On 7/8/11 2:58 AM, Amal Elmah wrote:
I built a model using Opennlp using the instructions in the documentation
so I used training data in such the following format:
<START:person> Pierre Vinken<END> , 61 years old , will join the board as a
nonexecutive director Nov. 29 .
Mr .<START:person> Vinken<END> is chairman of Elsevier N.V. , the Dutch
publishing group .
<START:person> Rudolph Agnew<END> , 55 years old and former chairman of
Consolidated Gold Fields
I am now trying to evaluate this model but I don't know if the test data should
conatin tag as the above or not.
The OpenNLP built-in evaluation always needs the tags in the test data,
otherwise it assumes that
the data with no-tags is correct and counts every name as a mistake.
You also need much more training data.
Please feel free to send us a patch to our documentation to clarify this.
Jörn
