived products

    • + later paper, water footprint benchmark values

• lists of (un)sustainable ingredients

  • One starting place is a list of ingredients considered 'sustainable' or 'unsustainable', either internationally or locally, and their suggested substitutions. Such ingredients could include:

    • the good fish guide

    • regional milk comparisons (actually based on the Poore and Nemecek data above)

• regional seasonality

  • Seasonality can be defined on both a global and a local basis

    • regional agronomy python library

    • worldwide hardiness zones

    • seasonal food availability calendar

    • good fish guide fish of the month recipe (uk)

• origins of crops in different locations

  • CIAT origin of crops (global)

    • dataset released on Dryad for Khoury et al 2016, Origins of food crops connect countries worldwide

  • origins of food supply (uk)

  • Spatial Production Allocation Model -- SPAM. Uses a cross-entropy model to estimate crop production and distribution

    • on Dataverse

• local agricultural legislation

  • combined with data about where food comes from, having information about different agricultural standards and practices can inform answers about relative sustainability:

    • EIU food sustainability index excel workbook -- compares different countries against a series of sustainability indicators

    • fairtrade global standards and agricultural legislation

    • global agricultural greenhouse gas emissions

• ingredient substitution

  • substituting resource-intensive produce for appropriate replacements, requires an ingredient substitution dataset:

    • schema

• sustainable recipes

  • suggesting recipes by default that consist of seasonal ingredients: requires a recipe dataset that intersects with some kind of seasonal ingredients dataset, plus some idea of user preferences and region.

    • recipe dataset schema

3. Flow + Filtering:

• What are some expected queries -> what should they return? How are they interpreted at each layer?

• (one) Overlay:

  • Receives a voice input. Needs to interpret as a text string as a query and process into keywords etc.

  • Sends processed query to the interlay ->

  • Interprets and outputs the results of the query

• Interlay:

  • Needs to read the context of the data

  • Takes form of an api endpoint/middleware

• Underlay (s):

  • Sources available as part of a schematised graph. Data has been preprocessed into collections.

4. Nexus:

• Who maintains this system?

  • multiple separate datasets maintained by sustainability orgs:

• Who maintains the interlay? depends on the specificity needed. some kind of search interface, wikidata?

5. Similar efforts

• For other types of sustainability, see: Food deserts, Food pantries, food cooperatives

• For other parts of the food supply chain, see: Farming sustainability