Using the Sentiment Analysis Converter

Sentiment analysis attempts to determine whether documents have a positive, negative, or neutral disposition. Watson Explorer Engine Engine supports sentiment analysis via a built-in converter known as the Sentiment Analyzer. The Sentiment Analyzer can only perform sentiment analysis on documents that are in English.

To add the Sentiment Analyzer to the conversion pipeline for a search collection:

1. Navigate to the Configuration tab for the search collection that will be indexing the documents on which you want to perform sentiment analysis.
   Note: Because it is a part of the conversion pipeline for a search collection, the Sentiment Analyzer can only perform sentiment analysis on documents as they are being indexed, whether for the first time or as the result of a refresh of that search collection. To perform sentiment analysis on documents that have been indexed in an existing search collection, you must reindex that collection.
2. Select the Converting sub-tab.
3. Click Add a New Converter. The Add a New Converter dialog displays.
4. Select Sentiment Analyzer from the scrollable list in the left portion of this dialog and click Add to add the Sentiment Analyzer to the conversion pipeline for the search collection.

After adding the Sentiment Analyzer to the conversion pipeline for a search collection, you can either click OK to accept the default settings in the Sentiment Analyzer or customize its settings, as explained in the next section.

After accepting the default settings for the Sentiment Analyzer or customizing its settings, it will be invoked as part of the conversion pipeline when crawling the seed(s) that are associated with that search collection. When crawling the data sources associated with the seed(s) used in your collection, the Sentiment Analyzer will add a content element whose name attribute is disposition and raw_score to each document that was identified as having a positive, negative, or neutral disposition:

• The disposition content element will contain one of the words Positive, Negative, or Neutral.
• The raw_score element will contain positive or negative values that are appropriate for a positive or negative disposition, or the number 0 for neutral disposition.