Difference between revisions of "Publications/abadie.22.das"

From LRDE

(Created page with "{{Publication | published = true | date = 2022-04-07 | authors = Nathalie Abadie, Edwin Carlinet, Joseph Chazalon, Bertrand Duménieu | title = A Benchmark of Named Entity Rec...")
 
 
(4 intermediate revisions by the same user not shown)
Line 6: Line 6:
 
| booktitle = Proceedings of the 15th IAPR International Workshop on Document Analysis System
  
| booktitle = Proceedings of the 15th IAPR International Workshop on Document Analysis System
 
| address = La Rochelle, France
  
| address = La Rochelle, France
−
| abstract = Named entity recognition (NER) is a necessary step in many pipelines targeting historical documents. Indeed, such natural language processing techniques identify which class each text token belongs toe.g. “person name”, “location”, “number”. Introducing a new public dataset built from 19th century French directories, we first assess how noisy modern, off-the-shelf OCR are. Then, we compare modern CNN- and Transformer-based NER techniques which can be reasonably used in the context of historical document analysis. We measure their requirements in terms of training data, the effects of OCR noise on their performance, and show how Transformer-based NER can benefit from unsupervised pre-training and supervised fine-tuning on noisy data. Results can be reproduced using resources available at https://github.com/soduco/paper-ner-bench-das22 and https://zenodo.org/record/6394464
 +
| abstract = Named entity recognition (NER) is a necessary step in many pipelines targeting historical documents. Indeed, such natural language processing techniques identify which class each text token belongs to, e.g. “person name”“location”, “number”. Introducing a new public dataset built from 19th century French directories, we first assess how noisy modern, off-the-shelf OCR are. Then, we compare modern CNN- and Transformer-based NER techniques which can be reasonably used in the context of historical document analysis. We measure their requirements in terms of training data, the effects of OCR noise on their performance, and show how Transformer-based NER can benefit from unsupervised pre-training and supervised fine-tuning on noisy data. Results can be reproduced using resources available at https://github.com/soduco/paper-ner-bench-das22 and https://zenodo.org/record/6394464
  +
| series = Lecture Notes in Computer Science
  +
| volume = 13237
  +
| pages = 445 to 460
  +
| publisher = Springer
 
| lrdenewsdate = 2022-04-07
  
| lrdenewsdate = 2022-04-07
  +
| lrdepaper = http://www.lrde.epita.fr/dload/papers/abadie.22.das.pdf
  +
| lrdeprojects = Olena
  +
| lrdekeywords = Image
  +
| lrderank = B
 
| type = inproceedings
  
| type = inproceedings
 
| id = abadie.22.das
  
| id = abadie.22.das
  +
| identifier = doi:10.1007/978-3-031-06555-2_30
 
| bibtex =
 
| bibtex =
 
@InProceedings<nowiki>{</nowiki> abadie.22.das,
  
@InProceedings<nowiki>{</nowiki> abadie.22.das,
Line 20: Line 29:
 
year = 2022,
  
year = 2022,
 
address = <nowiki>{</nowiki>La Rochelle, France<nowiki>}</nowiki>,
  
address = <nowiki>{</nowiki>La Rochelle, France<nowiki>}</nowiki>,
−
month = may,
 +
month = 5,
 
abstract = <nowiki>{</nowiki>Named entity recognition (NER) is a necessary step in many
  
abstract = <nowiki>{</nowiki>Named entity recognition (NER) is a necessary step in many
 
pipelines targeting historical documents. Indeed, such
  
pipelines targeting historical documents. Indeed, such
Line 37: Line 46:
 
available at
  
available at
 
https://github.com/soduco/paper-ner-bench-das22 and
  
https://github.com/soduco/paper-ner-bench-das22 and
−
https://zenodo.org/record/6394464<nowiki>}</nowiki>
 +
https://zenodo.org/record/6394464<nowiki>}</nowiki>,
  +
series = <nowiki>{</nowiki>Lecture Notes in Computer Science<nowiki>}</nowiki>,
  +
volume = 13237,
  +
pages = <nowiki>{</nowiki>445--460<nowiki>}</nowiki>,
  +
publisher = <nowiki>{</nowiki>Springer<nowiki>}</nowiki>,
  +
doi = <nowiki>{</nowiki>10.1007/978-3-031-06555-2_30<nowiki>}</nowiki>,
  +
lrderank = <nowiki>{</nowiki>B<nowiki>}</nowiki>
 
<nowiki>}</nowiki>
  
<nowiki>}</nowiki>
   

Latest revision as of 13:47, 30 August 2023

Authors
Nathalie Abadie, Edwin Carlinet, Joseph Chazalon, Bertrand Duménieu
Where
Proceedings of the 15th IAPR International Workshop on Document Analysis System
Place
La Rochelle, France
Type
inproceedings
Publisher
Springer
Projects
Olena
Keywords
Image
Date
2022-04-07

Abstract

Named entity recognition (NER) is a necessary step in many pipelines targeting historical documents. Indeed, such natural language processing techniques identify which class each text token belongs to, e.g. “person name”“location”, “number”. Introducing a new public dataset built from 19th century French directories, we first assess how noisy modern, off-the-shelf OCR are. Then, we compare modern CNN- and Transformer-based NER techniques which can be reasonably used in the context of historical document analysis. We measure their requirements in terms of training data, the effects of OCR noise on their performance, and show how Transformer-based NER can benefit from unsupervised pre-training and supervised fine-tuning on noisy data. Results can be reproduced using resources available at https://github.com/soduco/paper-ner-bench-das22 and https://zenodo.org/record/6394464

Documents

Bibtex (lrde.bib)

@InProceedings{	  abadie.22.das,
  author	= {Nathalie Abadie and Edwin Carlinet and Joseph Chazalon and
		  Bertrand Dum\'enieu},
  title		= {A Benchmark of Named Entity Recognition Approaches in
		  Historical Documents},
  booktitle	= {Proceedings of the 15th IAPR International Workshop on
		  Document Analysis System},
  year		= 2022,
  address	= {La Rochelle, France},
  month		= 5,
  abstract	= {Named entity recognition (NER) is a necessary step in many
		  pipelines targeting historical documents. Indeed, such
		  natural language processing techniques identify which class
		  each text token belongs to, e.g. ``person name'',
		  ``location'', ``number''. Introducing a new public dataset
		  built from 19th century French directories, we first assess
		  how noisy modern, off-the-shelf OCR are. Then, we compare
		  modern CNN- and Transformer-based NER techniques which can
		  be reasonably used in the context of historical document
		  analysis. We measure their requirements in terms of
		  training data, the effects of OCR noise on their
		  performance, and show how Transformer-based NER can benefit
		  from unsupervised pre-training and supervised fine-tuning
		  on noisy data. Results can be reproduced using resources
		  available at
		  https://github.com/soduco/paper-ner-bench-das22 and
		  https://zenodo.org/record/6394464},
  series	= {Lecture Notes in Computer Science},
  volume	= 13237,
  pages		= {445--460},
  publisher	= {Springer},
  doi		= {10.1007/978-3-031-06555-2_30},
  lrderank	= {B}
}





Retrieved from "https://www.lrde.epita.fr/index.php?title=Publications/abadie.22.das&oldid=128713"