def Explain(self, text, ID):
"""Does the NLP on the String and insert the result in BigQuery"""
from sparknlp.base import LightPipeline
print("Explain... " + ID)
# In this case we use a Light-Pipeline. For furter information see:
# https://medium.com/spark-nlp/spark-nlp-101-lightpipeline-a544e93f20f1
# To set up the Pipeline the prevous downloaded Modell is needed
lp = LightPipeline(self.Model)
# No let the Pipeline annotate our text
r = lp.annotate(text)
# Create empty list
toret = []
# iterate through entities
for i in r["entities"]:
# and append it to the list with the corresponding ID
toret.append((i, ID))
# Check if list is not empty
if toret != []:
# Create for each item a row in BigQuery
errors = self.client.insert_rows(self.table, toret) # API request
# Print possible errors
if errors != []:
print(errors)
else:
print("No entities found")
def annotate(model_class, name, target, target_column=None):
if not model_class.model:
model_class.model = ResourceDownloader().downloadPipeline(name, "en")
if type(target) is pyspark.sql.dataframe.DataFrame:
if not target_column:
raise Exception("annotate() target_column arg needed when targeting a DataFrame")
return model_class.model.transform(target.withColumnRenamed(target_column, "text"))
elif type(target) is list or type(target) is str:
pip = LightPipeline(model_class.model)
return pip.annotate(target)
def main():
spark, sc = init_spark()
tweetsDf = spark.read \
.load("C:\\sparkTmp\\why_i_wear_mask_tweets.csv",
format="csv", sep=",", inferSchema="true",
header="true", charset="UTF-8") \
.select("text")
pipeline = simplePipeline()
pipelineModel = pipeline.fit(tweetsDf)
# result = pipelineModel.transform(tweetsDf)
# result.show()
lightModel = LightPipeline(pipelineModel, parse_embeddings=True)
print(
lightModel.annotate(
"How did serfdom develop in and then leave Russia ?"))
prediction_data.show()
prediction_model.transform(prediction_data).show(truncate=False)
prediction_model.write().overwrite().save("ner_dl_model")
!cp -r "ner_dl_model" "gdrive/My Drive/Colab Notebooks/SparkNLP/utils/ner_dl_model_base"
from pyspark.ml import PipelineModel, Pipeline
loaded_prediction_model = PipelineModel.read().load("ner_dl_model")
prediction_data = spark.createDataFrame([["Maria is a nice place."],["any bbq places open before 5 nearby"]]).toDF("text")
prediction_data.show()
#loaded_prediction_model.transform(prediction_data).show(truncate=False)
prediction = loaded_prediction_model.transform(prediction_data)
prediction.select("finished_ner_metadata").show(truncate=False)
prediction.select("finished_ner").show(truncate=False)
prediction.select("finished_ner_converter_metadata").show(truncate=False)
prediction.select("finished_ner_converter").show(truncate=False)
#prediction.select("ner").show(truncate=False)
from sparknlp.base import LightPipeline
lp = LightPipeline(loaded_prediction_model)
result = lp.annotate("Peter is a good person.")
for e in list(zip(result['token'], result['ner']))[:10]:
print(e)
for stage in loaded_prediction_model.stages:
print(stage)
print(loaded_prediction_model.stages[-1].stages)
