AI_SENTIMENT¶

Syntax¶

AI_SENTIMENT( <text> [ , <categories> ] )

Arguments¶

Required:

text

A string containing the text from which sentiment is extracted.

Optional:

categories

An array containing up to ten categories (also called entities or aspects) for which sentiment should be extracted. Each category is a string. For example, if extracting sentiment from a restaurant review, you might specify ['cost', 'quality', 'service', 'wait time'] as the categories. Each category may be a maximum of 30 characters long.

If you do not provide this argument, AI_SENTIMENT returns only the overall sentiment.

Returns¶

An OBJECT value containing a categories field. categories is an array of category records. Each category includes these fields:

• name: The name of the category. The category names match the categories specified in the categories argument.

• sentiment: The sentiment of the category. Each sentiment result is one of the following strings.

  • unknown: The category was not mentioned in the text.

  • positive: The category was mentioned positively in the text.

  • negative: The category was mentioned negatively in the text.

  • neutral: The category was mentioned in the text, but neither positively nor negatively.

  • mixed: The category was mentioned both positively and negatively in the text.

The overall category record is always included and contains the overall sentiment of the text.

Example:

{
  "categories": [
    {
      "name": "overall",
      "sentiment": "mixed"
    },
    {
      "name": "Brand",
      "sentiment": "unknown"
    },
    {
      "name": "Cost",
      "sentiment": "negative"
    },
    {
      "name": "Professionalism",
      "sentiment": "unknown"
    }
  ]
}

Access control requirements¶

Users must use a role that has been granted the SNOWFLAKE.CORTEX_USER database role. See Required privileges for more information on this role.

Usage notes¶

AI_SENTIMENT can analyze sentiment in English, French, German, Hindi, Italian, and Portuguese. You can specify categories in the language of the text or in English.

Examples¶

The following example uses AI_SENTIMENT to get the overall sentiment of a food service review.

SELECT AI_SENTIMENT('A tourist\'s delight, in low urban light,
    Recommended gem, a pizza night sight. Swift arrival, a pleasure so right,
    Yet, pockets felt lighter, a slight pricey bite. 💰🍕🚀');

Return value:

{
  "categories": [
    {
      "name": "overall",
      "sentiment": "positive"
    }
  ]
}

In this example, a table named reviews contains a column named review_content containing the text of movie reviews submitted by users. The query returns the sentiment of several facets of up to ten reviews.

SELECT
  AI_SENTIMENT(
    review_content,
    ['concept', 'performance', 'script', 'cinematography', 'soundtrack']
  ),
  review_content
  FROM reviews LIMIT 10;

Regional availability¶

AI_SENTIMENT is available in the following regions:

Legal notices¶

Refer to Snowflake AI and ML.

Limitations¶

Snowflake Cortex functions do not support dynamic tables.