search

calgo-lab / green-db

The monorepo that powers the GreenDB.
https://calgo-lab.github.io/green-db/
22 stars 2 forks source link

readme

GreenDB

Preprint: arXiv\ Dataset: DOI

This repository contains code and infrastructure components (Helm charts) that powers the Green Database (GreenDB). This development is part of the research project Green Consumption Assistant (GCA).

The publicly available open GreenDB is a product database. It contains classical product attributes, e.g., name, description, colors, etc. On the other hand, it also includes information about the products' sustainability. Moreover, this sustainability information is transparently evaluated so that it is possible to rank products depending on their sustainability. For more information about the GreenDB dataset or technical details see the dataset on zenodo [1] or our preprint on arXiv [2].

The GreenDB enables the implementation of applications such as the prototypical Chrome Extension Koala - Ecosia Assistant Beta - Grün shoppen, which we built as part of the GCA project. The Koala detects when and what users would like to buy online and offers more sustainable alternatives. More information can be found in one of our working papers [3].

With the open GreenDB, we hope to contribute to the fight against climate crises by enabling others to build applications that support more sustainable consumption. Further, we plan to publish datasets to support the research community.

If you would like to participate with code, data, or anything else, reach out to use!

Where do we get product and sustainability information?

Nowadays, some online shops offer sustainability filters. We use scraping technologies to automatically find these sustainable products and integrate them into the GreenDB. Further, we evaluate the sustainability information and use easily accessible scores (0 - 100) for fine-grained aspects of three sustainability dimensions: social, ecological, and their credibility.

However, manual evaluating products does not scale well. Therefore, we leverage that products can be certified with sustainability labels. The German website Siegelklarheit systematically evaluates sustainability labels and makes their results publicly available. We use these to infer a products' sustainability if it is certified with a sustainability label.

GreenDB schema

The GreenDB schema is highly inspired by schema.org. However, this is still a proof of concept implementation and can differ from their definitions (e.g., the brand column of the green-db table). One of our future plans is to tackle this and make the GreenDB fully compatible with schema.org's Product definition. Our ultimate goal is to eventually contribute an extension of schema.org that integrates the products' sustainability information.

Table: green-db

column name id gtin asin timestamp url source merchant country category name description brand sustainability_labels price currency image_urls colors sizes gender consumer_lifestage
column data type int4 int8 text timestamp text text text text text text text text array[text] numeric text array[text] array[text] array[text] text text
column nullable no yes yes no no no no no no no no no no no no no yes yes yes yes

Please note: Currently, we use our own category strings, which are not documented. However, we plan to switch to the GS1 Global Product Classification taxonomy to rely on a public definition of the category column.

Table: sustainability-labels

column name id timestamp name description cred_credibility eco_chemicals eco_lifetime eco_water eco_inputs eco_quality eco_energy eco_waste_air eco_environmental_management social_labour_rights social_business_practice social_social_rights social_company_responsibility
column data type text timestamp text text int4 int4 int4 int4 int4 int4 int4 int4 int4 int4 int4 int4 int4
column nullable? no no no no yes yes yes yes yes yes yes yes yes yes yes yes yes

Cite us

If you are using our code or the GreenDB dataset, please reference the preprint as:

@misc{https://doi.org/10.48550/arxiv.2205.02908,
  doi = {10.48550/ARXIV.2205.02908},
  url = {https://arxiv.org/abs/2205.02908},
  author = {Jäger, Sebastian and Greene, Jessica and Jakob, Max and Korenke, Ruben and Santarius, Tilman and Biessmann, Felix},
  keywords = {Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences},
  title = {GreenDB: Toward a Product-by-Product Sustainability Database},
  publisher = {arXiv},
  year = {2022}
}

Future plans

Contact

Research

Publications that use the GreenDB

  1. Gossen, M., Jäger, S., Hoffmann, M.L., Biessmann, F., Korenke, R., & Santarius, T. (2022). Nudging Sustainable Consumption: A Large-Scale Data Analysis of Sustainability Labels for Fashion in German Online Retail. Frontiers in Sustainability. DOI: https://doi.org/10.3389/frsus.2022.922984
  2. Jäger, S., Flick, A., Garcia, J.A., Driesch, K.V., Brendel, K., & Biessmann, F. (2022). GreenDB - A Dataset and Benchmark for Extraction of Sustainability Information of Consumer Goods. ArXiv, abs/2207.10733. DOI: https://doi.org/10.48550/arXiv.2207.10733
  3. Flick, A., Jäger, S., Trajanovska, I., Biessmann, F. (2023). Automated Extraction of Fine-Grained Standardized Product Information from Unstructured Multilingual Web Data. In: , et al. Advances in Information Retrieval. ECIR 2023. Lecture Notes in Computer Science, vol 13982. Springer, Cham. https://doi.org/10.1007/978-3-031-28241-6_19

Disclaimer!

This is research code and under development and not supposed to be used in production.