"Standalone" is not standalone about heideltime

[GoogleCodeExporter]

Dear Ryan,

Thank you for opening this issue. HeidelTime is available as
- a UIMA component, and as
- a standalone version, which can be used without UIMA.

In both cases, HeidelTime requires some preprocessing, namely, sentence 
splitting, tokenization, and part-of-speech tagging. For all languages except 
Arabic and Vietnamese, we use the TreeTagger for these tasks.

As explained in the readme of the UIMA version and in the Manual of the 
standalone version, you have to download the TreeTagger and its modules for the 
languages you want to process. In the standalone version, you then have to set 
the path to the TreeTagger in the config.props. 

Please have a look in the Manual for more details.

Thanks,
Jannik

Original comment by [email protected] on 21 Jun 2013 at 7:51

• Changed state: Done

[GoogleCodeExporter]

Ok, thanks. My goal is to use heideltime from Hadoop. If it's all in Java, this 
will be easy.

Original comment by [email protected] on 21 Jun 2013 at 4:32

[GoogleCodeExporter]

Ok, I'm still real confused here. I can't use perl for what I am doing. Is 
Heideltime Java or not?

Original comment by [email protected] on 24 Jun 2013 at 10:41

[GoogleCodeExporter]

Dear Ryan,

I would kindly refer to the Manual where you can find a description how to run 
the standalone version from the command line. Make sure that you have the 
TreeTagger installed and the path to the Treetagger set correctly in the 
config.props, but as already mentioned, everything is explained in the Manual.

If you run into specific problems, we are happy to help, but then we need to 
know what you are actually trying to do in more detail.

Thanks,
Jannik

Original comment by [email protected] on 26 Jun 2013 at 12:33

[GoogleCodeExporter]

My goal is to deploy Heideltime on a Hadoop cluster. Currently, I search for 
dates with regex. I'd like to improve on that.

The Treetagger dependency is where I am stuck. It runs fine on my laptop, but, 
because it's not Java, it's difficult (impossible?) to install/run Treetagger 
on every node in my cluster.

Can I somehow remove Treetagger and still get ok results? Perhaps there is a 
Java library out there that I can use in Treetaggerwrapper.java instead?

Original comment by [email protected] on 26 Jun 2013 at 7:05

[GoogleCodeExporter]

Hi Ryan,

You can use the Stanford POS Tagger instead of the TreeTagger -- however 
currently not with the standalone version. We will add parameter to the 
standalone version to decide which POS tagger should be used. However, this is 
not implemented yet. 

You could replace the TreeTaggerWrapper with the Stanford POS Wrapper in the 
source code of the standalone version. What you should keep in mind is that 
HeidelTime requires Sentence information. Without sentence information, you 
won't get any results. Without token and pos information, you can get results, 
however, they will probably be worse.

Thanks,
Jannik

Original comment by [email protected] on 28 Jun 2013 at 9:30