Diff of
/medacy/pipelines/systematic_review_pipeline.py
[000000]
..
[6c353a]
Switch to unified view
| a | b/medacy/pipelines/systematic_review_pipeline.py | ||
|---|---|---|---|
| 1 | import spacy |
||
| 2 | |||
| 3 | from medacy.pipeline_components.feature_extractors.discrete_feature_extractor import FeatureExtractor |
||
| 4 | from medacy.pipeline_components.feature_overlayers.metamap.metamap import MetaMap |
||
| 5 | from medacy.pipeline_components.feature_overlayers.metamap.metamap_component import MetaMapOverlayer |
||
| 6 | from medacy.pipeline_components.learners.crf_learner import get_crf |
||
| 7 | from medacy.pipeline_components.tokenizers.systematic_review_tokenizer import SystematicReviewTokenizer |
||
| 8 | from medacy.pipelines.base.base_pipeline import BasePipeline |
||
| 9 | |||
| 10 | |||
| 11 | class SystematicReviewPipeline(BasePipeline): |
||
| 12 | """ |
||
| 13 | A pipeline for clinical named entity recognition. This pipeline was designed over-top of the TAC 2018 SRIE track |
||
| 14 | challenge. |
||
| 15 | |||
| 16 | Created by Andriy Mulyar (andriymulyar.com) of NLP@VCU |
||
| 17 | """ |
||
| 18 | |||
| 19 | |||
| 20 | def __init__(self, entities, metamap=None, **kwargs): |
||
| 21 | """ |
||
| 22 | Create a pipeline with the name 'clinical_pipeline' utilizing |
||
| 23 | by default spaCy's small english model. |
||
| 24 | |||
| 25 | :param entities: a list of entities |
||
| 26 | :param metamap: an instance of MetaMap |
||
| 27 | """ |
||
| 28 | |||
| 29 | super().__init__(entities, spacy_pipeline=spacy.load("en_core_web_sm"), **kwargs) |
||
| 30 | |||
| 31 | if metamap: |
||
| 32 | metamap = MetaMap(metamap) |
||
| 33 | self.add_component(MetaMapOverlayer, metamap) |
||
| 34 | |||
| 35 | def get_learner(self): |
||
| 36 | return "CRF_l2sgd", get_crf() |
||
| 37 | |||
| 38 | def get_tokenizer(self): |
||
| 39 | return SystematicReviewTokenizer(self.spacy_pipeline) |
||
| 40 | |||
| 41 | def get_feature_extractor(self): |
||
| 42 | return FeatureExtractor(window_size=10, spacy_features=['pos_', 'shape_', 'prefix_', 'suffix_', 'text']) |