John Slocum created OPENNLP-1307:
------------------------------------
Summary: Incorrect code example for Document Categorization (9.3)
Key: OPENNLP-1307
URL: https://issues.apache.org/jira/browse/OPENNLP-1307
Project: OpenNLP
Issue Type: Documentation
Components: Doccat
Affects Versions: 1.9.3
Environment: N/A
Reporter: John Slocum
Fix For: 1.9.3
in
[https://opennlp.apache.org/docs/1.9.3/manual/opennlp.html#tools.doccat.classifying.api,]
the code example feeds a String into DocumentCategorizerME.categorize(). The
method itself takes an array. I flagged priority as Major because this was a
killer - obviously it's a self-documenting bug when you run it, but I made the
mistake of assuming that the array actually needed would be an array of
documents - instead it needs to be an array of tokens from a single document,
i.e. one needs to split() the doc on whitespace. Lost 24 hours experimenting
with algos (maxent vs. naive_bayes) and params (cutoff, iterations, etc) before
figuring this one out.
Current(wrong) version:
{code:java}
String inputText = ...
DocumentCategorizerME myCategorizer = new DocumentCategorizerME(m);
double[] outcomes = myCategorizer.categorize(inputText);
String category = myCategorizer.getBestCategory(outcomes);
{code}
Should be more like:
{code:java}
String inputText = ... // sanitized document to be categorized
DocumentCategorizerME myCategorizer = new DocumentCategorizerME(m);
double[] outcomes = myCategorizer.categorize(inputText.split(" ");
String category = myCategorizer.getBestCategory(outcomes);
{code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
