My TEDxLjubljana talk from last November is finally online and if you like languages and don’t mind spending a few minutes on it, you can watch it below (in Slovenian). I have already written about the topic in a previous blog post here.

Linguistics is a big hobby of mine so it’s always exciting for me when people let me talk about it for a whole 15 minutes .

The presentation touches upon the latest developments in mainstream language technology and how all of this matters to “minor” languages.

The video is also available on the TEDxLjubljana site.

And here are the slides in English again:

I had the privilege to give a talk at the second TEDxLjubljana, themed language & communication.

My talk touches upon the latest developments in mainstream language technology and how all of this matters to “minor” languages.

In 100 years, language economics indicates that up to 90% of currently spoken languages may go extinct. Some of them are beyond saving, so it is important that we have the tools to preserve as much unique knowledge contained in these languages as possible. For the rest, we are seeing significant improvements in technologies that are making their daily use more viable.

Special mention goes to Google Translator Toolkit, that – provided that it takes off – may mean an unprecedented increase in the quality of the parallel corpora that Google Translate uses to facilitate a real-time multilingual web.

Also speaking at the event were Primož Jakopin, Maja Vardjan, Jaka Železnikar, Daniela Candillari, Igor Bezget, Janko Božič, Vuk Čosić, Pavel Koltaj and Boštjan Napotnik.

Pictures from TEDxLjubljana

Thanks to the organizers for the great time we all had there and I hope to see the next one really soon!

(and don’t forget to post the videos from the event )

Translate Server Error

Despite a myriad of translation engines and web sites, human translation is still a very important job that requires a lot of experience and knowledge to be done properly.

It’s not just the language.

No-one in their right mind should tackle e.g. legal or technical translations if they’ve never worked in any of the fields (unless the job pays really, really well and it’s worth digging through books and Googling everything ).

Translation work is well paid (and rightly so) – and thus quite expensive for the client.

After all, you usually pay for 1-3 passes of proofreading after the translation, depending on the type of document you are delivering.

And if it’s marketing or other commercial text, the translator needs to know how to write copy, too.

I am sure that in Brussels, where everyone and their dog have a translator in their rolodex (as just about everything is translated into more than 20 of EU’s official languages), this is a pressing issue.

There is a huge demand for cost-cutting through Machine Translation and Crowdsourced Translation – it’s one of those markets where the right solution can make a lot of people rich instantly – but we are not quite there yet.

We are still in the era of machine-aided human translation.

This means that the areas where money can be made right now are:

• Translation corpuses (example: http://evrokorpus.gov.si/)
• Translation memory tools (example: Trados)
• Seamless integration of dictionaries, thesauri, …
• Various tools for translators (e.g. tools that hide HTML and other tags in web pages)

Natural language processing, machine translation and various semantic technologies are already having beneficial effects on productivity in writing, research and language learning (look at Lingro, Zemanta, Google Translate, Systran), but we are still years away from being able to deliver quality localizations without an experienced human translator.

Fun fact: this post would cost about 25-50 EUR to translate by a professional

Disclosure: I used to do freelance translation and nowadays still occasionally do business as an agent and proofreader

In about 5 minutes, Mars lander Phoenix is going to attempt to land on the surface of Mars.
The event is being broadcast live at a number of museum events and at NASA TV here.

Here is a live blog from mission control (also following @MarsPhoenix on Twitter).

*just entered the atmosphere*

It’s exciting to watch this in real-time – it’s sad that space exploration gets so little mainstream attention nowadays.

* altitude 1700 meters … altitude 1000 meters … altitude 150 meters … altitude 80 meters … altitude 15 meters *

Everybody is on their toes …

* touchdown * * cheering *

Apparently, the landing was a success! Congratulations!

Now everybody is waiting for the solar arrays to be deployed and if this goes well, the mission that the NASA team worked on for 5 years will be a success.

Image via Wikipedia

I’ve just received another newsletter from Transparent Language advertising their Video Immersion Software and it got me thinking.

