Citation Parsing (NER)

The Citation Parsing (NER) model utilizes advanced Named Entity Recognition (NER) to extract key fields from citation texts. This model parses citations into structured data fields such as TITLE, AUTHORS, VOLUME, ISSUE, YEAR, DOI, ISSN, ISBN, FIRST_PAGE, LAST_PAGE, JOURNAL, and EDITOR.

Overview

• Model type: Language Model
• Architecture: DistilBERT
• Language: Multilingual
• License: Apache 2.0
• Task: Named Entity Recognition (NER) for Citation Parsing
• Dataset: Custom Citation Parsing Dataset
• Additional Resources:
  • GitHub

Model description

The Citation Parsing (NER) model is part of the Citation Parser package. It is fine-tuned for extracting structured information from citation texts into the following key fields:

• TITLE
• AUTHORS
• VOLUME
• ISSUE
• YEAR
• DOI
• ISSN
• ISBN
• FIRST_PAGE
• LAST_PAGE
• JOURNAL
• EDITOR

This model was trained using the DistilBERT-base-multilingual-cased architecture, making it capable of processing multilingual citation data.

Intended Usage

This model is designed for extracting citation information and parsing raw citation text into structured fields. It is ideal for automating citation metadata extraction in academic databases, manuscript workflows, or citation analysis tools.

How to use

from transformers import pipeline

# Load the model
citation_parser = pipeline("ner", model="SIRIS-Lab/citation-parser-ENTITY")

# Example citation text
citation_text = "MURAKAMI, H等: 'Unique thermal behavior of acrylic PSAs bearing long alkyl side groups and crosslinked by aluminum chelate', 《EUROPEAN POLYMER JOURNAL》"

# Parse the citation
result = citation_parser(citation_text)
print(result)

Training

The model was trained using the SIRIS-Lab/citation-parser-ENTITY dataset consisting of:

• Training data: 2419 samples
• Test data: 269 samples

The following hyperparameters were used for training:

• Base Model: distilbert/distilbert-base-multilingual-cased
• Batch Size: 16
• Number of Epochs: 10
• Learning Rate: 2e-5
• Weight Decay: 0.01
• Max Sequence Length: 512

Evaluation Metrics

The model's performance was evaluated on the test set, and the following results were obtained:

Class-wise Evaluation Metrics:

Additional Information

Authors

SIRIS Lab, Research Division of SIRIS Academic.

License

This work is distributed under an Apache License, Version 2.0.

Contact

For further information, send an email to either nicolau.duransilva@sirisacademic.com or info@sirisacademic.com.