Sentiment Analyzer

class vnlp.sentiment_analyzer.sentiment_analyzer.SentimentAnalyzer(model='SPUCBiGRUSentimentAnalyzer', evaluate=False)[source]

Main API class for Sentiment Analyzer implementations.

Available models: [‘SPUCBiGRUSentimentAnalyzer’]

In order to evaluate, initialize the class with “evaluate = True” argument. This will load the model weights that are not trained on test sets.

predict(text: str) int[source]

High level user API for discrete Sentiment Analysis prediction.

1: Positive sentiment.

0: Negative sentiment.

Parameters

text – Input text.

Returns

Sentiment label of input text.

Example:

from vnlp import SentimentAnalyzer
sentiment_analyzer = SentimentAnalyzer()
sentiment_analyzer.predict("Sipariş geldiğinde biz karnımızı çoktan atıştırmalıklarla doyurmuştuk.")

0
predict_proba(text: str) float[source]

High level user API for probability estimation of Sentiment Analysis.

Parameters

text – Input text.

Returns

Probability that the input text has positive sentiment.

Example:

from vnlp import SentimentAnalyzer
sentiment_analyzer = SentimentAnalyzer()
sentiment_analyzer.predict_proba("Sipariş geldiğinde biz karnımızı çoktan atıştırmalıklarla doyurmuştuk.")

0.08

SentencePiece Unigram Context BiGRU Sentiment Analyzer

class vnlp.sentiment_analyzer.spu_context_bigru_sentiment.SPUCBiGRUSentimentAnalyzer(evaluate)[source]

SentencePiece Unigram Context Bidirectional GRU Sentiment Analyzer class.

  • This is a Bidirectional GRU based Sentiment Analyzer that uses SentencePiece Unigram tokenizer and pre-trained Word2Vec embeddings.

  • It achieves 0.9469 Accuracy, 0.9380 F1 macro score and 0.9147 F1 score (treating class 0 as minority).

  • For more details about the training procedure, dataset and evaluation metrics, see ReadMe.

predict(text: str) List[Tuple[str, str]][source]
Parameters

text – Input text.

Returns

Sentiment label of input text.

predict_proba(text: str) float[source]
Parameters

text – Input text.

Returns

Probability that the input text has positive sentiment.