Born from the collaboration between Bing and Microsoft Research

Apr 29, 2010 15:43 GMT  ·  By

Microsoft Web N-gram Services was born through the collaboration between Bing and Microsoft Research, as a project focused on data-driven research. Evelyne Viegas, senior research program manager, Microsoft Research, revealed that the Redmond company had worked together with a limited pool of testers for the past year on the private Beta of the Web N-gram Services. At the World Wide Web Conference (WWW2010) this week, the software giant expanded the Beta testing program, making it public, and invited professors and students at accredited colleges and universities around the world to access and test the Web N-gram Services.

“The technologies included in Microsoft Web N-gram Services have been noted for their ability to assist in writing applications specific to search, translation, and speech processing. One of the immediate scenarios made possible by the technology is the ability to understand misspelled words and ungrammatical sentences by using the power of the sheer volume of language data, for any natural language that has lots of data published on the web. From a development perspective, this reduces the need for experts to develop grammars for all languages; users who conduct searches or network on the Internet will be enabled to share information in free form with stronger understanding and clarity. This is made possible by using predictions to contextualize the initial words in the query,” Viegas explained.

Bing’s Kristin Meldahl notes that the Microsoft Web N-gram Services are focused on supporting research using large data. In this regard, Microsoft Research and Bing have been looking at test scenarios involving natural computing experiences, such as linguistics, and leveraging speech NUI for search.

“As the technology and corresponding development efforts advance, Microsoft Web N-gram Services are expected to provide an accurate, consistent user experience, such as helping people learn another language or search for information with queries that are spoken rather than typed,” Viegas added.