Publication Metadata Extraction Pipeline

The publication extraction pipeline is used to extract publication metadata from web pages. It consists of two modules:

• A PublicationExtractor responsible for extracting the publication strings located within web pages. Everything related to this module is located in /publication_extraction.
• A MetadataExtractor responsible for extracting various metadata fields located within publication strings. Everything related to this module is located in /metadata_extraction

The pipeline itself is represented by the Pipeline class.

Prerequisites

• Some models assume that pre-trained GloVe word embeddings are located in data/glove/.
• Some models assume that the HomePub dataset is located in data/homepub-2500/.
• Some models assume that the UMass Citation Field Extraction Dataset is located in data/umass/.

The glove.sh, homepub.sh, and umass.sh scripts can be used to download these.