I’ve always been very interested in language learning and finding new ways of doing it more effectively. What I currently see as the main issues of this increasingly important process are:

• Time consumption – it takes considerable time to make significant process in learning grammar, scripts, pronunciation and building a solid vocabulary
• Context switching – learning a new language usually means stopping whatever you are doing, sitting down and starting to think about things largely disconnected from your everyday activities – in my opinion, this is a significant »barrier to entry« for most people

The entire process often consists of the following steps (in this or a slightly modified order):

1. Learning the basic grammar
2. Learning the basic pronunciation (including any »unfamiliar« sounds a language might have)
3. Learning the script (based on whether or not this is a priority, a transliteration might suffice at first, e.g. for Mandarin Chinese, learning the script is an extremely lengthy process that can usually be postponed with the help of pinyin)
4. Acquiring an initial vocabulary (by e.g. learning the Swadesh list for a language)
5. Passive contact – reading books, listening to audio clips, watching movies
6. Interactive contact – conversations with other speakers, IM-chatting, visiting the country and practising on-site

It is a widely recognised fact that actually living amongst native speakers produces the best results.

The reasons, as I see them, are mostly the following:

1. No choice – you have to learn how to communicate in order to survive
2. Constant contact with the language
3. No context switching – you use the language while doing what you would be doing anyway

I think understanding the last point is paramount to being able to build a next-generation language learning tool.

One very low-tech and available way of getting close to it today is by picking up a book you’ve always wanted to read in the language you want to learn and just trying to push yourself through using a dictionary.

For a lot of readers, Paulo Coelho is one of those authors whose books are both easy to read and available in a lot of the world’s languages.

But as you can imagine, having to browse a paper (or even electronic) dictionary while reading a book is cumbersome and most books end up sitting comfortably on the »will finish soon« shelf.

So, what does today’s technology hold in store for us to make a step forward?

Using the rising semantic web technologies, this approach is becoming increasingly viable.

Imagine reading a German book online (or offline in a PDF or other text document). You stumble upon a word you don’t know, you click it, and up pops a translation.

A browser plugin (or screen reader, as used by accessibility software enabling the blind to use a computer) gets the word under the mouse cursor, searches a database and provides the corresponding word in the target language.

Of course, the »searches« and »database« part are the tricky part.

Building a huge database out of hundreds of dictionaries, Wikipedia entries and books may do the trick to some extent, but this is still years away.

One approach I’ve described in another post was using a well granulated corpus, such as the Bible and all of its available translations, to find a passage of text in language B that likely includes the selected word in language A and display it as a reference.

Getting the actual translated word is a bit tricky and requires some statistical tweaking and an active participation from the users, but according to some initial tests I have done this approach looks quite promising.

One factor worth mentioning here is that preventing mistakes is not very crucial, as the most important thing is to give the user some kind of context for a quick conclusion, not provide a literal translation.

E.g., if you’re reading the sentence “Er musste sich entscheiden”, and you click on “entscheiden”, such a system will find the word in a corpus and display the corresponding translation of that text part to you as a hint, you could e.g. have “It was hard for her to decide”, which would help you guess the meaning of the word.

A reader can very well live without knowing the meaning of 1, 2 or even a dozen words on a single page, as long as the surrounding context is informative enough for a probable conclusion on their meaning.

In summary? You are building your vocabulary and probably your grammar while going through an everyday text in a foreign language.

For some, this means zero time waste.

But for most, it is not enough …

A YOUNG LADY’S ILLUSTRATED PRIMER

I have never been particularly drawn to nonsense such as Justin.TV, where a guy walks around with a camera strapped to his head and lets the whole Internet in on his everyday life.

But, as it happens, having a portable camera connected to the Internet with you at all times just might become something we won’t mind doing at all.

Please note that these are my impromptu musings and I might come back with one or two references or alterations at a later time.

Those of you who are familiar with Neal Stephenson’s Diamond Age know about the concept of “ractors”, ad lib actors that jump right into a computer generated environment and act out a live scene to give their clients (who are also participating) a sense of interactivity and liveliness.

