For 45 €, you can buy Machine Translation: From Real Users to Research, the book of the 6th Conference of the Association for Machine Translation in the Americas held in early October, 2004.

The 30 revised papers presented were carefully reviewed and selected for inclusion in the book. The papers address all current issues in machine translation ranging from theoretical and methodological topics to applications in various contexts and evaluation and analysis of user needs and systems.

Rodolfo M. Raya has recently published a second article in his series on XML for localization.

This second part focuses on XML Localisation Interchange File Format (XLIFF) and explains with practical examples how to use it for translating different kinds of documents. This article presents a step-by-step guide to translating multilingual documents using XLIFF as an intermediary file format, and provides useful resources for localizing Java applications.

The first article briefly explained the most relevant XML standards used in the localization industry. By the way, the s/z shift in spelling localized is intentional – and a good example of spelling localis/zation.

If you use MSN Messenger and like playing English language games with chatterbots, try the beta version of Spleak, a female avatar produced by IMT Labs, a Danish start-up dedicated to enhancing the web interface experience. You can download Spleak for free and then yak away.

The underlying idea is that Spleak will act as an interface to information and eventually other services over and from the web. For this beta version, you can ask for word definitions, find out facts on U.S. presidents (!) and even ask for a Shakespeare play to read. Spleak naturally has Hamlet, Prince of Denmark, which then unfurls a dozen or so lines at a time in your Messaging window. Well, it’s a start. This beta is designed to help IMT Lab build up a corpus of questions, language usage etc. to expand the knowledge base and enrich the user experience. But it is not clear whether the program actually uses your query to consult existing web resources.

If you start trying to chat Spleak up, she does pretty well at evading the issue. If you treat her like ELIZA, the mythic shrinkbot program built by Joseph Weizenbaum in 1966, and tell Spleak you have a problem, she’ll probably reply “Whatever.” And when I asked why she’s called Spleak, she offered me a spell checker. As Hamlet would have said, Glet thlee to a nlunnery!

One of the more endearing aspects of the Wikipedia (online open source inspired encyclopedia) project is its decision to include Simple English among the localized versions of various Wikipedia articles. Indeed, the main page of the Simple English section is itself written in, err… Simple English. It is targeted at English learners and teachers, a little like the Collins Cobuild dictionary for learners which used ‘simple’ English in the definitions. The Simple English versions are also designed to help the many translators working on English originals.

Wikipedia has come in for considerable shtick from professional librarians and the like since, as a grass-roots open publishing enterprise generated by unpaid amateurs, it is open to abuse of all kinds. But as this recent article suggests, the very scope and reactivity of the production community ensures that errors and idiocies are removed faster than you can say Britannica.

What intrigues me, though, is that Wikipedia offers an extraordinary resource for people interested in the writing process, as the whole history of modifications are stored in the history section of any given article. Unlike most documents, whether digital or ex-arboreous, you can inspect the way Wikipedia’s articles have been edited, expanded, rewritten for readability and so on. And then you can go to a translated version, of the article and see the same process of editing applied to the new version. A rare treasure after five hundred years of final-version-only print publishing.

Which is where the Simple English Wiki comes in. First, it shows that writers can more or less easily ‘simplify’ their discourse, even without much training or theory. Second, the S E discussion forum offer a instructive glimpse into how various aspects of simple writing are thrashed out by real world practitioners. And third, the actual encyclopedia entries together with their editing histories offer a superb training resource for people interested in the potential for using ‘simple’ expressive means (sentence length, vocabulary, grammatical rules, etc) in expository discourse.

In the wake of the great native speaker quality English machine translation system, how about a map of the London underground fully localized into German? Mind the gap at Nimdill Zirkus!

UK schools are to be given official guidance (a 60 page booklet called Introducing the Grammar of Talk) on how to teach pupils standard spoken English. The Qualifications and Curriculum Authority says teaching about speaking is more difficult for many teachers than teaching writing skills, for which there is a well-developed system of grammar. It hopes the initiative will improve their social skills and career prospects.

Meanwhile in Poland, there is outrage about how Wojciech Pormorski has been told by the Hamburg (Germany) city authorities that regular dialogue in Polish with his daughters, aged seven and four, would hinder their integration into German society. He is separated from his wife and has been denied access to their daughters after refusing to guarantee that he would speak to them only in German.

Babeling, a French terminology engineering software developer, has just released Motilus, an interface to multi-database term queries over the web. Or as Babeling puts it:

Motilus is a unique interrogation engine software of monolingual and multilingual terminology resources and dictionaries available on the Internet. Its innovation is its ability to select and query, simultaneously, relevant dictionaries and terminology resources in a particular language combination in seconds. With both single-user or network (intranet, extranet) applications, Motilus’ high processing speed and user-friendly, attractive design set it apart.

