orkgnlp.annotation.rfclf.classifier.ResearchFieldClassifier

class ResearchFieldClassifier(*args, **kwargs)[source]

Bases: ORKGNLPBaseService

The ResearchFieldClassifier requires a torch-script model trained on abstracts and a label dictionary file containing research fields and corresponding indices. The required files are downloaded while initiation, if it has not happened before

You can pass the parameter force_download=True to remove and re-download the previous downloaded service files.

Parameters:

  • service – Service name.

  • force_download – Indicates whether the required files are to be downloaded again. Defaults to False.

  • batch_size – Size of the batches used during model prediction. This argument is used by services that applies batch evaluation. Defaults to 16.

Methods

release_memory

Releases the memory of all available executors.

Attributes

SERVICE_NAME

__call__(raw_input, top_n=5)[source]

Classifies a paper into a research field by processing the given abstract.

Parameters:

  • raw_input (str) – Combined paper’s title and abstract

  • top_n (int) – The top n research fields to be retrieved

Return type:

Any

Returns:

A list of n most likely research fields for the paper

Raises:

orkgnlp.common.util.exceptions.ORKGNLPValidationError – If no abstract is given

release_memory()

Releases the memory of all available executors.