Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quantitative Information Extraction from Humanitarian Documents

Daniele Liberatore, Kyriaki Kalimeri, Derya Sever, Yelena Mejova

TL;DR

The paper tackles the challenge of rapidly extracting quantitative information from humanitarian documents to support crisis response and anticipatory action. It introduces a richly annotated dataset of 755 English humanitarian excerpts with a six-label schema (Number, Unit, Modifier, EventP, EventA, EventO) and a multistage NLP pipeline that detects numbers, links them to units and modifiers, and anchors them to events. The approach shows consistent improvements over baselines such as SpaCy and the Comprehensive Quantity Extractor, with particularly strong gains in shelter and WASH domains and across diverse geographic contexts. This work provides open data and code, enabling reproducibility and serving as a foundation for data-driven humanitarian decision-making and integration with spatial-temporal analyses and predictive models.

Abstract

Humanitarian action is accompanied by a mass of reports, summaries, news, and other documents. To guide its activities, important information must be quickly extracted from such free-text resources. Quantities, such as the number of people affected, amount of aid distributed, or the extent of infrastructure damage, are central to emergency response and anticipatory action. In this work, we contribute an annotated dataset for the humanitarian domain for the extraction of such quantitative information, along side its important context, including units it refers to, any modifiers, and the relevant event. Further, we develop a custom Natural Language Processing pipeline to extract the quantities alongside their units, and evaluate it in comparison to baseline and recent literature. The proposed model achieves a consistent improvement in the performance, especially in the documents pertaining to the Dominican Republic and select African countries. We make the dataset and code available to the research community to continue the improvement of NLP tools for the humanitarian domain.

Quantitative Information Extraction from Humanitarian Documents

TL;DR

The paper tackles the challenge of rapidly extracting quantitative information from humanitarian documents to support crisis response and anticipatory action. It introduces a richly annotated dataset of 755 English humanitarian excerpts with a six-label schema (Number, Unit, Modifier, EventP, EventA, EventO) and a multistage NLP pipeline that detects numbers, links them to units and modifiers, and anchors them to events. The approach shows consistent improvements over baselines such as SpaCy and the Comprehensive Quantity Extractor, with particularly strong gains in shelter and WASH domains and across diverse geographic contexts. This work provides open data and code, enabling reproducibility and serving as a foundation for data-driven humanitarian decision-making and integration with spatial-temporal analyses and predictive models.

Abstract

Humanitarian action is accompanied by a mass of reports, summaries, news, and other documents. To guide its activities, important information must be quickly extracted from such free-text resources. Quantities, such as the number of people affected, amount of aid distributed, or the extent of infrastructure damage, are central to emergency response and anticipatory action. In this work, we contribute an annotated dataset for the humanitarian domain for the extraction of such quantitative information, along side its important context, including units it refers to, any modifiers, and the relevant event. Further, we develop a custom Natural Language Processing pipeline to extract the quantities alongside their units, and evaluate it in comparison to baseline and recent literature. The proposed model achieves a consistent improvement in the performance, especially in the documents pertaining to the Dominican Republic and select African countries. We make the dataset and code available to the research community to continue the improvement of NLP tools for the humanitarian domain.
Paper Structure (13 sections, 5 figures, 5 tables)

This paper contains 13 sections, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Quantity Extraction
  4. Data Extraction for Humanitarian Domain
  5. Data Collection
  6. Humanitarian-focused Information Annotation
  7. Quantity Extraction System Design
  8. Number extraction
  9. Unit extraction
  10. Results
  11. Number extraction
  12. Unit extraction
  13. Discussion & Conclusions

Figures (5)

  • Figure 1: Distribution over sectors and countries of the selected excerpts from HumSet.
  • Figure 2: Annotation example. The annotation centers around the Number label, while the Unit, and Events are identified in the relation to the Number.
  • Figure 3: Frequencies of units.
  • Figure 4: Example of a dependency parsing tree
  • Figure 5: Improvement in F1 of the final model vs. baseline.