30 Eylül 2012 Pazar

Sweet Words vs Monstrosities

To contact us Click HERE

More than a century ago, Henry Sweet wrote The Practical Study of Languages and through it, criticized the existing methods of the day, much as we still do now.   The book's myth-busting objectives reviews phonetics, alphabets and pronunciation issues before diving into methods, grammar, vocabulary and texts. 

In fact, while scanning through the text, I honestly couldn't help but think I bet he'd have been a blogger if he were around today.  His prose is tight, easy to read and the language direct.

His obvious annoyance at the 'insufficient knowledge of the science of language' (1899:3) like my own, literally jumps off the page.  Given that this post is part 2 of No Evidence for a Fixed Aquisition Order, I'll hone in on this one quote which I wanted to share with you, for reflection, as it neatly wraps up the debate on authenticity vs manufactured texts:

...the dilemma is that if we try to make our texts embody certain definite grammatical categories, the texts cease to be natural: they become either trivial, tedious and long-winded, or else they become more or less monstrosities' (1899:192).

Really sounds like he was describing Headway long before it ever arrived to influence all the other copy-cat productions from then on and into today.  The question is though, will it influence tomorrow's or can we teachers at least try to stop it before it does?

Best,
Karenne

Image credit
Wikimedia commons, wolf in sheep's clothing

Reference
Sweet, H. (1899). The Practical Study of Languages.  London, UK. J.M. Dent & Co.
(Available for free online from Google Books)

Extrinsic "VS" or "AND" Intrinsic Motivation?

To contact us Click HERE
In the last couple of weeks prior to restarting classes, I've been watching a lot of television. The weather's not been particularly nice and I'm too poor to do anything else.  That's the downside to being a student at my age, I guess.  The upside is: imagine the best conference you've ever been to and think of one of the great presentations - one that has really had an impact on your teaching... now imagine that instead of a 45 minute session you get to have access to months of amazing lectures, group discussions and articles to read to follow up and challenge yourself with.  So, poverty is worth it, I guess.

But anyway, back to TV, one of the theme songs, from Weeds, has become a real earworm.  It goes, for those of you who don't know it,

"Little Boxes, little boxes, on the hillside, little boxes all the same
There's a green one and a pink one and a blue one and yellow one 
And they're all made out of ticky tacky and they all look just the same..."

The song makes me think about motivation, a lot.  Or maybe I was thinking about it before the song and it just drummed it in.  We all know that learning doesn't happen without motivation.  But what is it really?   Where does it come from?


Usually, it gets boiled down into three categories:


Extrinsic Motivation (external influences)
e.g. money, rewards, good grades, trophies, certificates, job position

Intrinsic Motivation (internal influences)
e.g. enjoyment of a task, passion, a drive to seek challenges, autonomy, inherent satisfaction

Intrinsic motivation refers to doing something simply because it is enjoyable while extrinsic motivation is more about getting a specific value or outcome based on what you have done (Ryan and Deci, 2000).

Amotivation is basically when you can't be bothered.




It has also been determined, through extensive empirical research by Deci & Ryan, Vallerand and others over the decades, that extrinsic rewards put a damper on intrinsic motivation.   I think though, that we have to be a bit cautious with this sort of thinking as it could very easily lead one into an assumption that extrinsic motivation is bad and that intrinsic motivation is best.  A dangerous position I feel, because for the most part, whether we like it or not, our adult language learners are more likely to come to us extrinsically motivated than intrinsically.

They want to learn English to integrate into society, to get a job promotion, to ensure job security, to get a better pay cheque, to speak to their foreign colleagues and close the deal.  If not this then they want to know that when they go on holiday, they won't get lost.  Sure, there are a handful of housewives who just fancy learning it, but usually because someone else told them it is the "thing" to do. And the teens mostly just want to pass the course, get the certificate, and get on with life.

So where it all gets a bit sticky for me, is that sometimes our extrinsically motivated learners really enjoy learning.  Why not, after all?  Sometimes we teachers can inspire them and sometimes their colleagues do and sometimes they develop an interest for the language - but all this interest and high from learning a second language does not take away their primary extrinsic goals.

In more recent research, Ryan and Deci have made a point of re-examining extrinsic motivation more closely, placing extrinsic motivation on a continuum and have created this taxonomy:



The idea is that learners can be in a state of external regulation (wanting rewards or avoiding criticism), or one of introjected regulation (constraints are internalized and set by the learner).  Identified regulation means that the behavior is thought of as being self-determined and finally the last type is integrated regulation - the person learns willingly because it fits in with the rest of the life activities and life goals (Vallerand, 1992).

