Description
This model is a BioBERT based classifier that can classify premise about health mandates related to Covid-19 from tweets. This model is intended for direct use as a classification model and the target classes are: Has no premise (no argument), Has premise (argument).
Predicted Entities
Has premise (argument), Has no premise (no argument)
Live Demo Open in Colab Copy S3 URI
How to use
document_assembler = DocumentAssembler() \
.setInputCol('text') \
.setOutputCol('document')
tokenizer = Tokenizer() \
.setInputCols(['document']) \
.setOutputCol('token')
sequenceClassifier = MedicalBertForSequenceClassification.pretrained("bert_sequence_classifier_health_mandates_premise_tweet", "en", "clinical/models")\
.setInputCols(["document",'token'])\
.setOutputCol("class")
pipeline = Pipeline(stages=[
document_assembler,
tokenizer,
sequenceClassifier
])
data = spark.createDataFrame(["""It's too dangerous to hold the RNC, but let's send students and teachers back to school.""",
"""So is the flu and pneumonia what are their s stop the Media Manipulation covid has treatments You're Speaker Pelosi nephew so stop the agenda LIES.""",
"""Just a quick update to my U.S. followers, I'll be making a stop in all 50 states this spring! No tickets needed, just don't wash your hands, cough on each other.""",
"""Go to a restaurant no mask Do a food shop wear a mask INCONSISTENT No Masks No Masks.""",
"""But if schools close who is gonna occupy those graves Cause politiciansprotected smokers protected drunkardsprotected school kids amp teachers""",
"""New title Maskhole I think I'm going to use this very soon coronavirus."""
], StringType()).toDF("text")
result = pipeline.fit(data).transform(data)
result.select("class.result", "text").show(truncate=False)
val documenter = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val tokenizer = new Tokenizer()
.setInputCols("sentences")
.setOutputCol("token")
val sequenceClassifier = MedicalBertForSequenceClassification.pretrained("bert_sequence_classifier_health_mandates_premise_tweet", "es", "clinical/models")
.setInputCols(Array("document","token"))
.setOutputCol("class")
val pipeline = new Pipeline().setStages(Array(documenter, tokenizer, sequenceClassifier))
val data = Seq(Array("It's too dangerous to hold the RNC, but let's send students and teachers back to school.",
"So is the flu and pneumonia what are their s stop the Media Manipulation covid has treatments You're Speaker Pelosi nephew so stop the agenda LIES.",
"Just a quick update to my U.S. followers, I'll be making a stop in all 50 states this spring! No tickets needed, just don't wash your hands, cough on each other.",
"Go to a restaurant no mask Do a food shop wear a mask INCONSISTENT No Masks No Masks.",
"But if schools close who is gonna occupy those graves Cause politiciansprotected smokers protected drunkardsprotected school kids amp teachers",
"New title Maskhole I think I'm going to use this very soon coronavirus.")).toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("en.clasify.health_premise").predict("""Just a quick update to my U.S. followers, I'll be making a stop in all 50 states this spring! No tickets needed, just don't wash your hands, cough on each other.""")
Results
+------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------+
|text |result |
+------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------+
|It's too dangerous to hold the RNC, but let's send students and teachers back to school. |[Has premise (argument)] |
|So is the flu and pneumonia what are their s stop the Media Manipulation covid has treatments You're Speaker Pelosi nephew so stop the agenda LIES. |[Has premise (argument)] |
|Just a quick update to my U.S. followers, I'll be making a stop in all 50 states this spring! No tickets needed, just don't wash your hands, cough on each other.|[Has no premise (no argument)]|
|Go to a restaurant no mask Do a food shop wear a mask INCONSISTENT No Masks No Masks. |[Has no premise (no argument)]|
|But if schools close who is gonna occupy those graves Cause politiciansprotected smokers protected drunkardsprotected school kids amp teachers |[Has premise (argument)] |
|New title Maskhole I think I'm going to use this very soon coronavirus. |[Has no premise (no argument)]|
+------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------+
Model Information
| Model Name: | bert_sequence_classifier_health_mandates_premise_tweet |
| Compatibility: | Healthcare NLP 4.0.2+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [document, token] |
| Output Labels: | [class] |
| Language: | en |
| Size: | 406.5 MB |
| Case sensitive: | true |
| Max sentence length: | 128 |
References
The dataset is Covid-19-specific and consists of tweets collected via a series of keywords associated with that disease.
Benchmarking
label precision recall f1-score support
Has_no_premise_(no_argument) 0.9000 0.6992 0.7870 502
Has_premise_(argument) 0.6576 0.8815 0.7532 329
accuracy - - 0.7714 831
macro-avg 0.7788 0.7903 0.7701 831
weighted-avg 0.8040 0.7714 0.7736 831