GIZ Development Project Reports: 490 Evaluations Using Text Mining
by Korb, Sebastian / Harvard Dataverse·Updated 2mo ago
Available on 1 platform
Sign in to view source links and access this dataset
Description
490 project reports from 159 German development projects, commissioned by GIZ's Corporate Evaluation Unit, form the empirical basis for two evaluative studies completed in 2023 and 2025. The studies assess project effectiveness for BMZ-funded projects valued below EUR 3 million, covering projects ending between October 2018 and December 2023. Methodologically, they combine automated table extraction with computational text mining, including a fine-tuned BERT model to classify evaluative statements.
Use Cases
Training text classification models for evaluative statements based on the described fine-tuned BERT approach.
Conducting topic modeling to identify common themes in project success or failure factors mentioned in the description.
Performing dictionary-based sentiment analysis on narrative project report content as described.
Benchmarking automated evaluation methods against traditional assessment criteria like OECD DAC effectiveness.
Strengths
Includes 490 reports from 159 distinct projects, providing a substantial corpus for analysis.
Covers a defined time range with project end dates from October 2018 to December 2023.
Applies a multi-method approach combining automated table extraction, BERT classification, topic modeling, and word frequency analysis.
Addresses a specific gap in evaluation for smaller projects (below EUR 3 million) as recommended by the German Federal Audit Office.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count for the underlying structured data is unknown, which may limit suitability assessment.
The data scope is limited to projects commissioned by a single German ministry (BMZ) and implemented by GIZ.
Provenance
Source
Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) Corporate Evaluation Unit.
Collection Method
Combined automated extraction from project database metadata and computational text mining of narrative reports.
Time Range
Projects with end dates from October 2018 to December 2023.
Freshness
Last updated 2026-05-06 21:02:41; freshness should be verified.
Geography
null
License is unknown; terms of use must be verified before application.