Despite the fact that there is so much literature on extrinsic and intrinsic motivations and I'll continue reviewing it all, I really can't help but wonder if motivation is not actually something quite fluid. Can't you (or our learners) be one type and the next day, another?


But more importantly, if by categorizing motivation into boxes and then onto further hazy sub-boxes, might we be missing out on the fact that humans are infinitely complex creatures who can be both intrinsically and extrinsically motivated at exactly the same time?


What d'ya think?


Best,
Karenne

image credits:
plant fondo oscuro by eric caballero

References:
Ryan, R. & Deci, E. (2000). Intrinsic and Extrinsic Motivations: Classic Definitions and New Directions. Contemporary Educational Psychology. 25, 54-67.

Vallerand, R. & Bissonnette, R. (1992). Intrinsic, Extrinsic and Amotivational Styles as Predictors of Behavior: A Prospective Study. Journal of Personality. 60:3. 599-620.

English: time 4 a revamp?

To contact us Click HERE
Nothing so amazes me more than the fact, that despite so many other languages havie large governing bodies which analyze, stay on top of and make changes to their language in order to better fit the times, that English doesn't.

I think we should, especially as its reach extends across the globe.

If I could change the English language., then I would





- add an extra grammatical tense:  The "Ever Present" Tense
  • it annoys me somewhat to tell students to use the present for "habits, permanence, facts." If it's for all time, then there should be a specific tense that refers to this because for most people, present = now.


- I'd add two extra pronouns to reflect gender reality.
  • heshe and shehe


- I'd also love to revamp spelling entirely to make it better reflect the way words sound
  • if the ch sounds like an sh, it should be an sh
  • regular past tense endings are a waste of time teaching.  Why not write workt not worked, filld not filled - loaded can stay loaded.

What do you think?  Any bug-bears you've noticed while teaching our fair language?  What would you change if you could?


Best,
Karenne


image credit
Teaching an old dog new tricks by Fouquier on Flickr

Why I don’t like Second Life (by Jacqueline Goulbourne)

To contact us Click HERE


Imagine a world where you can make a cartoon avatar of yourself and do whatever you like in an international community of English speakers.

Well, it already exists, in Taiwan where I have spent most of my career teaching,  it's called ‘World of Warcraft‘ and the mission of almost every parent of teenagers there to wrest their kids off of it!  But imagine if something similar existed, primarily for education, business and sex.  That exists too and I’m fairly certain that it’s nothing we’d really want to get students involved in.

I’m sure many of you disagree and are using Second Life in your practice to good effect, so my intent is not to disrespect your work, but to explain the problems I have with Second Life and other such virtual worlds.

I first experimented with Second Life in 2003 when friends of mine had set up an experimental teacher training space.  I didn’t really get interested.  More recently I came across Second Life in a wee teacher training course I’m doing.  I fought against my inherent tendency to not engage with things that don’t resonate with me and decided to ‘join in’ and not be such a negative Nancy.  Although I have an aversion to ‘virtual worlds’ and to speaking through machines in any way (I don’t use telephones) I put this aside in order to try and retrieve any babies in this bathwater.

I Googled up Second Life and was met with this screen:

[screenshot from https://join.secondlife.com/ ]
The user sees this screen with avatars that can be chosen.  Your avatar can be customised later on but I immediately took real exception to the prototype avatars that were presented for customisation.  These include a slightly scary rabbit, a robot and lots of white kids who clearly resembled extras from a 90s vampire film.  The only black woman present is wearing a short dress with a flower in her hair and high heeled shoes - a very different aesthetic to the white women presented.  One of the men has a shirt open to show rippling pecs but generally, the male figures are covered up.  For me, it doesn’t represent an acceptable aesthetic to present to students in my care, particularly young women.

There are two strands to my  misgivings: firstly the absence of ethnic diversity and secondly, the overt sexualisation of young women, in the Second Life avatar choices.   As a teacher joining in, I tried to choose an avatar that could represent my self without making myself ridiculous.  I’m 37, pretty fat, greying, and generally get dressed in the dark.  My choices ranged from avatars that mostly looked like a 14 year old version on myself: pale skinned, bone thin, and dressed in black with lots of eyeliner and it seemed undignified and self-abasing. 

I’m not scarred by the experience. But then I’m a confident, English-speaking adult.  I can articulate why I don’t want to be this 1990 version of myself and tried to change it.  Can we expect young, perhaps not so good at English students to raise their own misgivings to their educators or will they simply go along with what the teacher, or the authority wants them to do?  Thinking back to my time teaching junior high school girls in Taiwan, none of those prototype avatars is of Han Chinese or any other Taiwanese ethnicities.

