AppTek Releases the New State-of-the-Art in Machine Translation Software With Hybrid MT

Hybrid MT Provides Breakthrough Advances in Translation Software and User Operating Economics

AppTek, a leader in human language technology (HLT), today released a complete hybrid machine translation (HMT) system, which is a full integration of a statistical and a rule-based system to provide the performance and benefits of both in one solution. This is a marked improvement in machine translation (MT) software in which all of the key criteria of MT are addressed by one system. The three criteria of translation systems -- fluency, informativeness, and adequacy -- are now supported in one comprehensive system with greater performance.

This system is the result of a hybridization that integrated and enhanced our two complete and proven systems -- one statistical and the other rule-based. This was a significant effort to provide a full integration of two complete systems, rather than simply taking a shortcut approach and adding some rules to the statistical system or adding a minor statistical module to the rule-based engine. We have released a system that provides all the features of both MT approaches.

The advantages of statistical systems are that they are very consistent and they have a higher level of fluency -- the translated text reads better. Rule-based systems do not always read as well, but have a higher level of informativeness -- the translated information is often more accurate. AppTek's Chief Scientist, Hassan Sawaf, said that, "Today, no single system provides a fully-automatic, high-quality machine translation. Our portfolio of MT applications came the closest before today, but our hybrid MT solution now combines the benefits of each approach."

AppTek's Hybrid MT uses the company's statistical machine translation platform and augments it with its rich rule-based machine translation engine, pushing the state-of-the-art in machine translation (MT) design to the next level. The HMT system doubled the level of fluency and further increased the level of informativeness, which was already high in the rule-based system. These are significant advances in MT performance.

Organizations that utilize MT will find that Hybrid MT offers very attractive operating economics on a life-cycle basis compared to deploying multiple platforms to achieve the same level of translation performance.

ABOUT APPTEK

AppTek, headquartered in McLean, Virginia, is a developer of human language technology products with a complete suite for text and speech (voice) processing and recognition. The Company also leads major research and development efforts to further the advancement in the field of developing better methods and technologies in the field of HLT. AppTek's product offerings include machine translation (MT) and automatic speech recognition (ASR) for a growing list of more than 23 languages; multilingual information retrieval with query and topic search capabilities; name-finding applications; and integrated suites providing automatic speech recognition and machine translation in media monitoring of broadcast and telephony speech as well as handheld and wearable speech-to-speech translation devices. The company has language professionals and computer scientists in offices around the world.

Add to Digg Bookmark with del.icio.us Add to Newsvine