Hello, I used especially the Name Finder with quite a few European languages, depending on the language the performance is worse (e.g. German, lots of initial capital words which are not names) or comparable to English (e.g. Danish, French).
We try to make the components as flexible as possible so they can be adapted/optimized for a specific language. If you need support for a new language we could enhance our feature generation to deal with it. The feature generation of the name finder for example can be configured with an xml descriptor. Jörn On 07/05/2012 02:24 AM, Lance Norskog wrote:
The sourceforge model site only includes European languages. Are any of the OpenNLP algorithms useful in other language families? I'm sure this is a large research area! I am just asking for concrete opinions. For that matter, how effective are the algorithms for the various European languages? All NLP research is funded by the US Intelligence canker, so I assume it works best in English :)
