MaltOptimizer also gathers information about some properties, such as the percentage of non-projective arcs/trees which is by the way crucial in order to select the best parsing algorithm. And indeed it optimizes the feature model. Finally it tunes the parameters of the learning algorithm and creates an option file and a feature specification file.
About the learning time, it depends on the size of the training corpora. Of course if the system comes up with a very complex feature model or suggests a slower parsing algorithm (some Malt algorithms require more time than others) it affects the learning time. Nevertheless, the system guarantees that the suggested configuration is the best in performance that it can find. By the way, the time differences with the same training corpus are not very wide.
I recommend you to visit the website (http://nil.fdi.ucm.es/maltoptimizer) and if you are at all interested in MaltOptimizer or you have any further questions (or need help) do not hesitate to contact me directly.
On 17 February 2012 15:29, Mohammad Sadegh Rasooli <rasooli.ms at gmail.com>wrote:
> Hello Miguel,
> Do you mean that the mentioned tool works as a combined feature selection
> and learner model parameter selection tool? How this optimization affects
> learning speed?
> Mohammad Sadegh Rasooli
> Head of the Persian Dependency Treebank Project:
> On Fri, Feb 17, 2012 at 5:30 PM, Miguel Ballesteros <
> miguelballesteros1 at gmail.com> wrote:
>> Let us introduce MaltOptimizer, an optimization tool for MaltParser.
>> MaltOptimizer has been (and is) developed by Miguel Ballesteros<http://nil.fdi.ucm.es/index.php?q=node/449>from Complutense
>> University of Madrid <http://www.ucm.es/> (Spain) and Joakim Nivre<http://stp.lingfil.uu.se/%7Enivre/>from Uppsala
>> University <http://www.uu.se/en/> (Sweden).
>> Freely available statistical parsers often require careful optimization
>> to produce state-of-the-art results, which can be a non-trivial task
>> especially for application developers who are not interested in parsing
>> research for its own sake. MaltOptimizer is a freely available tool
>> developed to facilitate parser optimization with the open-source system
>> MaltParser, which offers a wide range of parameters for optimization,
>> including nine different parsing algorithms, two different machine learning
>> libraries (each with a number of different learners), and an expressive
>> specification language that can be used to define arbitrarily rich feature
>> models. MaltOptimizer is an interactive system that first performs an
>> analysis of the training set in order to select a suitable starting point
>> for optimization and then guides the user through the optimization of
>> parsing algorithm, feature model, and learning algorithm parameters.
>> The system will be demonstrated in the System Demonstration Session at EACL
>> 2012 <http://eacl2012.org/home/index.html> and is further described in a
>> paper to appear at LREC 2012 <http://www.lrec-conf.org/lrec2012/>.
>> For further information and download:
>> Best regards,
>> Miguel and Joakim
>> UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
>> Corpora mailing list
>> Corpora at uib.no
-- Miguel Ballesteros Universidad Complutense de Madrid NIL, Natural Interaction based on Language Website <http://nil.fdi.ucm.es/index.php?q=node/449> -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: text/html Size: 5502 bytes Desc: not available URL: <http://www.uib.no/mailman/public/corpora/attachments/20120217/69bb04cd/attachment.txt>