This post is also available in: heעברית (Hebrew)

Due to the absence of information required for automated training systems, dealing with minority language content has been regarded difficult for many translation and retrieval technologies until recently.

The information retrieval multilingual technology from Raytheon BBN enables English speakers to search for text and analyze speech segments in a foreign language, as well as obtain relevant summaries translated from a low-resource foreign language into English. The application facilitates the retrieval of information in the form of text or speech, as well as the translation of words and phrases, and aids in the comprehension of the material’s context and meaning without the need to learn a new language.

According to stripes.com, company’s revolutionary solution allows users to search through texts and spoken snippets from several languages using English search phrases. The tool was produced as part of the IAPRA MATERIAL initiative and took four years to build.

Raytheon’s new technology allows users to get documents and recordings in other languages for use in English queries, as well as give a specialized translation of pertinent terms. The user can grasp the context and meaning without having to learn the language in this way. Kazakh, Pashto, Swahili, Somali, and Tagalog were used to demonstrate the system’s performance even in languages with limited machine learning resources, such as