You make me feel like a natural language

0

The expression natural language seems to be going mainstream. It is shifting from a more technical application (natural L versus programming L) in the domain of the computer modeling and processing of ‘human’ languages (natural language processing) and now includes the concepts of ‘informal’ and ‘un-controlled’; the sort of thing you say spontaneously when speaking in everyday contexts. As it happens, typing a ‘natural’ style of language query into a search engine is usually a far more longwinded and less ‘natural’ process than it is to keyboard in a medley of isolated nouns, adjectives and verbs and hope for the best. Humans are pretty good at tailoring language to the demands of speed and effectiveness, witness the archetypal ‘unnatural’ communication style of the telegram.

One area where this new “natural language” agenda is popping up like fungi in the fall is in the field of search technology. Search’s cutting edge is clearly moving towards giving “natural language” answers to “natural language” queries, rather than plunking a list of highlighted documents in front of you for perusal. On the mobile front, the private German firm Linguit has rolled out its Nuggets program, which delivers automatic answers (by trawling websites) to questions sent in SMS over mobile phones. This is certainly one up from the Any Questions Answered service, also launched in the UK, which uses human agents to prepare the answers to queries and SMS them back to the sender. So far, these services seem primarily targeted at players of quiz games.

On the desktop front, watch out for yet another new search engine called Stochasto that will also be able to answer questions in, yes, “natural language” rather than return documents. Apparently there is already a Russian version of this operating, and although registered in Norway, the company appears to be driven by Russian language tech skills. It’s interesting to note that FAST, the real time search and filter company, was also started up in Norway.

Cross-lingual search report

0

Multilingual information retrieval R&D in Europe has been the focus of a digital libraries initiative called Cross-Language Evaluation Forum (CLEF) for some years now. Anyone wanting to know about the state of the art can consult Carol Peters report of the latest CLEF workshop, which summarizes how the systems tested for text, image and actual information retrieval benchmarks.

“CLEF 2004 … marked a breaking point with respect to previous campaigns. The focus was no longer concentrated on multilingual document retrieval but was diversified to include different kinds of text retrieval across languages (exact answers in the question-answering track) and retrieval on different kinds of media (not just plain text but collections containing image and speech as well). In addition, increased attention was given to issues that regard system usability and user satisfaction with tasks to measure the effectiveness of interactive systems or system components being included in both the QA and the ImageCLEF tracks, with the collaboration of the coordinators of iCLEF (the interactive track).”

Mobile Qur’an

0

The Gulf Daily News reports that the first Islamic mobile phone, the ilkone i800, was launched in Bahrain on Wednesay.

The new product is the first phone to have full Islamic applications and features, such as a complete text of the Holy Quran with English translation, automatic prayer-time alarm, Qibla directions from anywhere in the world and automated silence or vibration mode for 40 minutes during prayer time.

Reading matter

0

Barbara Cassin, a French philosopher working with a team of other Europeans, has just published VOCABULAIRE EUROPEEN DES PHILOSOPHIES ; DICTIONNAIRE DES INTRADUISIBLES. A somewhat hard to translate title – is it European Vocabulary of Philosophies: Dictionary of Untranslatable terms, or European Philosophical Vocabulary, or Vocabulary of European Philosophies, etc? Fifteen hundred pages on how to (sometimes fail to) translate such terms as German Dasein, English agency, Russian mir, Greek logos, Hebrew Torah, Italian virtù and even Portuguese saudade. Probably not for the commercial translator’s bookshelf, but a fascinating journey through the joys and sufferings of language.

Nervad Kermani, the German-Iranian writer, has published an interesting article in the October 1 issue of The Times Literary Supplement (probably need to be a subscriber) on ‘The language of Islam and how Osama bin Laden betrays it’. Sample:

Watching Osama bin Laden’s first video broadcast after the start of the American air offensive on Afghanistan, I was struck by the exquisite Arabic he spoke. Not once did he slip into dialect, as usually happens with the modern generation of Arabic leaders, nor did he confuse the complicated flexional endings, a mistake made even by intellectuals. He chose antiquated vocabulary, familiar to educated Arabs from religious literature and classical poetry, and avoided neologisms. It was indeed the stiff, puritanical, conformist, artificial Arabic as described above, but it was immaculate. And for the first time, watching bin Laden’s broadcast, I found myself falling under its spell.

It sounded like a traditional speech, but it represented a break with tradition. Arab theologians speak very differently – if they are rhetorically well educated – with an exquisitely varying enunciation of high Arabic consonants, precise modulation and length of vowels, the result of many years of learning rhetoric and Qur’anic recitation. Osama bin Laden, being a businessman by profession, lacks this training, and although he speaks antiquated Arabic, it sounds simple, clear and modest. In fact, his rhetoric works precisely because of the absence of rhetorical ornament, and a conscious modesty of expression.

