This option measures precision and recall between extraction runs by comparing the expected results of entity extraction with the actual results.
Expected results are created once and remain as a base reference for ongoing tests. Actual results are generated as required each time a grammar is modified. The two results are compared to generate precision and recall information.
To generate expected results, run edktool -extract
, and then revise the generated output file so that it contains the correct matches. From then on, edktool -extract
is used only to create the actual results, and the two files are compared against each other to generate precision and recall information on an ongoing basis.
-e <expectedfile>
|
The expected results file from edktool -extract . |
-a <actualfile>
|
The actual results file from subsequent extraction runs with modified grammar files. |
-o <resultsfile > |
The results, including precision, recall, and differences. |
-q
|
(Optional) Sets “Quiet Mode” so that descriptive messages are removed, and the output consists of only an XML document containing the differences between the expected and actual output. |
For more information on how to use the Measure feature to check the effectiveness and performance of your grammar files, refer to IDOL Expert.
|