The sentiment insight uses a powerful deep learning model to analyze the tone of the top 1000 documents in your search results and classify them on a scale from Negative to Positive. For example, praise would be classified as positive and criticism would be classified as negative, whilst a statement of fact would be classified as neutral. When the classifier is uncertain about a document it distributes the document across the scale according to probability. Please note that perfectly accurate classification is impossible - even a state-of-the-art model that is 90% accurate will be wrong about every tenth document. Do not expect every document to be correctly classified, but statistics about the distribution of sentiment will be accurate.

Sentiment is currently supported in: English, French, German & Italian. If you have an unsupported language selected you will see the warning symbol appear.