isag91 / histlm

Neural Language Models for Historical Research

histLM

Neural Language Models for Historical Research

Table of contents

• Language models
  • Download
  • Load models
• Language models in use
• Installation and setup
  • Method 1: Anaconda + install dependencies manually
• License

Language models

Download

We have pre-trained four types of neural language models trained on a large historical dataset of books in English, published between 1760-1900 and comprised of ~5.1 billion tokens. The language model architectures include static (word2vec and fastText) and contextualized models (BERT and Flair). For each architecture, we trained a model instance using the whole dataset. Additionally, we trained separate instances on text published before 1850 for the two static models, and four instances considering different time slices for BERT.

⚠️ The language models can be downloaded from zenodo. (see License)

Each .zip file on zenodo contains model instances for one neural network architecture (i.e., bert, flair, fasttext and word2vec). After unzipping the four .zip files, the directory structure is as follows:

histLM_dataset
├── README.md
├── bert
│   ├── bert_1760_1850
│   │   ├── config.json
│   │   ├── pytorch_model.bin
│   │   ├── special_tokens_map.json
│   │   ├── tokenizer_config.json
│   │   ├── training_args.bin
│   │   └── vocab.txt
│   ├── bert_1760_1900
│   |   └── ...
│   ├── bert_1850_1875
│   |   └── ...
│   ├── bert_1875_1890
│   |   └── ...
│   └── bert_1890_1900
│       └── ...
|
├── flair
│   └── flair_1760_1900
│       ├── best-lm.pt
│       ├── loss.txt
│       └── training.log
|
├── fasttext
│   ├── ft_1760_1850
│   │   ├── fasttext_words.model
│   │   ├── fasttext_words.model.trainables.syn1neg.npy
│   │   ├── fasttext_words.model.trainables.vectors_ngrams_lockf.npy
│   │   ├── fasttext_words.model.trainables.vectors_vocab_lockf.npy
│   │   ├── fasttext_words.model.wv.vectors.npy
│   │   ├── fasttext_words.model.wv.vectors_ngrams.npy
│   │   └── fasttext_words.model.wv.vectors_vocab.npy
│   └── ft_1760_1900
│       └── ...
|
└── word2vec
    ├── w2v_1760_1850
    │   ├── w2v_words.model
    │   ├── w2v_words.model.trainables.syn1neg.npy
    │   └── w2v_words.model.wv.vectors.npy
    └── w2v_1760_1900
        └── ...

Load models

After downloading the language models from zenodo (refer to Download section):

1. Go to histLM directory:

cd /path/to/histLM

2. Create a directory called histLM_dataset:

mkdir histLM_dataset

3. Move the unzipped directories to histLM. The directory structure should be:

histLM
├── README.md
├── histLM_dataset
│   ├── README.md
│   ├── bert
│   │   ├── bert_1760_1850
│   │   ├── bert_1760_1900
│   │   ├── bert_1850_1875
│   │   ├── bert_1875_1890
│   │   └── bert_1890_1900
│   ├── fasttext
│   │   ├── ft_1760_1850
│   │   └── ft_1760_1900
│   ├── flair
│   │   └── flair_1760_1900
│   └── word2vec
│       ├── w2v_1760_1850
│       └── w2v_1760_1900
└── notebooks
    ├── BERT_model.ipynb
    ├── Flair_model.ipynb
    ├── fastText_model.ipynb
    └── word2vec_model.ipynb

4. Finally, open one of the jupyter notebooks stored in the notebooks directory:

$ cd notebooks
$ jupyter notebook

Language models in use

So far, the language models presented in this repository have been used in the following projects:

• When Time Makes Sense: A Historically-Aware Approach to Targeted Sense Disambiguation (Findings of ACL 2021): repository and paper (forthcoming).
• Living Machines: A Study of Atypical Animacy (COLING 2020): repository and paper.
• Assessing the Impact of OCR Quality on Downstream NLP Tasks (ARTIDIGH 2020): repository and paper.

Installation

We strongly recommend installation via Anaconda:

• Refer to Anaconda website and follow the instructions.

Method 1

• Create a new environment for histLM called py38_histLM:

conda create -n py38_histLM python=3.8

• Activate the environment:

conda activate py38_histLM

• Clone histLM source code:

git clone https://github.com/Living-with-machines/histLM.git 

• Install dependencies:

pip install torch
pip install transformers
pip install flair
pip install gensim
pip install notebook
pip install jupyter-client
pip install jupyter-core
pip install ipywidgets

• To allow the newly created py38_histLM environment to show up in the notebooks:

python -m ipykernel install --user --name py38_histLM --display-name "Python (py38_histLM)"

License

Codes/notebooks are released under MIT License.

Models are released under open license CC BY 4.0, available at https://creativecommons.org/licenses/by/4.0/legalcode.