SENTIMENT_CLASSIFIER

Use the SENTIMENT_CLASSIFIER type to create a preference for sentiment analysis queries. This classifier specifies preferences associated with a user-defined sentiment classifier preference. You must define a preference of this type before you use the CTX_CLS.SA_TRAIN_MODEL procedure to train the user-defined sentiment classifier. Table 2-43 lists the attributes for the SENTIMENT_CLASSIFIER type.

Table 43 SENTIMENT_CLASSIFIER Attributes

Attribute Data Type Default Minimum Value Maximum Value Description
MAX_DOCTERMS I 50 10 8192 Specify the maximum number of distinct terms representing one document
MAX_FEATURES I 3000 1 100000 Specify the maximum number of distinct features used to build a sentiment classifier
THEME_ON B False     Specify if themes must be extracted as features
TOKEN_ON B True     Specify if tokens must be extracted as features
STEM_ON B True     Specify if stemmed tokens must be extracted as features
MEMORY_SIZE I 500 10 4000 Specify the typical memory size, in MB, used to build the sentiment classifier.
SECTION_WEIGHT I 2 0 100 Specify the integer multiplier for term occurrence within a field section
NUM_ITERATIONS I 600     Specify the maximum number of iterations for which the sentiment classifier is run before it converges

See Also: Oracle Text Application Developer’s Guide for an example of using the SENTIMENT_CLASSIFIER type