This linguistic asceticism marks a rejection of the burden of tradition, a return to roots. In the video his prophetic aura was reinforced by his austere attire and location in a cave in Afghanistan, a clear reference to the cave in which the Prophet received his first revelation. Even the lack of accentuation in bin Laden’s rhetoric echoes the puritanical Wahabitic spirit, which is allegedly identical with the divine spirit of the Prophet. This break with prevailing tradition was most obvious when bin Laden cited phrases from the Qur’an: while other speakers grotesquely raise and lower their voices when they recite the Revelation, Osama bin Laden proceeded in the same solicitous tone, as if he wished to persuade his audience through the clarity of his message alone.

Michel, Muhammed and Sarkozy: spot the foreign name

0

Should call-center operatives reveal their real offshore names? The issue has come up in a recent French ruling to ostensibly bring call center behavior in line with other consumer information requirements. Most data in English on call-centers naturally focus on outsourced call centers in India, and how well-educated English speaking locals are prepared to play Sandra or Roger for the likes of British bank account owners or U.S. insurance premium purchasers. And the language offer is widening. According to this report, some Indian operatives may soon be switching their names to Sabine or Reinhardt:

Courtesy a recent decision of the Electronics and Computer Software Export Promotion Council (ESC), even students studying information technology at engineering colleges can now learn German at ESC-recognized institutions by paying just 50 per cent of the course fee, subject to a ceiling of Rupees 12,000.

This scheme would be applicable at 11 German-training institutions in Bengal.

“The decision to extend the benefits of the language-related market access facilitation initiative to even students interested in German stems from our desire to help them become more employable,” a ESC official said.

In the French case, the new ruling (from Finance Minister Nicholas Sarkozy) was mooted in April and turned into a decree last week. The first two Articles say: “All goods and service suppliers who use a call center to issue or receive phone communications must inform callers of the geographical locale of the call center. This information must be provided at the beginning of the call and feature on all marketing collateral giving the coordinates of the center. Single calls to multiple call center geographies must also be declared.”

In other words, when you ring in France at night these days to change your mobile phone account, there is every likelihood that you will talk to Monique (aka Rim) in Casablanca, Morocco, Michel (or Hedi) in a suburb of Tunis, or Robert (or Mamadou) in Dakar, Senegal. Many call centers switch their night and weekend services to offshore locations in North or West Africa, where French is widely (and well) spoken and labor cheaper. If the Sarkozy ruling goes through, Rim, Hedi and Mamadou will have to identify themselves to each caller. Presumably the idea behind this is that French citizens will be so outraged that they are not speaking to real Monique & Co that they will switch to a service that promotes the authentic Gallicity of their call center. Which will in turn mean keeping or creating more jobs in France, and put the brakes on what the French call ‘délocalisations’ or offshoring.

As in India, call centers in Africa are considered a growth industry, providing job outlets for educated young adults who are hitting the glass ceiling of custom, under investment, and unfair work practices in oppressive regimes. Obviously these people are a source of cheaper labor for French firms (Moroccan women are often used by publishers and other companies needing large-scale text and data inputting jobs performed by French-speaking operatives), yet according to Le Monde, there are far more call centers in France (205,000 jobs and 3,300 call centers) than in Africa, which only gets 5% of France’s call center traffic. So the real effect of this new ruling is hard to guess at, beyond the consumer transparency issue. Especially when a France-located call center probably employs plenty of operatives whyose real names are Mohamed and Mounira already, call centers being one way for the ‘children of immigration’ as they are called to find a job in a country where unemployment is still stuck at around 9.5%.

Most of the comment in France and in North Africa (here, here and here) draws attention to the confused message this ruling gives the consumer – will my call cost more because I hear Habib say he’s in Tunis? – and the faint effluvium of racism implicit in the ‘identify yourself’ requirement.

New language technology evaluation website

2

The industrious Jeff Allen has built a site collecting together a number of useful references in the field of translation technology evaluation. He plans to extend it to the evaluation of other tools in future. At first scroll, it may look like a single author site, but if you keep on going, the list grows more names!

Albanian in Albion and other fantasies

0

The United Kingdom is usually roundly mocked as Europe’s multilingual dunce. And we know why. However, the UK government’s brand new MyBusinessRates website illustrates an interesting trend in what we could call the U.K.’s ’domestic localization’ – localizing to minority languages in specific locations.

This Rates site is designed to offer one-stop access to, well the ‘rates’ (or local taxes) that businesses have to pay. What’s interesting about it is that it has been launched in 10 languages – from Czech, Arabic and Soomaali, to French, Albanian and Russian – in compliance with UK accessibility standards that government websites seem to be spearheading. This is a national site, and with its localization repertoire, presumably addresses the strategic languages of business ownership in the UK.