What’s the message here?

White people (and the token African American) own English. You are different. Go get yourself a white identity to join our English-speaking world!

OK, yes, you can choose to brown/yellow/black up later if you want to, but that’s not the standard issue human in this community.

Don’t we want our students to come and sit at the table as equals? To join the English-speaking communities that they are passionate about?

In my Real Life classes, every kid is a beautiful prototype of the individual they want to become as they are: be they fat or thin, with braces, wearing unsexy clothes - with a wonderful inheritance of Chinese ethnicity, not something to be tagged on later, once they have chosen their ‘core’ avatar.

Am I over-thinking this?

Perhaps, but if we use these tools in the classroom, we are also raising the question of who ‘owns’ English.  We have to ask who funds a lot of these communities and to what business or political ends?  To promote ELT as a British, Australian, White American/European activity?  Why?

Is that congruent with our principles, desirable for our personal teaching contexts?

Cartoon images hurt as much as photographic images.

As a teacher I’m rarely didactic and I know I can’t change the world, but I am absolutely committed to making every child  in my care feel like they can be whoever they want to be and that includes valuing and celebrating every child’s individuality and identity and not promoting certain images as a norm within an educational context.



Jackie has been a teacher for 15 years, around the world,
but mostly in Taiwan.

Those Greedy Publishers (by Sue Jones)

To contact us Click HERE
It's a super honour for me to present a special guest post for you on the issues of finance within the ELT publishing industry, by a true veteran of the business.  The post is in response to a discussion we were both a part of in the IATEFL Young Learner Yahoo Group and I'm delighted to host it here.




Are you amongst those who can’t even look at a textbook without feeling annoyed?

Let me try and explain why you needn’t feel like this, even if you don’t like text books, never use one and may have personal views about individuals in the business as well!




How did textbooks start?

Usually because some smart teacher arranged,organized, sequenced, sliced and diced and generally managed the material to be taught in a way that helped her, produced results, and her colleagues found it helpful and time-saving as well. So those smart teacher-publishers rapidly became rather richer, so of course they started to look around for other subjects to apply the same approach to.

To do this, they learned how to deal with the mimeograph machine, or haggle with printers, how to store and distribute in practical and economic ways, howto find other teachers who could assemble content in a meaningful way to other teachers, how to find others who could manage and improve these teachers’ work, and so on. Thus publishing companies were born.

In that original print-based world, an initial one-time investment is made in the creation of the fixed format of the book, and then of course the more copies sold, the more attractive that fixed amount looks when spread over a very large number of copies, amortizing (paying off a debt over a period of time) the initial investment to reach a break-even point, after which it’s all profit.


The cost of printing

Of course, there are some costs related to each individual book – the cost of printing, and the royalty paid to the author. Those two sums have to be paid on every single book. But there’s an upside to printing - the larger the number of copies printed, the cheaper each one becomes, because the key cost in traditional printing is associated with the initial setting up of the print run – cleaning printing plates and presses, making and mounting the film and so on.

This outlines the basic print publishing business model – find out and offer what people really feel is valuable to them, invest a fixed sum once to create it, and replicate it at reasonable cost many times over.

So two different sorts of publishing investment is required. One is a fixed, one-time investment (the editing, design, planning, research etc to get to a point where there is something to print) and the other is a variable cost, which depends on the number of books printed and sold (the per-copy printing cost and author royalty cost), and which is paid on every single copy.

This means it is extremely difficult, not to say impossible, to say exactly what each individual book costs to create. If the book sells well, the amount of the fixed, pre-press investment will look quite small on a per-book basis, and of course the converse is true.

A wrinkle in this picture is that the costs of some of the items in the pre-press investment may be considered overhead – the salaries and overhead connected with maintaining editorial, marketing and sales staff, for example. This overhead is probably not taken into the costs for each individual book.

Of course the publisher’s goal is to invest once, cover all those costs as well as contribute to the cost of maintaining the necessary overhead, have the cash flow to continue to be able to print as required, and when all that has been paid for, to have a profit – a surplus after all the bills have been paid and the revenue from sales has come in. It sometimes works and it sometimes doesn’t. So it helps to have a lot of books, so that some are moving into the market and will take a little while to become profitable, while others are at their peak and yet others declining.


