@OSTAustria_DC on Twitter

OSTA

When it comes to our universities we like to hit the ground running with our "frontrunner system" as highlighted in… https://t.co/C7aNUNCmN9

by OSTA

OSTA

Fully-funded PhD positions ✅ World-class research environment ✅ Make yourself an academic 🎁 this 🎄 and apply for… https://t.co/xCrvSeb4vE

by OSTA

bridges vol. 26, July 2010 / Feature Articles

language_weaver_small.jpg Within a period of two years, the company "Language Weaver" has developed a fully functional commercial software product from a novel, statistics-based, translation technology brought to a research prototype by the company founders - professors and researchers at the University of Southern California's Information Sciences Institute.
The translations of documents, newscasts, and other source materials for defense and commercial purposes include Arabic, Farsi, Somali, Hindi, Chinese, French, and Spanish.
The company founders are still professors at the university. The company now has about 35 employees, many of them attracted from the university's Infor¬mation Sciences Institute.


Technology

The machine-based software uses computational algorithms and probability statistics to learn from existing translated parallel texts, analyze words and word groupings, and build translation parameters that will afford the highest statisti¬cal probability of providing a correct translation. Language Weaver's technology offers societal benefits in several ways: First, it reportedly achieves a significantly higher level of accuracy in translation than counterpart rule-based machine translations, delivering greater value to customers. Second, it is able to provide translation systems in languages for which there is a shortage of available translators and a considerable demand for translations, particularly for defense purposes. Third, it can be more cost-effective than human translators for translation of large volumes of information. Fourth, the technology may offer a faster means for obtaining needed translations by its ability to process large volumes of data quickly. For example, it reportedly can process in one minute what a human translator would take several days to produce.  


Funding

Funding sources for the company are federal government grants, venture capital, and licens¬ing revenue.

In 2001 the founders submitted a proposal to the National Science Foundation (NSF), and received an SBIR grant the following year. The technology was thought to be interesting and subsequently the company was able to obtain venture capital funding as well.


At the end of 2002, the company was incorporated when it got the first STTR grant. Language Weaver was then given a chance to convert the STTR into an SBIR grant, which was accepted, because the SBIR offered more advantages.

Since its founding in 2002, Language Weaver has received a total of $150,000 in Phase I SBIR grants and $1,500,000 in Phase II grants. The SBIR grants were given by the NSF and the US Army.

In addition to its SBIR grants, the company received a multi-year grant from the Advanced Technology Program (ATP) for a large-scale syntax-based system, expected to bear fruit several years out. The ATP funds amounted to $1,972,557 for the period 12/1/2004 -11/30/2007.

For the detailed case study, please visit this link .

Click here to go back to the article.