astroBERT / README.md
fgrezes's picture
SciX Categorizer tutorial and updated readme
3bf62ed
metadata
license: mit
language:
  - en
task_categories:
  - fill-mask
task_ids:
  - masked-language-modeling
pipeline_tag: fill-mask
widget:
  - text: M67 is one of the most studied [MASK] clusters.
    example_title: M67
  - text: >-
      A solar twin is a star with [MASK] parameters and chemical composition
      very similar to our Sun.
    example_title: solar twin
  - text: >-
      The dynamical evolution of planets close to their star is affected by
      [MASK] effects
    example_title: dynamical evolution
  - text: >-
      The Kepler satellite collected high-precision long-term and continuous
      light [MASK] for more than 100,000 solar-type stars
    example_title: Kepler satellite
  - text: >-
      The Local Group is composed of the Milky Way, the [MASK] Galaxy, and
      numerous smaller satellite galaxies.
    example_title: Local Group
  - text: >-
      Cepheid variables are used to determine the [MASK] to galaxies in the
      local universe.
    example_title: Cepheid
  - text: >-
      Jets are created and sustained by [MASK] of matter onto a compact massive
      object.
    example_title: Jets
  - text: A single star of one solar mass will evolve into a [MASK] dwarf.
    example_title: single star
  - text: The Very Large Array observes the sky at [MASK] wavelengths.
    example_title: Very Large Array
  - text: Elements heavier than [MASK] are generated in supernovae explosions.
    example_title: Elements
  - text: Spitzer was the first [MASK] to fly in an Earth-trailing orbit.
    example_title: Spitzer
  - text: Galaxy [MASK] can occur when two (or more) galaxies collide
    example_title: galaxies collide
  - text: >-
      Dark [MASK] is a hypothetical form of matter thought to account for
      approximately 85% of the matter in the universe.
    example_title: hypothetical matter
  - text: >-
      The cosmic microwave background (CMB, CMBR), in Big Bang cosmology, is
      electromagnetic radiation which is a remnant from an early stage of the
      [MASK].
    example_title: CMBR
  - text: The Local Group of galaxies is pulled toward The Great [MASK].
    example_title: galaxies pulled
  - text: The Moon is the only [MASK] of the Earth.
    example_title: Moon
  - text: >-
      Galaxies are categorized according to their visual morphology as [MASK],
      spiral, or irregular.
    example_title: morphology
  - text: Stars are made mostly of [MASK].
    example_title: Stars moslyl
  - text: Comet tails are created as comets approach the [MASK].
    example_title: Comet tails
  - text: Pluto is a dwarf [MASK] in the Kuiper Belt.
    example_title: Pluto
  - text: >-
      The Large and Small Magellanic Clouds are irregular [MASK] galaxies and
      are two satellite galaxies of the Milky Way.
    example_title: Magellanic Clouds
  - text: The Milky Way has a [MASK] black hole, Sagittarius A*, at its center.
    example_title: Milky Way
  - text: >-
      Andromeda is the nearest large [MASK] to the Milky Way and is roughly its
      equal in mass.
    example_title: Andromeda
  - text: The [MASK] medium is the gas and dust between stars.
    example_title: gast and dust

astroBERT: a language model for astrophysics

This public repository contains the work of the NASA/ADS on building an NLP language model tailored to astrophysics, along with tutorials and miscellaneous related files. This model is cased (it treats ads and ADS differently).

astroBERT models

  1. Base model: Pretrained model on English language using a masked language modeling (MLM) and next sentence prediction (NSP) objective. It was introduced in this paper at ADASS 2021 and made public at ADASS 2022.
  2. NER-DEAL model: This model adds a token classification head to the base model finetuned on the DEAL@WIESP2022 named entity recognition task. Must be loaded from the revision='NER-DEAL' branch (see tutorial 2).
  3. SciX Categorizer: This model was finetuned to classify text into one of 7 categories of interest to SciX (Astronomy, Heliophysics, Planetary Science, Earth Science, NASA-funded Biophysics, Other Physics, Other, Text Garbage).

Tutorials

  1. generate text embedding (for downstream tasks)
  2. use astroBERT for the Fill-Mask task
  3. make NER-DEAL predictions
  4. categorize texts for SciX

BibTeX

@ARTICLE{2021arXiv211200590G,
       author = {{Grezes}, Felix and {Blanco-Cuaresma}, Sergi and {Accomazzi}, Alberto and {Kurtz}, Michael J. and {Shapurian}, Golnaz and {Henneken}, Edwin and {Grant}, Carolyn S. and {Thompson}, Donna M. and {Chyla}, Roman and {McDonald}, Stephen and {Hostetler}, Timothy W. and {Templeton}, Matthew R. and {Lockhart}, Kelly E. and {Martinovic}, Nemanja and {Chen}, Shinyi and {Tanner}, Chris and {Protopapas}, Pavlos},
        title = "{Building astroBERT, a language model for Astronomy \& Astrophysics}",
      journal = {arXiv e-prints},
     keywords = {Computer Science - Computation and Language, Astrophysics - Instrumentation and Methods for Astrophysics},
         year = 2021,
        month = dec,
          eid = {arXiv:2112.00590},
        pages = {arXiv:2112.00590},
archivePrefix = {arXiv},
       eprint = {2112.00590},
 primaryClass = {cs.CL},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2021arXiv211200590G},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}