Complexities to costing

While this business model is essentially simple, there are complexities this picture doesn’t capture. For example, the publisher is almost certainly not the final sales point for the book. This may be a local retailer in the international market, a local education or school district distribution point in some countries, or Amazon etc. To make the book available to the buying public, the publisher offers a discount, which inevitably means the price the ultimate consumer pays is not the price the publisher receives. In overseas markets, the publisher may give 60% or 70% discount, so the amount the publisher receives from what the customer pays may be only 30% or 40%.

In addition, publishing is a very time-sensitive business. Books must be available when schools open. Often teachers won’t commit to a book without seeing that the whole series is available. The effect of all this is that the publisher has to make all investment up front, before a single book has been sold. If there has been any delay in receiving payments from the previous year (more common than you would think), or if a book didn’t sell as well as expected, the publisher may need to buy cash flow from the bank. This of course adds to the publisher’s costs.

So if you pay £20 in a bookshop, the publisher will receive between £14 (if the discount is only 30%) and £6 (if the discount is 70%). Nowadays most books are targeted at specific local markets and it is rare for the exact same book to sell throughout the world - there are exceptions, mostly amongst books published before the mid 90s, and there’s a reason for this, which is that the massive increase in computer and desk top publishing since the mid 1990s has made it much easier to be a publisher.

This led to massive fragmentation and increase in numbers of publishers serving their own specific, local market. This increased level of competition leads in turn to greater sales and marketing cost, which may now be 5-7 times greater than editorial overhead costs. In attempting to manage this risk, publishers avoid anything experimental or outside the normal publishing pattern, leading to a lot of over-similar books. (Perhaps in any case be argued that since books are chiefly used by mainstream, time-poor teachers, this is the area of greatest demand – adventurous teachers can and will create their own experimental materials.)

The end result

The result of all this is that negotiations for adoptions rapidly become focused on price or added value (additional give-away items, seminars, sometimes even downright bribery which is of course now very severely punished under new UK legislation). So although there are of course authors who have made a good living, there are no longer authors who have literally made a fortune.

The publisher will already know what the expected price will be and have made all calculations with the likely end revenue in mind. It’s likely that the print cost will be in the region of £1 – this varies according to the size of the book (format), number of pages (extent), the number printed (print run), number of colours, weight and type of paper, type of cover. The author royalty will be in the region of 10% of the final price the publisher receives. After that, the publisher needs to use the balance to cover all other costs – fixed pre-press investment, overhead and other costs, some of which I have mentioned above. So the picture of profitability (or loss) only really emerges over all the years the book is actively in the market.


Still think they're so greedy?


A 35-year veteran of ELT publishing, Sue sees this career as three distinct stages – then, in between, and now. ‘Then’ was when ELT course materials were special, valuable things, so there were fewer of them around and they were easy to sell (which suited the hedonistic lifestyle she enthusiastically pursued as a rep in Greece). ‘In between’ got going in the 90s, when the wide availability of computers removed barriers to entry to the publishing business. For a lengthy period Sue published materials for Latin America after a spell Asia, based in Hong Kong. ‘Now’ has seen periods as managing director of two different publishers in the US and UK, as publishers struggle to identify a way forward into the flexible, creative, digital world they’d like to be in, but which hasn’t really settled into identifiable business models with clear market segments. 
imagecredit: Fat Cat, Russell Heisteman

29 Eylül 2012 Cumartesi

Ах, Александр Сергеевич, милый .... (с) or Pushkin's Birthday

To contact us Click HERE
Today, on the 6th of June ,we celebrate Alexander Pushkin's birthday. Only lazy one didn't mention it today in his account anywhere. This just proves that Pushkin is indeed "наше в�е" - 'our everything', as we use to call him now, following the phase of some critic.

Indeed, if you ask many interesting and talented professors in Russia 'who is your favourite writer', many of them will say 'Pushkin...', and that is of course not because they didn't read anything else! Pushkin is indeed greatest poet and a real treasure of Russian literature

Let's see why. Let's read the quotations about Pushkin by the writers and critics of his time  - here we will find interesting facts about him


