Beyond single-stage classification: A hybrid AI-expert validation framework for environmental monitoring of European food waste prevention.

European food loss and waste prevention records are reported through heterogeneous sources and formats, limiting their comparability and reuse for environmental monitoring. We developed a hybrid Zero-Shot Learning and Human-in-the-Loop framework to transform these records into a traceable, multidimensional corpus while retaining expert oversight. The workflow integrated 1164 Action Tracker records with 161 expert-curated records. 1324 records were retained after one semantically insufficient entry was excluded. A confidence-based routing step sent 331 Action Tracker records to full expert review, while a separate blind audit examined 125 of 832 higher-confidence records. Expert review showed that classification confidence and eligibility as a discrete prevention action are distinct quality dimensions. On 75 qualifying audit records, model-expert agreement remained limited, supporting automated inference as a first-pass routing tool rather than a substitute for expert assessment. The resulting taxonomy also showed that 512 of 1163 classified Action Tracker records (44.0%) were assigned to two or more food supply chain stages. These are model-assigned labels rather than verified proportions of qualifying interventions, but they retain cross-stage information that a strictly single-stage representation would omit. The framework does not estimate environmental impacts directly. Instead, it structures heterogeneous prevention evidence so that intervention characteristics, data gaps, and cross-stage coverage are explicit before subsequent linkage to activity data, life-cycle inventories, and policy assessment.

Authors

Institutions

Publication Details

Journal
PubMed
Published
2026-11-14
DOI
https://doi.org/10.1016/j.jenvman.2026.131313
Primary Topic
Food Waste Reduction and Sustainability
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Beyond single-stage classification: A hybrid AI-expert validation framework for environmental monitoring of European food waste prevention.

Wilmer Garzón, Ainhoa Alonso‐Vicario, Manuel Amador Cervera, Haizea Cruz Elejalde
PubMed
Food Waste Reduction and Sustainability
article

Beyond single-stage classification: A hybrid AI-expert validation framework for environmental monitoring of European food waste prevention.

Wilmer Garzón, Ainhoa Alonso‐Vicario, Manuel Amador Cervera, Haizea Cruz Elejalde
article en

Abstract

European food loss and waste prevention records are reported through heterogeneous sources and formats, limiting their comparability and reuse for environmental monitoring. We developed a hybrid Zero-Shot Learning and Human-in-the-Loop framework to transform these records into a traceable, multidimensional corpus while retaining expert oversight. The workflow integrated 1164 Action Tracker records with 161 expert-curated records. 1324 records were retained after one semantically insufficient entry was excluded. A confidence-based routing step sent 331 Action Tracker records to full expert review, while a separate blind audit examined 125 of 832 higher-confidence records. Expert review showed that classification confidence and eligibility as a discrete prevention action are distinct quality dimensions. On 75 qualifying audit records, model-expert agreement remained limited, supporting automated inference as a first-pass routing tool rather than a substitute for expert assessment. The resulting taxonomy also showed that 512 of 1163 classified Action Tracker records (44.0%) were assigned to two or more food supply chain stages. These are model-assigned labels rather than verified proportions of qualifying interventions, but they retain cross-stage information that a strictly single-stage representation would omit. The framework does not estimate environmental impacts directly. Instead, it structures heterogeneous prevention evidence so that intervention characteristics, data gaps, and cross-stage coverage are explicit before subsequent linkage to activity data, life-cycle inventories, and policy assessment.

PubMedVol. 420
Universidad de Deusto (ES)
Food Waste Reduction and Sustainability
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon
Beyond single-stage classification: A hybrid AI-expert validation framework for environmental monitoring of European food waste prevention. — Wilmer Garzón, Ainhoa Alonso‐Vicario, et al. · PubMed (2026) | TGRS Research Map | TGRS