Motilus’ originality is to spare the user from painstaking research and repeated querying of the lexicographical and terminological resources, by selecting, for each resource, the source language, the target language(s), and other possible criteria as well as typing in the searched for word. With Motilus, the user selects once: the source language, target language, and types in the word. With one click he/she will obtain a comprehensive list of resources containing the word. On clicking again, the content of the chosen resource will be brought up: the words found as a result of the search, their definitions, their translations etc.

Was the company aware of an English near-homonym of its corporate name? The product sounds useful, but I found it hard to get my head around the website blurb and services.

GistXL, an English (or Singlish) to Simplified Chinese SMS translation platform embedded in SingTel’s Singapore network, won a Merit Award in the recent Singapore National Infocomm Awards held last week. According to The Straits Times:

GistXL lays claim to be one of the first (free translation systems) in the world to build artificial intelligence into the system. This means, while other “rule-based” systems match an English word to a Chinese one, GistXL sieves through its database – within a fifth of a second – to draw links and contexts between words and phrases.

To be picky, perhaps they meant ‘draw links between words and their context” or “contextualize words in phrases” rather than actually “draw contexts”. And it’s a pity to trot out the old AI saw again. The system is presumably what MT folks call ‘data driven’. According to their blurb, GistXL plans to scale up to 18 languages by the end of 2005.

Clearly, the stripped-down expressive forms (aka controlled language) found in text messaging, together with the platform constraints on volume (GisXL will do a around 25 words per translation) are proving to be a fruitful field for commercial language engineers. It would interesting to know in detail how standardized actual SMS messages are becoming in the various linguistic sub-groups that use them. GistXL for example is focused on the specifics of Singlish (Singapore English), as filtered through the space and typing constraints of SMS. As it happnes, Singaporeans can be pretty good at inputting ‘standard’ English too. In June this year, a student called Kimberly Yeo, broke the world record by typing 26 words into his cell phone in 43.24 seconds. At that rate, he could send 83 messages (or 2,165 words) an hour. A golden egg for his telecoms supplier.

Excessive verbiage typed into a chat room or IMS could suggest you’re lying. A report from Cornell University has found that monologuing liars

“talk too much, use more pronouns about others and use more terms about the senses, such as “see,” “hear” and “feel,” than people telling the truth, according to a new study by Cornell communication experts.

“Our study suggests that people who are lying to another person in a chat room or in instant messaging use approximately one-third more words, probably in their attempt to construct a more cohesive and detailed story in order to seem believable,” said Jeff Hancock, assistant professor of communication in the College of Agriculture and Life Sciences (CALS).”

On the other hand, if the ambient temperature isn’t warm enough in your office, you tnde to ùmake more typring mishtakes. Simson Garfinkel blogs yet another Cornell report :

“When the office temperature in a month-long study increased from 68 to 77 degrees Fahrenheit, typing errors fell by 44 percent and typing output jumped 150 percent. Hedge’s study was exploring the link between changes in the physical environment and work performance.

“The results of our study also suggest raising the temperature to a more comfortable thermal zone saves employers about $2 per worker, per hour,” says Hedge, who presented his findings this summer at the 2004 Eastern Ergonomics Conference and Exposition in New York City.

In other words, a chatroom liar in a cold room will type more and with more errors, than their truth-telling counterpart in a warm room. True or flase?

Alessandro Barrico, the Italian translator of Homer, has outperformed The Da Vinci Code in sales in Italy with his new Italian version of the Iliad, published by Feltrinelli.

Enthusiasm for the book has led to public readings, starting 2,600 people listening for 12 hours a night for three nights to the spoken translation in Rome. On October 8, the book’s 20,000 or so lines were read in relays by people from different backgrounds in Verona for 24 hours. A new. The next step in this mass vote for the old bard and his martial themes could be to memorize the whole thing.

It wouldn’t be the first time some one has tried, though. Although it’s hard to find out much about it, a retired U.S. businessman called Steven V.N. Powelson decided to memorize the whole Iliad, and was able to perform vast chunks of it around 1994. This meant getting your personal Mnemosyne to get its aural archive around some 200,000 spoken syllables. Powelson took 16 years, working about an hour a day (making a total of 5,840 hours) as befits a 76 year old. He apparently used to recite it as part of a national tour to promote the classics, but I cannot discover whether he ever stood up like Italy’s war-scarred public (Iraq not Troy) and performed the whole thing from beginning to end. For a weird commentary on the ‘autoerotics’ of memorizing poems see this

Here’s how Barrico summarized his motivations:

Quando racconto questo lavoro, spesso la gente mi chiede: perché proprio l’Iliade? Alcuni vorrebbero l’Odissea (che io non amo, tranne il finale), o magari Dante o Ariosto. Ho due risposte: la prima è che l’Iliade mi sembra una storia bellissima. La seconda è che godere del racconto di una guerra mi sembra una cura efficace per allontanare il desiderio (tragico ma legittimo) di godere facendo la guerra.