Natural Language Processing with Transformers

About the Transformers Library

The transformers library provides an interface to interact with models.

In this post, I’ll share how we can set up a simple sentiment analysis of text.

Setting things up

pip install transformers

from transformers import pipeline

Sentiment Analysis

Positive, Negative

classifier = pipeline("sentiment-analysis", model = "distilbert/distilbert-base-uncased-finetuned-sst-2-english")

data = ["This product is good", "I don't like this product"]
classifier(data)

[{'label': 'POSITIVE', 'score': 0.9998680353164673},
 {'label': 'NEGATIVE', 'score': 0.9953964352607727}]

Emotion Detection

classifier = pipeline("sentiment-analysis", model = "SamLowe/roberta-base-go_emotions", top_k=None)

data = [ "It was perfect for the first two days and then it was awful"]
results = classifier(data)
print(results)

[[{'label': 'disgust', 'score': 0.7053156495094299}, 
{'label': 'annoyance', 'score': 0.12410394847393036}, 
{'label': 'approval', 'score': 0.0792454406619072}, 
{'label': 'disapproval', 'score': 0.050548385828733444}, 
{'label': 'neutral', 'score': 0.04294545575976372}, 
{'label': 'disappointment', 'score': 0.03646270930767059},
{'label': 'admiration', 'score': 0.018407044932246208}, 
{'label': 'embarrassment', 'score': 0.016334818676114082}, 
{'label': 'anger', 'score': 0.01377098634839058}, 
{'label': 'realization', 'score': 0.01139384787529707}, 
{'label': 'fear', 'score': 0.01103982049971819}, 
{'label': 'love', 'score': 0.00794794037938118}, 
{'label': 'sadness', 'score': 0.007545181084424257}, 
{'label': 'gratitude', 'score': 0.005044595338404179}, 
{'label': 'desire', 'score': 0.0035985559225082397}, 
{'label': 'amusement', 'score': 0.0033641094341874123}, 
{'label': 'confusion', 'score': 0.003023572964593768}, 
{'label': 'excitement', 'score': 0.0023223150055855513}, 
{'label': 'curiosity', 'score': 0.002252183621749282}, 
{'label': 'optimism', 'score': 0.002141808858141303}, 
{'label': 'joy', 'score': 0.0018355769570916891}, 
{'label': 'remorse', 'score': 0.0016548713902011514}, 
{'label': 'surprise', 'score': 0.0016250367043539882}, 
{'label': 'caring', 'score': 0.001429390744306147}, 
{'label': 'nervousness', 'score': 0.0010988765861839056}, 
{'label': 'relief', 'score': 0.0010947092669084668}, 
{'label': 'pride', 'score': 0.0010576138738542795}, 
{'label': 'grief', 'score': 0.0008262579794973135}]]