Sentiment Analysis

Introduction

Sentiment analysis is a stage in the content processing of the URL. The aim is to identify content with very negative sentiment, and label these content profiles with the keyword sentiment:negative. Note that there are no other valid values for the key sentiment. Below is an example of a content profile which is labeled with negative sentiment.

Both for ad serving and recommendation purposes, it may be benificial to identify content with very negative sentiment, as many publishers do not want to do contextual advertising/content recommendation on these pages (e.g. advertising for a car in a news article about a car accident). Therefore, we do not do contextual ad targeting/content recommendation targeting for URLs which have been labeled as "negative".

Supported languages

Sentiment analysis is currently available for English, Norwegian, Spanish and Swedish.

Enable/disable sentiment analysis for a site

All new sites which are added as of May 20th, 2015 will by default do sentiment analysis if the content language is supported. This means that when a new site is added, and the publisher does NOT want to utilize sentiment analysis, this must be disabled using the API. Similarly, sentiment analysis could be enabled for existing sites where this feature by default have been disabled. The full API documentation for this features is provided here. See below for examples on how sentiment analysis is enabled/disabled for a site.

Example

$ python cx.py /site/update '{"siteId":"1234", "sentimentDetection":true}'
{
  "id": "1234"
}
 
$ python cx.py /site '{"siteId":"1234"}'
{
  "sites": [
    {
      "id": "1234", 
     (...)
      "sentimentDetection": true
    }
  ]
}

$ python cx.py /site/update '{"siteId":"1234", "sentimentDetection":false}'
{
  "id": "1234"
}
 
$ python cx.py /site '{"siteId":"1234"}'
{
  "sites": [
    {
      "id": "1234", 
     (...)
      "sentimentDetection": false
    }
  ]
}

Last updated: January 28, 2022