Hello Ihab, Unfortunately having huge amounts of data is not enough. The methods used also matter. To improve the quality of translation you will need to improve on the existing algorithms. One way would be to focus on the peculiarities of the languages you are working with. Maybe you can look into factored SMT. Also hierarchical SMT might be another thing you can explore. Are you working with Arabic as one of the languages ? If yes, then you might be interested in lattice based translation. There are so many ways but no one can give a 100% guarantee as to which method will work best for your specific purposes. All this information is available on the moses page. I would also suggest that you read up the publications associated with the methods to understand the workings of these methods. Regards.
On Thu, Feb 19, 2015 at 8:00 PM, Ihab Ramadan <[email protected]> wrote: > Thanks Mr. Philipp for your reply > > But I have this plenty of data and I did not satisfy with the quality what > I miss then !!! > > How to improve the quality of output as much as possible > > Thanks > > > > *From:* [email protected] [mailto:[email protected]] *On Behalf Of *Philipp > Koehn > *Sent:* Tuesday, February 17, 2015 4:25 PM > *To:* [email protected] > *Cc:* [email protected] > *Subject:* Re: [Moses-support] Number of enough segments > > > > Hi, > > > > 2 million segments is plenty. > > > > This question is generally hard to answer - the more data you have the > better. > > There has been some success with already only 1 million words in narrow > > domains - the systems for news translation have typically at least a > magnitude > > more than that. > > > > -phi > > > > > > On Tue, Feb 17, 2015 at 4:00 AM, Ihab Ramadan <[email protected]> > wrote: > > Dears, > > I just wonder how much data should I use to say I have enough data to > build a qualified MT > > For example If I have 2 million segments in the parallel files is that > enough? > > Thanks > > > > > *Regards,**Ihab Ramadan | *Senior Developer* | **Saudisoft-Egypt | ** Tel: > *+2 023 303 2037 - *ext *128 | *M *+2 01007570826 | *Fax *+2 023 303 2036 > | > *Follow us on | | * > > > > > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support > > > > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support > > -- Raj Dabre. Research Student, Graduate School of Informatics, Kyoto University. CSE MTech, IITB., 2011-2014
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