При имени Пушкина тотчаÑ� оÑ�енÑ�ет мыÑ�ль о руÑ�Ñ�ком национальномпоÑ�те. Ð’ Ñ�амом деле, никто из поÑ�тов наших не выше его и не может болееназватьÑ�Ñ� национальным; Ñ�то право решительно принадлежит ему. Ð’ нем, как будтов лекÑ�иконе, заключилоÑ�ÑŒ вÑ�е богатÑ�тво, Ñ�ила и гибкоÑ�ть нашего Ñ�зыка. Он болеевÑ�ех, он далее раздвинул ему границы и более показал вÑ�е его проÑ�транÑ�тво.Пушкин еÑ�ть Ñ�вление чрезвычайное и, может быть, единÑ�твенное Ñ�вление руÑ�Ñ�когодуха: Ñ�то руÑ�Ñ�кий человек в его развитии, в каком он, может быть, Ñ�витÑ�Ñ� чрездвеÑ�ти лет. Ð’ нем руÑ�Ñ�каÑ� природа, руÑ�Ñ�каÑ� душа, руÑ�Ñ�кий Ñ�зык, руÑ�Ñ�кий характеротразилиÑ�ÑŒ в такой же чиÑ�тоте, в такой очищенной краÑ�оте, в какой отражаетÑ�Ñ�ландшафт на выпуклой поверхноÑ�ти оптичеÑ�кого Ñ�текла.Ð�.Ð’. ГогольWhen I think about Pushkin,I immediately think of the idea of ​​theRussian national poet. In fact, none of our poets is above him. In him, as ifin the lexicon, is combined all the wealth, power and flexibility of ourlanguage. More than anyone else, he further extended the boundaries of the language and showed all itsspace. Pushkin is a phenomenon, unique phenomenon of the Russian spirit: he is aRussian man in his development, as he might be in two hundred years. He is ofRussian spirit, the Russian soul, Russian language, Russian character reflectedin the same purity, in such beauty, in which the landscape is reflected on theconvex surface of the optical glass.NV Gogol
It isdifficult to estimate the contribution of  Pushkin in the formation of Russian literatureand Russian language. This is already the undisputed truth, and confirmed by awide variety of studies.


Russianliterature and Russian language is constantly experiencing the impact of thepolitical situation in the country. Depending on who rules a great power, beganto set up their own rules and regulations that speak about how to write or saywhat books will be sold and which will fall under the ban. Language  assimilated, feeling the impact of populardialects, as well as European language. Gradually, language and literatureacquired properties of other cultural traditions that took root in Russia, butall were strange. Everything has changed since then, as theliterary arena joined Alexander Pushkin. In his works, the author uses thepopular lexicon. Despite the fact that he is considered a good connoisseur ofFrench language, is widely established in various areas, Pushkin refersspecifically to the people's speech, using specific words or concepts. Thus, hebrings in a whole new language of communication elements that have beenforgotten or replaced by a foreign vocabulary. Pushkin understood that theRussian language - is primarily a national treasure, and however  it had been changed and distorted by the greatminds of his time, ordinary people will still speak the national language –same that had been used by their ancestors.
 In addition, the author introduces thelanguage of a large variety of beautiful epithets and metaphors. He imbues hisworks phraseology and expressive epithets, the writer tries his readers not thinkstraight, and try to see the image and beauty of language and words denoting theusual subject and ordinary action. The literary language of Pushkin became aclassic of Russian language, which was later followed by other writers. Pushkin wasthe first Russian writer who began to make experiments with the stylisticorientation of the text. He never followed the strict framework of genre whenwriting texts. Pushkin is more appealing to the notion of style, highlightingthe common features of it, playing with it, creating a unique product. Inaddition, Pushkin laid the foundation of a new stylistic direction in Russia -realism. All the writer's works are multi-faceted and diverse.
ReadingPushkin, you hear his voice. He was the first poet who destroyed the image of aconvention of copyright: in "Ruslan and Lyudmila," "EugeneOnegin", Pushkin speaks to the reader "on an equal footing", asit is between close friends, all grasping at a glance 


За�луги Пушкина перед Ро��ией велики и до�тойны народной признательно�ти. Он дал окончательную обработку нашему �зыку, который теперь по �воему богат�тву, �иле, логике и кра�оте формы признает�� даже ино�транными филологами едва ли не первым по�ле древнегрече�кого; он отозвал�� типиче�кими образами, бе��мертными звуками на в�е ве�ни� ру��кой жизни. Он первый, наконец, водрузил могучей рукою знам� по�зии глубоко в ру��кую землю...И.С. Тургенев
The merits of Pushkin for Russia are  great and worth getting national recognition. He gave a final transormation to our language, which is now by its richness, power, logic and beauty of form is recognized even by foreign philologists as almost the first after the ancient Greek.  He was the first, who finally hoisted the banner of poetry, deep in the Russian land by the mighty hand...IS Turgenev



source of photos - vk.com
text (Russian) here and here

21 June, Victor Tsoy's birthday

To contact us Click HERE
Only a lazy one doesn't know Tsoy...
Виктор Цой, Soviet rock-musician, poet, leader of the group KINO, has become a symbol of the epoch that 'was waiting for changes'

Tsoi contributed a plethora of musical and artistic works, including ten albums. He died in a car accident on August 15, 1990, aged 28.



