Description
This Spanish Text Classifier will identify from the viewpoint of a target whether a financial statement is positive, neutral or negative. This model is trained from the competition - IBERLEF 2023 Task - FinancES. Financial Targeted Sentiment Analysis in Spanish. We have used the participation dataset which is a small subset of the main one to train this model.
Predicted Entities
positive, neutral, negative
How to use
documentAssembler = nlp.DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
tokenizer = nlp.Tokenizer()\
.setInputCols("document")\
.setOutputCol("token")
sequenceClassifier = finance.BertForSequenceClassification.pretrained("finclf_bert_consumer_sentiments","es","finance/models")\
.setInputCols("token", "document")\
.setOutputCol("class")\
.setCaseSensitive(True)
pipeline = nlp.Pipeline(
stages=[
documentAssembler,
tokenizer,
sequenceClassifier
]
)
Results
+-------------------------------------------------------------------------+----------+
|text |result |
+-------------------------------------------------------------------------+----------+
|Renfe afronta mañana un nuevo dÃa de paros parciales de los maquinistas|[negative]|
+-------------------------------------------------------------------------+----------+
Model Information
| Model Name: | finclf_bert_consumer_sentiments |
| Compatibility: | Finance NLP 1.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [document, token] |
| Output Labels: | [class] |
| Language: | es |
| Size: | 408.7 MB |
| Case sensitive: | true |
| Max sentence length: | 128 |
References
https://codalab.lisn.upsaclay.fr/competitions/10052#learn_the_details
Benchmarking
labels precision recall f1-score support
negative 0.59 0.72 0.66 36
neutral 0.77 0.79 0.80 80
positive 0.72 0.55 0.64 42
accuracy - - 0.73 158
macro-avg 0.69 0.69 0.70 158
weighted-avg 0.71 0.71 0.72 158