Neal Stephenson goes to considerable lengths to describe the way these ractors are hired for the jobs – it reminds me (to some extent) of Amazon’s Mechanical Turk – using a computer application that notifies them of available jobs, they pick one and in a couple of seconds, they’re either racting out a particular role or – narrating.

And this is where I am trying to get to –

Imagine waking up on a sunny day, putting on your camera or video-capturing sunglasses, walking out into the street, and listening to your day’s story – in German.

“Bob walks out into the street on a sunny day, looking at the people passing by. He stops in front of a building, opens a large wooden door and steps in. He is now inside a restaurant”.

The narrator/ractor sees what you see, hears what you hear and describes it to you in a language you want to learn.

You pay them by the minute. They can either work from home or from an office, all they need is a broadband connection and a headset.

Privacy/legal issues? Probably lots.

But there are some measures that come to mind.

First of all, you can keep the ractor-ractee pairing such that longer “revisitations” are at a minimum. Client-relationships such as the one between Nell and her ractor are a complex matter.

Also, strict guidelines on how and where the ractor performs the job are required.

But all in all, I’m sure that a couple of minutes of exposure and a couple of bucks off your credit card are not much to ask for the probably best and most effective language learning method ever.

After you are done, you rate the ractor/narrator and say goodbye, as the next time, it might or might not be the same person watching.

Image via Wikipedia

A while ago, when I was beginning my struggle with Finnish, I found it quite difficult to find translations online, since Finnish is a highly inflected language and word roots are not easy to identify for a beginner.

After having googled enough Finnish words, I was starting to see a possibility – almost every search yielded an excerpt from a specific source.

I had never read more than a few pages of it before, but I soon saw that the Bible is the perfect language resource:

• It’s available in almost all languages
• Text is highly consistent across most languages
• It’s very well structured
• The smallest unit of text in many parts is 1-3 sentences long

The last point is very important as it makes the book ideal for being used as a text corpus.

Let’s take the word kansalle, for example.

If you google for the search term “kansalle finnish bible”, the first result gives you Kings, Chapter 12.

12:6 Kuningas Rehabeam neuvotteli vanhain kanssa, jotka olivat palvelleet hänen isäänsä Salomoa, kun tämä vielä eli, ja kysyi: “Kuinka te neuvotte vastaamaan tälle kansalle?”

And the English version:

12:6 And king Rehoboam consulted with the old men, that stood before Solomon his father while he yet lived, and said, How do ye advise that I may answer this people?

So, kansalle clearly means to (the) people, the suffix -lle being an indication of the allative case.

I am currently working on a system that would automate this.

Wouldn’t it be nice to have a Bible phrase in both languages pop up whenever you searched for a word? What do you think?

Links:

Biblija.net – a magnificent corpus of Slovenian, English, French, Dutch, Spanish, Catalan and Euskara (Basque) bible versions

A list of Finnish cases – the best explanation of the cases I could find

Finnish Bible

I’ve always been fascinated with this sleek and elegant language.

Luckily, there is an abundance of resources online to fast-forward a learner’s acquisition process.

As with any other language, I’d promote the following steps:

1. Start with an introductory course and get to know the basic grammar and pronunciation
2. Next, get a Swedish children’s book (e.g. this) and read it a couple of times using a dictionary
3. Get some Swedish friends online and practice your conversation skills
4. Watch a good Swedish film (e.g. Ondskan) – start slowly, Swedish pronunciation is the tricky part
5. Spend some time in Sweden
6. Get free Swedish books online and read them side-by-side with their English counterparts
7. Listen to Swedish podcasts, watch Swedish news, watch Swedish TV channels

A quick list of resources:

• Lexin - Swedish Dictionary (English, Albanian, Arabic, Bosnian, Croatian, Finnish, Greek, Kumandji, Russian, Serbian, Somalian, Sorani, Spanish, Turkish) – a perfect learning aid, I even use it to learn Finnish
• Project Runeberg – a Swedish version of Project Gutenberg (includes texts in Norwegian, Danish and English)
• BeelineTV – watch TV channels from various countries online
• Online courses: link1, link2, link3, link4

Anyway, låt mig veta hur det går!