At the age of 17, Tsoi began writing songs. In the 1970s and the 1980s, rock was an underground movement limited mostly to Leningrad; Moscow pop stars ruled the charts and received the most exposure from the media. However, rock music was not popular with the government. Thus rock bands received little or no funding, were not given any exposure by the media. By this time Tsoi had begun to perform the songs he wrote at parties.Tsoi went to underground concerts of Leningrad rock musicians. After a Boris Grebenshchikov solo concert he returned with Grebenshchikov on an elektrichka train from Petergof to Leningrad and played two of his songs to him. Grebenshchikov, who had already been a relatively established musician in the Leningrad underground scene, was very impressed by Tsoi's talent and took him under his wing and helped him start up his own band. This signaled the beginning of Tsoi's rock music career.Kino's impact on Soviet music and society was huge. The group introduced a sound and lyrics that no other Soviet artist before them was able or willing to produce. Kino opened the doors for modern Russian rock bands. It's displayed today in many places around Russia, from graffiti on the fences of St. Petersburg to an entire wall dedicated to Viktor Tsoi in a bylane of the famousArbat street in Moscow, where fans still gather to remember their hero. In 2000 some of the nation's top rock bands came together and released their interpretations of Kino's best songs as a tribute to Viktor Tsoi on what would have been his 38th birthday. Even though he is gone, Viktor Tsoi still lives in the minds of many Russian youths.There are lots of web-sites dedicated to his music and song. Like the fan's site http://vitya-tsoy.ru/ where you may listen to the music and find his texts or watch some videos with the legend of Russian rock or web-sites with the info about all his albums etcThe most symbolic songs are for sure these one
'The star named the Sun'

Listen or download КИ�О Звезда по имени Солнце for free on Prostopleerand 'We are waiting for changes'

Listen or download Кино Мы ждем перемен for free on Prostopleer
Here I'd just share some of my most favorite songs! 

Listen or download КИ�О Пачка �игарет for free on Prostopleer
Listen or download Кино В.Цой Кукушка for free on Prostopleer
Listen or download Кино �люминиевые огурцы for free on Prostopleer

С днем рождени�, Вит�!


Little men by 'Happiness supplier no.1'

To contact us Click HERE



The author of this picture is Eugenia Gapchinskaya, Kiev artist who has made happiness as her brand, and describes herself as ' happiness supplier number 1 ." Pictures of Evgenia are cheerful and bright, inhabited by small funny and touching people



Eugenia Gapchinskaya bornin 1974 is a Ukrainian artist,painter, illustrator of children's literature.


In 2008, the Ukranian post has releaseda series of 12 stamps, "Zodiac" with herworks.
Annually she holds over a dozen newexhibitions in Ukraine, Russia, France, Belgium, England, theNetherlands and other countries. He has several of his own galleriesin Ukraine and Russia. Her works are in European museums and privatecollections of fans and artists.


Her style is very recognizable. Andeven if she thinks of experiment and something new, it is just aboutsomewhere inside herself, at thelevel of these strange little men from herpictures. 'This is the way I thinkh –thourgh these little beloved men', says the artist











personal web-site http://gapchinska.com
photos are taken from 2photo



Cinema club: July, 17 at 7 pm, 'Hipsters'/ "Стиляги"

To contact us Click HERE

RUSSIANLANGUAGE CENTER invitesyou to the 
RUSSIANCINEMACLUBJuly, 17 at 7pm
HIPSTERS/ Cтил�ги(2008,Russia,120 min)
Director Valery TodorovskyStarring Anton Shagin, Oksana Akinshina
Stilyagi (СтилÑ�ги) is a 2008 film, named Hipsters for its American release. It is a musical that deals with Soviet youth subculture "hipsters" or literally "obsessed with fashion" of the 1950s.Mels, a member of Komsomol (the youth wing of Soviet Communist Party), helps break up a hipster’s party. He briefly meets and is intrigued by a hipster girl named Polly, who invites him to hang out with her friends on "Broadway." Mels is drawn to Polly and seeks to win her over by becoming part of their world of colorful fashions, dancing, and loud music. He begins to adopt their fashions and even purchases a saxophone off the black market which he learns to play in the illicit jazz style…



  • audience favorite at the Toronto International Film Festival,
  • audience favorite at the Nashville Film Festival,
  • audience favorite at the the Cleveland International Film Festival, winner of the Audience Choice Award at the Anchorage International Film Festival
  • Golden Eagle Awards
  • Best Film at the Nika Awards

VENUE:the classroom of RLC, 701,7/F, Arion Commercial Centre, No. 2-12 Queen's Road West, SheungWan
Admissionis free,however please write to us (2teacher@rlc.hk)or call us (25988389)to book a seat.

