Model documentation & parameters

Model type: Type of PGT model to be used:

PGTGenerator: A model for part-of-patent generator.
PGTEditor: An algorithm for part-of-patent editing.
PGTCoherenceChecker: An algorithm for patent coherence check.

Generator task: Task in case the PGTGenerator model is used. Options are:

title-to-abstract
abstract-to-title
abstract-to-claim
claim-to-abstract

Editor task: Task in case the PGTEditor model is used. Options are:

abstract
claim

Coherence task: Task in case the PGTCoherenceChecker model is used. Options are:

title-abstract
title-claim
abstract-claim

Primary text prompt: The main text prompt for the model

Secondary text prompt: The secondary text prompt for the model (only used for PGTCoherenceChecker).

Maximal length: The maximal number of tokens in the generated sequences.

Top-k: Number of top-k probability tokens to keep.

Top-p: Only tokens with cumulative probabilities summing up to this value are kept.

Model card -- PatentGenerativeTransformer

Model Details: Patent Generative Transformer (PGT), a transformer-based multitask language model trained to facilitate the patent generation process. Published by Christofidellis et al. (ICML 2022 Workshop KRLM)

Developers: Dimitrios Christofidellis and colleagues at IBM Research.

Distributors: Model natively integrated into GT4SD.

Model date: 2022.

Model type:

PGTGenerator: A model for part-of-patent generator
PGTEditor: An algorithm for part-of-patent editing.
PGTCoherenceChecker: An algorithm for patent coherence check

Information about training algorithms, parameters, fairness constraints or other applied approaches, and features: N.A.

Paper or other resource for more information: The Patent Generative Transformer (PGT) paper by Christofidellis et al. (ICML 2022 Workshop KRLM).

License: MIT

Where to send questions or comments about the model: Open an issue on GT4SD repository.

Intended Use. Use cases that were envisioned during development: N.A.

Primary intended uses/users: N.A.

Out-of-scope use cases: Production-level inference, producing molecules with harmful properties.

Metrics: N.A.

Datasets: N.A.

Ethical Considerations: Unclear, please consult with original authors in case of questions.

Caveats and Recommendations: Unclear, please consult with original authors in case of questions.

Model card prototype inspired by Mitchell et al. (2019)

Citation

@inproceedings{christofidellis2022pgt,
  title={PGT: a prompt based generative transformer for the patent domain},
  author={Christofidellis, Dimitrios and Torres, Antonio Berrios and Dave, Ashish and Roveri, Manuel and Schmidt, Kristin and Swaminathan, Sarath and Vandierendonck, Hans and Zubarev, Dmitry and Manica, Matteo},
  booktitle={ICML 2022 Workshop on Knowledge Retrieval and Language Models},
  year={2022}
}