ModernBERT-base-literary-NER-novelties

ModernBERT-base-literary-NER-novelties is NER model based on ModernBERT-base and trained on the Novelties literary dataset for usage in the Renard library.

Performance

Book	Precision	Recall	F1
Ulysses	89.23	85.50	87.33
The Way we Live Now	78.66	82.17	80.37
Tinker Tailor Soldier Spy	82.19	83.54	82.86
The Wheel of Time	88.63	84.68	86.61
Vanity Fair	88.57	92.38	90.44

Class	F1
CHR	92.53
GRP	68.47
LOC	81.97
MSC	37.76
ORG	37.14

Usage in Renard

from renard.pipeline import Pipeline
from renard.pipeline.tokenization import NLTKTokenizer
from renard.pipeline.ner import BertNamedEntityRecognizer
from renard.pipeline.character_unification import GraphRulesCharacterUnifier
from renard.pipeline.graph_extraction import CoOccurrencesGraphExtractor

pipeline = Pipeline(
    [
        NLTKTokenizer(),
        BertNamedEntityRecognizer(
            model="compnet-renard/ModernBERT-base-literary-NER-novelties"
        ),
        GraphRulesCharacterUnifier(),
        CoOccurrencesGraphExtractor(co_occurrences_dist=10),
    ],
    # by default, a Renard pipeline uses "PER" as the tag to recognize characters,
    # but this model is trained to ouput "CHR" tags instead.
    character_ner_tag="CHR"
)

out = pipeline("Princess Liana felt sad, because Zarth Arn was gone. The princess went to sleep.")
print((out.character_network.nodes, out.character_network.edges))

This outputs:

(NodeView((<Princess Liana, Gender.UNKNOWN, 1 mentions>, <Zarth Arn, Gender.UNKNOWN, 1 mentions>)), EdgeView([(<Princess Liana, Gender.UNKNOWN, 1 mentions>, <Zarth Arn, Gender.UNKNOWN, 1 mentions>)]))