Master and Margarita

To contact us Click HERE

TheMaster and Margarita («ÐœÐ°́Ñ�тери Маргари́та»)is a 1967 novel by MikhailBulgakov,woven around the premise of a visit by the Devil tothe fervently atheistic SovietUnion.Many criticsconsiderit to be one of the best novels of the 20th century, and the foremostof Soviet satires, directed against a suffocatingly bureaucraticsocial order.

 `I  am  a master.'  He grew  stern and  took  from  the pocket  of  hisdressing-gown a completely greasy black cap  with the letter 'M' embroidered on it in yellow silk.  He put this cap on and showed himself to Ivan both in profile and  full face,  to prove that he was a master. `She sewed it for me with her own hands,' he added mysteriously.
(chapter 13)

'She  was  carrying repulsive, alarming  yellow flowers in  her hand. Devil knows  what they're called, but for some reason they're the first to appear in Moscow. And these flowers stood  out clearly against  her black spring coat. She was carrying yellow flowers! Not a nice colour. She turned down a lane from Tverskaya and then looked back. Well, you know Tverskaya! Thousands of people were walking along Tverskaya, but I can assure you that she  saw me alone, and looked not really  alarmed, but even as if in pain. And I was struck not so much by her beauty  as by an extraordinary loneliness  in  her eyes, such as no one had ever seen before! Obeying this yellow  sign, I also turned down the lane and followed  her.  We walked along the crooked, boring lane silently, I  on one side, she  on  the other. And, imagine, there was  not a soul in the lane. 
(chapter13)
Read the book here (English or Russian)The info about the book - (c) wikipediaThe pictures are from the TV-film 'Master and Margarita' (2005) by V.Bortko, from the web-site www.kino-teatr.ru and www.kinopoisk.ru

28 Eylül 2012 Cuma

Nonlinear dimensionality reduction by locally linear embedding

To contact us Click HERE
In this paper, the authors want to find the low-dimension representation of the observed data which considers the original data structure. In real world, we receive lots of stimuli at the same time, such as  the pixel intensities of images, the power spectra of sounds, and so on. These oberservations typically have a much more compact description, and are lied on or close to a smooth low-dimensional manifold.  Thus, in this paper, they assume the data are generated on some low-dimension manifold as Fig1 shows.






















Traditional method, such as PCA, cannot preserve the distances on the manifold and might lower the performance. LLE (locally linear embedding) wants the find the axes along with the manifold, as figure(c) shows.
























Figure 2 is the overview of the algorithm. I will describe them in detail in the following :
For every point, they first find K-nearest neighbors, and use these neighbors to learn the weights which minimize the lose function. (Step 1 in Figure 2)






These weights can be found by solving a least-squares problem.  (Step 2 in Figure 2)
After finding these weights, try to find the low-dimension vector Yi representing the axes on each point.
This can be done by minimizing the following function






That is almost the same as the function in Step 2. The only difference is that the fixed values are Wij now, and we want to find suitable Y. This can be solve by finding the eigenvalues. (Step 3 in Figure 2)
After finding the new positions of all points, we can put all the points on the low-dimension space and the result is like figure 1(c).

They do the experiments on the facial images and text documents, and I only show the result of facial images here.
This figure is generated by projection all the points on the ï¬�rst two coordinates, and the bottom images correspond to points along the top-right path (linked by solid line). We can find the expressions of these facial images change smoothly. That means, this dimension reduction method really somewhat preserves the semantic distances. 
========================================================Comment:
The authors first observe the behaviors the data. Data often have some correlation between certain dimensions of the high-dim space. That is, data can't expend all the space, and they want to find the underlying structure. They find the distance on the manifold is almost the same as the euclidean distance between near points, and assume this property holds for the data. Then, use this property to resolve the underlying structure. That is a good example for solving the problem by observing the property of the data. 

A Global Geometric Framework for Nonlinear Dimensionality Reduction

To contact us Click HERE
The goal of this paper is almost the same as the paper I read last week (Nonlinear Dimensionality Reduction by Locally Linear Embedding). Actually, they are coming form the same year Science magazine, and the position of this paper is one former. The same concept is derived from two different teams at almost the same time. That is quite interesting.

In this paper, they also want to find the low dimension underlying manifold structure from high dimension space.
The example is also "Swill roll".