This gradual shift towards greater accessibility for citizens, using localization, display technology and the like, almost certainly began as a response to very local needs. Take Haringey Council, the local authority for the London suburb that fields the Spurs football team. Earlier this summer it went online with a six-language spread, including Kurdish and Turkish. If you drill down, you naturally end up with English language content etc, but the thrust seems to be in the right direction, and the language range suggests that localization is a genuine budget item. It also proves that those advertisements back in the late 1990s promoting London as Europe’s most multilingual city, with over a hundred tongues blooming on street corners and in back-offices, were on the mark.

But the domestic localization effort goes beyond London. If you check out the website for say the Western England city of Bristol, you will not find a localized site, but the offer of city hall information services in 10 core languages (Bengali, Chinese, Gujarati, Hindi, Italian, Polish, Punjabi, Somali, Urdu and Vietnamese), all European languages, and “other languages such as Farsi, Pashto, Albanian, Croatian, Bulgarian, Bosnian, Asian and African languages.”

This UK localization wave naturally needs benchmarking carefully against comparable efforts in other European cities. This could then generate a score card for localized e-services to EU citizens, and over time prepare the continent for effective e-democracy (voting, etc.), where the question of providing online language availability/ translation would have to be thrashed out in all countries. It would also be necessary to inquire more deeply into localization processes as work on these sites, and how local government is able to budget for them and take them on board.

From a rapid scan, France will not yet be tiptoeing down the domestic localization path – ici on parle français. Spain is already multilingual by definition, but on top of regional locales there is a fairly high rate of immigration by people speaking Arabic, Romanian, English, French, Russian, Chinese, Croatian, Ukrainian, Armenian, as well as African and Indian languages such as Pular, Wolof and Urdu. The region of Valencia, for example, has set up a 24 hour interpreting service to handle various incidents and problems. But website localization remain limited in scope to European languages. The same goes for Germany, where Frankfurt and Berlin provide German and English only. Finland’s Helsinki city hall site is in Finnish, Swedish, French, German, Russian and English, but it seems to be a more tourism oriented site. In Denmark, Københavns Kommune, the capital’s city hall, only has an English option, as does Holland’s Rotterdam, despite the fact that these countries (along with Sweden and Germany) host numerous Asian, Middle-Eastern and African residents who would probably like to be informed (if second class) citizens. Is anyone out there monitoring government/local authority sites for language accessibility?

Arresting news from both sides of the Atlantic

0

Two contrasting ways of leaping the language barrier when you’re busting crime. First old tech. In Norwich (UK), shopkeepers ganged together under project Siren. Part of this effort involved having foreign students speaking ‘East European’ languages ‘doing some translation work’ on flashcards that would contain warnings in various languages. The idea was to help Siren catch foreign (i.e. mainly Russian) shoplifters who flocked into town and emptied the shelves. No use of professional translators – who might have inquired into the literacy levels etc of the target readers and provided other services to Siren into the bargain. And no use of technology. Why bother when you’ve got all those students who wanted to go the U.S. but have settled for England as a second best?

Over in the in U.S. of course, the same sort of problem is being solved with high tech. Why? No handy student manpower, one can only suppose. In Burnsville, Minnesota, the local police have started sticking a VoxTech Phraselator in their holster to offset the difficulties of getting hold of an operational interpreter (or foreign student) when making arrests or giving warnings to members of the various language communities. The Phraselator can display/speak police type messages in 15 languages, and instruct the addressee to signal yes or no etc with their hands in reply. For a handy journalistic overview of the speech translator device scene in the U.S., read this.

Crosswords at the crossroads with il computer enigmista?

0

Nature reports that Italian researchers Marco Gori and Marco Ernandes have developed Web Crow, a program that automatically reads a crossword clue, then Googles it to get an answer from the web, and works in any (?) language. This spells the end of crosswords as a frontier of human intelligence, rather as chess machines have relegated the game of queens to database technology, even though people will continue to get enjoyment from solving them. Interestingly, the Nature report is headlined ‘multilingual algorithm’, but the exact nature of Web Crow may be more complicated. An Italian report of September 27 says:

Per il momento il programma enigmista funziona soltanto in lingua inglese (it only works in English for the time being). Il prossimo sarà farlo funzionare in italiano (Italian next). “Ma la vera sfida – ha detto Gori – sarà farlo funzionare in tutte le lingue, grazie alla tecnologia dei motori di ricerca (but the real challenge (sic) will be to make it work in all languages, using search engine technology)”.

In other words, the multilinguality dimension is in the total application environment – including search technology – rather than in the core technology.

Question: do all literate alphabetic cultures have a repertoire of crossword type games?