Agricultural Research Word Vectors Trained on PubAg and Ag Data Commons

A Word2Vec model with 300-dimensional vectors, fine-tuned from Google News vectors on agricultural research text. It was trained on titles and abstracts from the USDA's PubAg database and titles and descriptions from the Ag Data Commons. The model was created by the Department of Agriculture for a recommendation system and program analysis.

Use Cases

• Build a recommendation system for agricultural datasets and papers based on semantic similarity of titles and abstracts.
• Analyze semantic relationships between research projects within national programs based on their textual descriptions.
• Create visualizations of research topics and program management layers based on word vector similarities.
• Perform semantic search or text classification on documents containing agricultural research terminology.

Strengths

• Model uses 300-dimensional vectors, a standard and computationally efficient size for word embeddings.
• Vectors are fine-tuned on domain-specific text from two authoritative USDA sources: PubAg and Ag Data Commons.
• Trained using the Continuous Bag of Words (CBOW) algorithm with negative sampling, a common and effective method.

Limitations

• Row count and vocabulary size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; the specific structure of the model files must be inferred after download.
• Data may reflect temporal or source bias inherent to the specific agricultural publications and datasets used.

Provenance

Source

Department of Agriculture

Collection Method

Generated using Word2Vec, starting with vectors trained on Google News and further trained on text from PubAg and Ag Data Commons.

Time Range

null

Freshness

Last updated 2026-03-13 22:44:49.977678; freshness should be verified.

Geography

null