DESCRIPTION

Package: golgotha
Type: Package
Title: Contextualised Embeddings and Language Modelling using BERT
Version: 0.2.0
Maintainer: Jan Wijffels <jwijffels@bnosac.be>
Authors@R: c(
    person('Jan', 'Wijffels', role = c('aut', 'cre', 'cph'), email = 'jwijffels@bnosac.be'), 
    person('BNOSAC', role = 'cph'))
Description: Wraps the transformers Python module described in 
    "HuggingFace's Transformers: State-of-the-art Natural Language Processing" by Wolf T. et al. (2020) <arXiv:1910.03771> in order to easily obtain contextualised embeddings of sentences.
    This work was done in order to ease the work of building predictive models using BERT-like embeddings.
    BERT stands for "Bidirectional Encoder Representations from Transformers" as described in Devlin J. et al. (2018) <arXiv:1810.04805>. 
    It is a deep learning model which provides for pieces of words in a text a set of numbers called embeddings. These embedding
    capture the meaning of the words and depend on the context in which each word appears.
    The package provides an interface to BERT-like "Transformer" models such that R users 
    can just download these pretrained models and retrieve the embeddings of text with a straightforward call to predict.
License: MPL-2.0
Encoding: UTF-8
Archs: x64
SystemRequirements: Python (>= 2.7.0)
Suggests: 
    tinytest
Depends:
    reticulate (>= 1.14)
Config/reticulate:
  list(
    packages = list(
      list(package = "torch", pip = TRUE),
      list(package = "transformers", version = "2.4.1", pip = TRUE)
    )
  )    
RoxygenNote: 7.1.0