🙂 Sentiment Analysis model

Use the default Sentiment Analysis model

Need a general-purpose Sentiment Analysis model? You can use Artifex's default Sentiment Analysis model, which is trained to recognize five sentiments out-of-the-box:

• very_positive
• positive
• neutral
• negative
• very_negative

from artifex import Artifex

sentiment_analysis = Artifex().sentiment_analysis

print(sentiment_analysis("The product exceeded my expectations and works flawlessly!"))

# >>> [{'label': 'very_positive', 'score': 0.9955}]

Learn more about the default Sentiment Analysis model on our Sentiment Analysis HF model page.

Create & use a custom Sentiment Analysis model

Need more control over the sentiments recognized, or do you want to tailor the model to your specific domain for better results? Fine-tune your own Sentiment Analysis model, use it locally on CPU and keep it forever:

from artifex import Artifex

sentiment_analysis = Artifex().sentiment_analysis

model_output_path = "./output_model/"

sentiment_analysis.train(
    domain="e-commerce product reviews", # change to your desired domain
    classes={ # define your custom sentiments
        "positive": "Indicates a positive sentiment towards the product.",
        "negative": "Indicates a negative sentiment towards the product.",
        "neutral": "Indicates a neutral sentiment towards the product, or the absence of strong feelings.",
    },
    output_path=model_output_path
)

sentiment_analysis.load(model_output_path)
print(sentiment_analysis("The product exceeded my expectations and works flawlessly!"))

# >>> [{'label': 'positive', 'score': 0.9914}]