Description
Automatically identify Joy, Surprise, Fear, Sadness emotions in Tweets.
Predicted Entities
surprise, sadness, fear, joy
Live Demo Open in Colab Download
How to use
document_assembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
use = UniversalSentenceEncoder.pretrained('tfhub_use', lang="en") \
.setInputCols(["document"])\
.setOutputCol("sentence_embeddings")
classifier = ClassifierDLModel.pretrained('classifierdl_use_emotion', 'en') \
.setInputCols(["document", "sentence_embeddings"]) \
.setOutputCol("sentiment")
nlpPipeline = Pipeline(stages=[document_assembler,
use,
classifier])
data = spark.createDataFrame([["@Mira I just saw you on live t.v!!"],
["Just home from group celebration - dinner at Trattoria Gianni, then Hershey Felder's performance - AMAZING!!"],
["Nooooo! My dad turned off the internet so I can't listen to band music!"],
["My soul has just been pierced by the most evil look from @rickosborneorg. A mini panic attack and chill in bones followed soon after."]]).toDF("text")
result = nlpPipeline.fit(data).transform(data)
val documentAssembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("document")
val use = UniversalSentenceEncoder.pretrained('tfhub_use', "en")
.setInputCols(Array("document"))
.setOutputCol("sentence_embeddings")
val classifier = ClassifierDLModel.pretrained("classifierdl_use_emotion", "en")
.setInputCols(Array("document", "sentence_embeddings"))
.setOutputCol("sentiment")
val pipeline = new Pipeline().setStages(Array(documentAssembler,
use,
classifier))
val data = Seq(Array("@Mira I just saw you on live t.v!!",
"Just home from group celebration - dinner at Trattoria Gianni, then Hershey Felder's performance - AMAZING!!",
"Nooooo! My dad turned off the internet so I can't listen to band music!",
"My soul has just been pierced by the most evil look from @rickosborneorg. A mini panic attack and chill in bones followed soon after.")).toDS.toDF("text")
val result = pipeline.fit(data).transform(data)
import nlu
text = ["""@Mira I just saw you on live t.v!!"""]
emotion_df = nlu.load('en.classify.emotion.use').predict(text, output_level='document')
emotion_df[["document", "emotion"]]
Results
+-------------------------------------------------------------------------------------------------------------------------------------+---------+
|document |sentiment|
+-------------------------------------------------------------------------------------------------------------------------------------+---------+
|@Mira I just saw you on live t.v!! |surprise |
|Just home from group celebration - dinner at Trattoria Gianni, then Hershey Felder's performance - AMAZING!! |joy |
|Nooooo! My dad turned off the internet so I can't listen to band music! |sadness |
|My soul has just been pierced by the most evil look from @rickosborneorg. A mini panic attack and chill in bones followed soon after.|fear |
+-------------------------------------------------------------------------------------------------------------------------------------+---------+
Model Information
| Model Name: | classifierdl_use_emotion |
| Compatibility: | Spark NLP 2.7.1+ |
| License: | Open Source |
| Edition: | Official |
| Input Labels: | [sentence_embeddings] |
| Output Labels: | [class] |
| Language: | en |
| Dependencies: | tfhub_use |
Data Source
This model is trained on multiple datasets inlcuding youtube comments, twitter and ISEAR dataset.
Benchmarking
label tp fp fn support
fear 0.78 0.67 0.72 2253
joy 0.71 0.68 0.69 3000
sadness 0.69 0.73 0.71 3075
surprise 0.67 0.73 0.70 3067
accuracy - - 0.71 11395
macro-avg 0.71 0.70 0.71 11395
weighted-avg 0.71 0.71 0.71 11395
PREVIOUSCyberbullying Classifier