Description
The Violence and Abuse classifier employs MedicalBertForSequenceClassification embeddings within a robust classifier architecture. Trained on a diverse dataset, this model provides accurate label assignments and confidence scores for its predictions. The primary goal of this model is to categorize text into four key labels: Domestic_Violence_Abuse, Personal_Violence_Abuse, No_Violence_Abuse and Unknown.
-
Domestic_Violence_Abuse:This category refers to a pattern of behavior in any relationship that is aimed at gaining or maintaining power and control over an intimate partner or family member. -
Personal_Violence_Abuse: This category encompasses any form of violence or abuse that is directed towards an individual, whether admitted by the perpetrator or recognized by the victim. -
No_Violence_Abuse: This category denotes the complete absence of violence and abuse in any form. -
Unknown: This category covers when the nature or type of violence or abuse within a given text cannot be clearly identified or defined.
Predicted Entities
Domestic_Violence_Abuse, Personal_Violence_Abuse, No_Violence_Abuse, Unknown
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_sdoh_violence_abuse", "en", "clinical/models")\
.setInputCols(["document","token"])\
.setOutputCol("prediction")
pipeline = Pipeline(
stages=[
document_assembler,
tokenizer,
sequenceClassifier
])
sample_texts = [
["Repeated visits for fractures, with vague explanations suggesting potential family-related trauma."],
["Patient presents with multiple bruises in various stages of healing, suggestive of repeated physical abuse."],
["There are no reported instances or documented episodes indicating the patient poses a risk of violence."] ,
["Patient B is a 40-year-old female who was diagnosed with breast cancer. She has received a treatment plan that includes surgery, chemotherapy, and radiation therapy."]
]
sample_data = spark.createDataFrame(sample_texts).toDF("text")
result = pipeline.fit(sample_data).transform(sample_data)
result.select("text", "prediction.result").show(truncate=100)
val documenter = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val tokenizer = new Tokenizer()
.setInputCols("document")
.setOutputCol("token")
val sequenceClassifier = MedicalBertForSequenceClassification.pretrained("bert_sequence_classifier_sdoh_violence_abuse", "en", "clinical/models")
.setInputCols(Array("document","token"))
.setOutputCol("prediction")
val pipeline = new Pipeline().setStages(Array(documenter, tokenizer, sequenceClassifier))
val data = Seq(Array("Repeated visits for fractures, with vague explanations suggesting potential family-related trauma.",
"Patient presents with multiple bruises in various stages of healing, suggestive of repeated physical abuse.",
"There are no reported instances or documented episodes indicating the patient poses a risk of violence." ,
"Patient B is a 40-year-old female who was diagnosed with breast cancer. She has received a treatment plan that includes surgery, chemotherapy, and radiation therapy.",
)).toDF("text")
val result = pipeline.fit(data).transform(data)
Results
+----------------------------------------------------------------------------------------------------+-------------------------+
| text| result|
+----------------------------------------------------------------------------------------------------+-------------------------+
| Repeated visits for fractures, with vague explanations suggesting potential family-related trauma.|[Domestic_Violence_Abuse]|
|Patient presents with multiple bruises in various stages of healing, suggestive of repeated physi...|[Personal_Violence_Abuse]|
|There are no reported instances or documented episodes indicating the patient poses a risk of vio...| [No_Violence_Abuse]|
|Patient B is a 40-year-old female who was diagnosed with breast cancer. She has received a treatm...| [Unknown]|
+----------------------------------------------------------------------------------------------------+-------------------------+
Model Information
| Model Name: | bert_sequence_classifier_sdoh_violence_abuse |
| Compatibility: | Healthcare NLP 5.1.4+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [document, token] |
| Output Labels: | [prediction] |
| Language: | en |
| Size: | 406.4 MB |
| Case sensitive: | false |
| Max sentence length: | 512 |
References
Trained with the in-house dataset
Benchmarking
label precision recall f1-score support
Domestic_Violence_Abuse 0.921687 0.905325 0.913433 169
No_Violence_Abuse 0.978417 0.860759 0.915825 158
Personal_Violence_Abuse 0.889908 0.858407 0.873874 226
Unknown 0.937500 0.975610 0.956175 738
accuracy - - 0.931836 1291
macro-avg 0.931878 0.900025 0.914827 1291
weighted-avg 0.932106 0.931836 0.931234 1291