language: fr
tags:
- Early Modern French
- Historical
license: apache-2.0
datasets:
- freemmax
D'AlemBERT base model
This model is a RoBERTa base model pre-trained on the FreEMmax corpus for Early Modern French. It was introduced in this paper. This model is Cased and was trained with a mix of normalized and unnormalized data.
Model description
DistilBERT is a transformers model, smaller and faster than BERT, which was pretrained on the same corpus in a self-supervised fashion, using the BERT base model as a teacher. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts using the BERT base model. More precisely, it was pretrained with three objectives:
- Distillation loss: the model was trained to return the same probabilities as the BERT base model.
- Masked language modeling (MLM): this is part of the original training loss of the BERT base model. When taking a sentence, the model randomly masks 15% of the words in the input then run the entire masked sentence through the model and has to predict the masked words. This is different from traditional recurrent neural networks (RNNs) that usually see the words one after the other, or from autoregressive models like GPT which internally mask the future tokens. It allows the model to learn a bidirectional representation of the sentence.
- Cosine embedding loss: the model was also trained to generate hidden states as close as possible as the BERT base model.
This way, the model learns the same inner representation of the English language than its teacher model, while being faster for inference or downstream tasks.
Intended uses & limitations
You can use the raw model for masked language modeling, but it's mostly intended to be fine-tuned on a downstream task.
Note that this model is primarily aimed at being fine-tuned on tasks that use the whole sentence (potentially masked) to make decisions, such as sequence classification, token classification or question answering. For tasks such as text generation you should look at model like GPT.
The model is primarily intended for use in Digital Humanities and Historical NLP.
Limitations and bias
This model is trained with historical French data from starting from the 16th c., so it might produce results that seem extremely biased by today standards. It might not work well on contemporary data and it is not intended to be used on it.
This bias will also affect all fine-tuned versions of this model.
Training data
D'AlemBERT was pretrained on the non-freely available version of the FreEMmax corpus, a dataset consisting of more than 180k tokens coming from 22 different sources, and comprising French textual data going from the 16th c to the early 20th c.
BibTeX entry and citation info
@inproceedings{gabay-etal-2022-freem,
title = "From {F}re{EM} to D{'}{A}lem{BERT}: a Large Corpus and a Language Model for Early {M}odern {F}rench",
author = "Gabay, Simon and
Ortiz Suarez, Pedro and
Bartz, Alexandre and
Chagu{\'e}, Alix and
Bawden, Rachel and
Gambette, Philippe and
Sagot, Beno{\^\i}t",
booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
month = jun,
year = "2022",
address = "Marseille, France",
publisher = "European Language Resources Association",
url = "https://aclanthology.org/2022.lrec-1.359",
pages = "3367--3374",
abstract = "anguage models for historical states of language are becoming increasingly important to allow the optimal digitisation and analysis of old textual sources. Because these historical states are at the same time more complex to process and more scarce in the corpora available, this paper presents recent efforts to overcome this difficult situation. These efforts include producing a corpus, creating the model, and evaluating it with an NLP task currently used by scholars in other ongoing projects.",
}