def setup(args):
"""
Sets up dataset and pipeline/model since it gets used by every command.
:param args: Argparse args object.
:return dataset, model: The dataset and model objects created.
"""
dataset = Dataset(args.dataset)
entities = list(dataset.get_labels())
if args.test_mode:
dataset.data_limit = 1
if args.entities is not None:
with open(args.entities, 'rb') as f:
data = json.load(f)
json_entities = data['entities']
if not set(json_entities) <= set(entities):
raise ValueError(
f"The following entities from the json file are not in the provided dataset: {set(json_entities) - set(entities)}"
)
entities = json_entities
if args.custom_pipeline is not None:
logging.info(
f"Using custom pipeline configured at {args.custom_pipeline}")
# Construct a pipeline class (not an instance) based on the provided json path;
# args.custom_pipeline is that path
Pipeline = json_to_pipeline(args.custom_pipeline)
else:
# Parse the argument as a class name in module medacy.pipelines
module = importlib.import_module("medacy.pipelines")
Pipeline = getattr(module, args.pipeline)
logging.info('Using %s', args.pipeline)
pipeline = Pipeline(entities=entities,
cuda_device=args.cuda,
word_embeddings=args.word_embeddings,
batch_size=args.batch_size,
learning_rate=args.learning_rate,
epochs=args.epochs,
pretrained_model=args.pretrained_model,
using_crf=args.using_crf)
model = Model(pipeline)
return dataset, model
def setup(args):
"""
Sets up dataset and pipeline/model since it gets used by every command.
:param args: Argparse args object.
:return dataset, model: The dataset and model objects created.
"""
dataset = Dataset(args.dataset)
entities = list(dataset.get_labels())
pipeline = None
if args.pipeline == 'spacy':
logging.info('Using spacy model')
model = SpacyModel(spacy_model_name=args.spacy_model, cuda=args.cuda)
elif args.custom_pipeline is not None:
# Construct a pipeline class (not an instance) based on the provided json path;
# args.custom_pipeline is that path
Pipeline = json_to_pipeline(args.custom_pipeline)
# All parameters are part of the class, thus nothing needs to be set when instantiating
pipeline = Pipeline()
model = Model(pipeline)
else:
# Parse the argument as a class name in module medacy.pipelines
module = importlib.import_module("medacy.pipelines")
Pipeline = getattr(module, args.pipeline)
logging.info('Using %s', args.pipeline)
pipeline = Pipeline(entities=entities,
cuda_device=args.cuda,
word_embeddings=args.word_embeddings,
batch_size=args.batch_size,
learning_rate=args.learning_rate,
epochs=args.epochs,
pretrained_model=args.pretrained_model,
using_crf=args.using_crf)
model = Model(pipeline)
return dataset, model
import argparse
import glob
import os
from collections import defaultdict
from xml.etree import cElementTree
from medacy.data.dataset import Dataset
# Setup
parser = argparse.ArgumentParser(description='n2c2: Evaluation script for Track 2')
parser.add_argument('folder1', help='First data folder path (gold)')
parser.add_argument('folder2', help='Second data folder path (system)')
args = parser.parse_args()
gold_dataset = Dataset(args.folder1)
prediction_dataset = Dataset(args.folder2)
global_tags = tuple(gold_dataset.get_labels() & prediction_dataset.get_labels())
class ClinicalCriteria(object):
"""Criteria in the Track 1 documents."""
def __init__(self, tid, value):
"""Init."""
self.tid = tid.strip().upper()
self.ttype = self.tid
self.value = value.lower().strip()
def equals(self, other, mode='strict'):
"""Return whether the current criteria is equal to the one provided."""
if other.tid == self.tid and other.value == self.value:
return True
