Jan Krikke After 50 years of research and tinkering, machine translation might be ready to compete with human translators. Several companies have announced breakthroughs or substantial progress in MT research in recent months. In January, for example, Steven Klein, CEO of New York-based Meaningful Machines, announced that his company successfully tested new translation algorithms that he says could lead to translation engines replacing human translators. "While our current prototype is already outperforming other systems on limited resources," says Klein, "we expect to see significant improvement to our quality as both the target language corpus and the dictionary continue to increase in size, with a realistic goal of reaching human quality." "Although the prototype is only partially complete," says Klein, "we recently began blind testing from Spanish to English, and our system is already performing at higher quality levels on the BLEU (Bilingual Evaluation Understudy) scale than any system we are aware of--0.6092. Systran, whose Spanish-to-English system is one of the best, scored a 0.5494 when we ran it through the same test, and the Systran system has been through many decades of development and incremental improvements." Meaningful Machines' test has not been independently verified, and the goal of reaching near-human quality translation will probably depend on some degree of pre- and post-editing for years to come. But, the growing number of global corporations (such as Philips, Samsung, and HP) and international agencies and institutions (such as the UN and the European Commission) using the technology illustrates that machine translation--the first nonnumerical application of AI--is finally delivering practical solutions. Popular perception of MT has suffered from low-quality "gisting" translation that Web-based translation engines, such as Babelfish and other online services, generate. But MT engines designed for limited domains, and tailor-made systems that use controlled language, are already delivering services. The site makes available a wealth of information previously inaccessible to non-Japanese speakers. MT has also made it to the desktop. The system is self-learning--it improves over time as its associative memory grows.

As the United States and China move forward in both collaboration and competition, the ability to communicate becomes ever more critical. While emerging technologies such as Google Translate have shown promise, much work must be done to improve the language translation applications that America will need as one of its most important 21st century relationships develops. To move the technology forward, the Defense Advanced Research Projects Agency (DARPA) has awarded a $13.7 million grant, called "Linguistic Resources for Multilingual, Genre-Independent Language Technologies via its Broad Operational Language Translation" (BOLT) Program to the Linguistic Data Consortium at the University of Pennsylvania to develop linguistic resources. Brandeis has been given $2 million of that amount as a collaborator. Nianwen Xue, assistant professor of linguistics in the Language and Linguistics Program and the Department of Computer Science at Brandeis, is the principal investigator on the four-year project.

Following the recent adoption by the machine translation community of automatic evaluation using the BLEU/NIST scoring process, we conduct an in-depth study of a similar idea for evaluating summaries. The results show that automatic evaluation using unigram co-occurrences, i.e. ROUGE, between summary pairs correlates surprising well with human evaluations, based on various statistical metrics; while direct application of the BLEU evaluation procedure does not always give good results. For the inception of ROUGE, please read Lin & Hovy's HLT-NAACL 2003 (Lin and Hovy 2003) paper. For more details, please read Lin's paper "ROUGE: a Package for Automatic Evaluation of Summaries" (Lin 2004a).

An app offering real-time translations is to allow people in Japan to speak to foreigners over the phone with both parties using their native tongue. NTT Docomo - the country's biggest mobile network - will initially convert Japanese to English, Mandarin and Korean, with other languages to follow. It is the latest in a series of telephone conversation translators to launch in recent months. Alcatel-Lucent and Microsoft are among those working on other solutions. The products have the potential to let companies avoid having to use specially trained multilingual staff, helping them cut costs. They could also aid tourism.

Software that can translate spoken English into spoken Chinese almost instantly has been demonstrated by Microsoft. The software preserves intonation and cadence so the translated speech still sounds like the original speaker. Microsoft said research breakthroughs had reduced the number of errors made by the instant translation system. It said it modelled the system on the way brains work to improve its accuracy. Details about the project were given by Microsoft research boss Rick Rashid in a blogpost following a presentation he gave in Tianjin, China, in late October that had, he said, started to "generate a bit of attention". In the final few minutes of that presentation the words of Mr Rashid were almost instantly turned into Chinese by piping the spoken English through Microsoft's translation system.

Language barriers are starting to crumble. This month Japan's dominant mobile phone operator, NTT DoCoMo, introduced the world's first app for real-time voice translation. When a user with a DoCoMo smartphone places a call through the app, he speaks in Japanese and his words are promptly translated into English, Mandarin, or Korean. To complete the conversational circuit, the other person's words are translated from any of those languages back into Japanese. With this debut we've taken one step closer to building a mechanical Babel fish, the extraordinarily useful creature imagined by Douglas Adams in The Hitchhiker's Guide to the Galaxy.

Google has brought in optical character recognition (OCR), a feature that helps translate images of words, to its update Google Translate for Android applicaiton. The Android version of Google's language translation service has translated several languages worldwide for some time now. The new Google Translate app for its mobile operating system, however, includes the ability to translate text from images. The search giant said that the updated Google Translate taps on the power of OCR from its Google Goggles service to break down images and translate them to text. "This makes Google Translate for Android one of our most intelligent and machine learning-intensive apps. Speech recognition, handwriting recognition, OCR, and machine translation all rely on powerful statistical models built on billions of samples of data," said Etienne Deguine, associate product manager for Google Translate.

I've made an interesting observation recently while talking to people about Thinkudo Enlighten. It regards the misunderstanding between Text Analytics and Machine (automated) Translation. More than once people've asked "How did you do the Chinese translation?" So in this post, I'd like clarify the difference between them. Whether or not Machine Translation should be a substudy of Text Analytics, I will leave it to the readers within academia to discuss.

Recent discussions force me to return to the subject of translation versus SEO -- particularly machine translation -- as it seems this old topic has not yet gone away. For multinational sites, maintaining your site can be an expensive affair, and the cost savings of machine translation seem outstandingly attractive. But as my mother always says to me, "If something looks too good to be true, it usually is." Machine translation is a good case in point. Often abbreviated to "MT", machine translation involves using computers to do the work which human translators would normally do.

Machine learning (ML) is all the rage, riding tight on the coattails of the "big data" wave. Like most technology hype, the enthusiasm far exceeds the realization of actual products. Arguably, not since Google's tremendous innovations in the late '90s/early 2000s has algorithmic technology led to a product that has permeated the popular culture. That's not to say there haven't been great ML wins since, but none have as been as impactful or had computational algorithms at their core. Netflix may use recommendation technology, but Netflix is still Netflix without it.