Description
This model was imported from Hugging Face (link) and it’s been trained on tasosk/airlines dataset, leveraging Distil-BERT embeddings and DistilBertForSequenceClassification for text classification purposes. The model classifies texts into two categories: YES for positive comments, and NO for negative.
Predicted Entities
YES, NO
How to use
document_assembler = DocumentAssembler() \
.setInputCol('text') \
.setOutputCol('document')
tokenizer = Tokenizer() \
.setInputCols(['document']) \
.setOutputCol('token')
sequenceClassifier = DistilBertForSequenceClassification\
.pretrained('distilbert_base_sequence_classifier_airlines', 'en') \
.setInputCols(['token', 'document']) \
.setOutputCol('class')
pipeline = Pipeline(stages=[document_assembler, tokenizer, sequenceClassifier])
example = spark.createDataFrame([["Jersey to London Gatwick with easyJet and another great flight. Due to the flight time, airport check-in was not open, however I'd checked in a few days before with the easyJet app which was very quick and convenient. Boarding was quick and we left a few minutes early, which is a bonus. The cabin crew were friendly and the aircraft was clean and comfortable. We arrived at Gatwick 5-10 minutes early, and disembarking was as quick as boarding. On the way back, we were about half an hour early landing, which was fantastic. For the short flight from JER-LGW, easyJet are ideal and a bit better than British Airways in my opinion, and the fares are just unmissable. Both flights for two adults cost £180. easyJet can expect my business in the near future."]]).toDF("text")
result = pipeline.fit(example).transform(example)
val document_assembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val tokenizer = new Tokenizer()
.setInputCols(Array("document"))
.setOutputCol("token")
val tokenClassifier = DistilBertForSequenceClassification.pretrained("distilbert_base_sequence_classifier_airlines", "en")
.setInputCols(Array("document", "token"))
.setOutputCol("class")
val pipeline = new Pipeline().setStages(Array(document_assembler, tokenizer, sequenceClassifier))
val example = Seq("Jersey to London Gatwick with easyJet and another great flight. Due to the flight time, airport check-in was not open, however I'd checked in a few days before with the easyJet app which was very quick and convenient. Boarding was quick and we left a few minutes early, which is a bonus. The cabin crew were friendly and the aircraft was clean and comfortable. We arrived at Gatwick 5-10 minutes early, and disembarking was as quick as boarding. On the way back, we were about half an hour early landing, which was fantastic. For the short flight from JER-LGW, easyJet are ideal and a bit better than British Airways in my opinion, and the fares are just unmissable. Both flights for two adults cost £180. easyJet can expect my business in the near future.").toDF("text")
val result = pipeline.fit(example).transform(example)
import nlu
nlu.load("en.classify.distil_bert.base.by_datasets").predict("""Jersey to London Gatwick with easyJet and another great flight. Due to the flight time, airport check-in was not open, however I'd checked in a few days before with the easyJet app which was very quick and convenient. Boarding was quick and we left a few minutes early, which is a bonus. The cabin crew were friendly and the aircraft was clean and comfortable. We arrived at Gatwick 5-10 minutes early, and disembarking was as quick as boarding. On the way back, we were about half an hour early landing, which was fantastic. For the short flight from JER-LGW, easyJet are ideal and a bit better than British Airways in my opinion, and the fares are just unmissable. Both flights for two adults cost £180. easyJet can expect my business in the near future.""")
Results
['YES']
Model Information
| Model Name: | distilbert_base_sequence_classifier_airlines |
| Compatibility: | Spark NLP 3.4.0+ |
| License: | Open Source |
| Edition: | Official |
| Input Labels: | [sentence, token] |
| Output Labels: | [class] |
| Language: | en |
| Size: | 249.8 MB |
| Case sensitive: | true |
| Max sentence length: | 256 |
References
https://huggingface.co/datasets/tasosk/airlines
Benchmarking
label score
accuracy 0.9288
f1 0.9289