Account confirmers, canmove, Confirmed users
2,357
edits
Intellego: Difference between revisions
Jump to navigation
Jump to search
→Project details
(→Discussion: Add mention of newsgroup.) |
|||
| Line 11: | Line 11: | ||
; Our Motto | ; Our Motto | ||
: We Will Be Understood. | : We Will Be Understood. | ||
; The Technical Details | |||
<ol><li>Intellego will be a machine translation platform consisting of various open source engines based on the most prominent approaches to MT (i.e., SMT, RBMT, EBMT, and hybrid).</li> | |||
<br> | |||
'''Why?''' Research has revealed that certain approaches produce better output for certain language pairs and content types. For example, Russian's grammar is incredibly complex. So much so, that SMT output for Russian is usually very flawed. The RBMT approach has demonstrably produced better output for language pairs that include Russian. In addition, RBMT is best suited for long sentences and structured content, like wikis, whereas SMT is best suited for short sentences and user generated content. | |||
<li>Intellego will aim to provide a single API for engine developers and a unified web service that hosts a number of different language pairs/engines/implementations in the back end.</li> | |||
<br> | |||
'''Why?''' This aims to increase accessibility to smaller, more efficient MT engines on the web. Kevin Scannell gave this example: "If you look at Apertium for example, there are some language pairs that are better performing than Google Translate, and many pairs that Google doesn't support at all. But they don't have the infrastructure to keep a web service up and running (they've tried and it's been up and down)." This will help break up the proprietary nature of MT and allow for a greater presence of open MT on the web. | |||
In addition, it will help to satisfy our aim to make the Intellego platform available through an open API and web services, as is stated on the wiki. | |||
<li>The GSoC terminology-based project will serve as a pre-processing utility in the Intellego MT process to provide accuracy in translation.</li> | |||
<br> | |||
'''Why?''' Research suggests that when a user evaluates MT output, they tend to be more accepting of MT error when it is grammar based, rather than terminology based. | |||
</ol> | |||
== Project meetings == | == Project meetings == | ||