On Mon, Jan 6, 2014 at 8:47 AM, A. Mani <a.mani.cms(a)gmail.com> wrote:
On Mon, Jan 6, 2014 at 6:44 AM, sankarshan
<foss.mailinglists(a)gmail.com> wrote:
And, my take is that what is required is more focused and stronger
investment over repeated cycles into automated translation systems,
especially perhaps Statistical-MT.
Automated translation systems can never be sufficiently accurate.
Problems include those relating to handling of emotion words, tacit
knowledge etc.
I didn't make the claim that automated translation systems would be
accurate or even, sufficiently accurate. But here's the bit - the
choice is either generating enough new contributors, holding workshops
to get them trained and, then iterating this over and over again. Or,
it could be about adding to this with generated content which can then
be reviewed, curated and, the systems improved by cycles of such
processes.
There is a significant volume of content that is being auto-translated
and reviewed. And, the world is none the worse for it.
In a multi-user scenario, where people are not going
to follow a
strict mode of expression, it is bound to be worse.
Integration with things like linked open data project is yet to
happen, but would be useful.
Which open data project and, how would such integration help create content?
--
sankarshan mukhopadhyay
<https://twitter.com/#!/sankarshan>