Part 2 of the Example Template File - Teradata Vantage

In Part 2 of the feature template file, each row is a rule that indicates which nearby token to examine and which extractor class to apply to that token. The function applies each rule to each token in the input text. Based on the results of applying the rules, and the tagged entities in the training data set, the function builds a model.

The row format is %x[j,k] where j indicates the relative position from the current token (0 is the current token, -1 is the previous token, 1 is the next token, and so on) and k is the serial number of the extractor class to apply (from Part 1 of the feature template file).

For example, if the input text is "More restaurants open in San Diego." and the current token is "San", then the following table shows the selected feature for each template.