"A" is the original data structure in high dimension space, and they want to find the underlying structure and expand them as "C".
They also assume that geodesic distance can be approximated by adding up a sequence of “short hops” between neighboring points (This concept is the same as previous paper).
However, they use different method to solve the underlying structure.
They proposed isometric feature mapping (Isomap) algorithm to solve it.The following table is their algorithm overview.




















At the first step, they find the neighbor points for every point (the distance of them less then a threshold or found by KNN) . And construct a weighted graph G over the data points by connecting their neighbors.
Every point is viewed as a node in G, and the weight of one edge is defined as dx(i,j). dx(i,j) represent the distance between near points i and j.
After constructing this graph, we can compute the shortest distance of every two nodes by dynamic programming method (Floyd–Warshall algorithm). They denote this distance as dG(i,j).
Then, we know the distances of every two points in the underlying structure. Thus, we can start put them in the new low dimension space. They use classical  multidimensional scaling (MDS) to put the points into d-dimension  Euclidean space Y that best preserves the distances on the graph.
It can be fulfill by minimizing the following cost function.





Where DY denotes the matrix of Euclidean distances {dY(i,j) = |yi -yj|}.  The Tau operator converts distances to inner products  that supports efficient optimization. This function can be solved by setting the coordinates yto the top d eigen-vectors of the matrix Tau(DG).
The true dimensionality of the data can be estimated from the decrease in error as the dimensionality of Y is increased. That is shown in the figure below (The true dimensionality of underlying structure is marked by arrows).









They put three experiments result on this paper. For all three data sets, the natural appearance of linear interpolations between distant points in the low dimensional coordinate space confirms that Isomap has captured the data’s perceptually relevant structure.




Latent Dirichlet Allocation

To contact us Click HERE
Latent Dirichlet Allocation (LDA) is an improved version of pLSI (if you want to check the detail, please click here). Its concept is very similar to pLSI. pLSI tried to learn all possible p(wi|zn), which are the probabilities that word i appears in latent topic n. That is also the same for LDA. However, pLSI has some critical problems. Because pLSI only learns topic mixtures p(z|d) for those documents on which it is trained, there is no natural way to use it to assign probability to a previously unseen document. The other problem is that the number of parameters pLSI needs to learn is kV+kM, which is linear to the number of training document. The linear growth in parameters suggests that the model is prone to overï¬�tting and, empirically, overï¬�tting is indeed a serious problem for pLSI.

LDA solves these problems by treating the topic mixture weights as a k-parameter hidden random variable rather than training a set of individual parameters for each training document.Thus, it only needs to learn k + kV parameters. k parameters for generating the topic mixture weights (α), and kV parameters for Î² . Furthermore,  the k + kV parameters in a k-topic LDA model do not grow with the size of the training corpus, and will not suffer from overfitting.

The following are the mathematical details.
Assume that given parameters Î± and β, the joint distribution of a topic mixture Î¸, a set of N topics z, and a set of N words w is


where p(θ|α) is a Dirichlet distribution.It's graphical representation isThen we can get p(w|α, β) by integrating over Î¸ and summing over z.After that, we can derive p(D|α, β) by multiplying the probabilities p(w|α, β) of all documents appear in the corpus D.

That is the mathematical equation of LDA.
The main difference of pLSI and LDA can be highlighted in the following figure. pLSI learns a latent topic combination for every document in the training set. Therefore, the learned combination of one document can be viewed as one point in the above figure. Instead of learning the combination of every document in the training set, LDA tries to learn a distribution to describe the probability that some combination of latent topic appears. That is, it tries to learn a distribution that documents appear in the latent topic space to describe the corpus. 
When new data comes, both pLSI and LDA fix the latent topics and tried to find one point in the latent topic space to minimize the reconstruction error. They need almost the same computational cost to locate the new document in the latent topic space. However, after the document is located, LDA can give one more information then pLSI. LDA can tell us the probability that the new document appears in the current corpus, while pLSI gives every document the same weight and thus cannot tell us that probability.

Support vector learning for ordinal regression

To contact us Click HERE
In this paper, they proposed a problem about ordinal regression, which is very different from the traditional learning problems, such as classification and metric regression.

In traditional classification problem, the loss function is defined as 0-1 loss.
In regression estimation, the loss function will take into account the full metric structure.
In ordinal regression problem, the problem this paper want to solve, they consider a problem which shares properties of both classification and metric regression. Like classification, the outcome space is a finite set, and like metric regression, there exists an ordering among the elements.

The detail of the problem is defined as follows.
An input space X belongs to R^n with objects being represented by feature vectors x = (x1,...,xn)^T.
Y denotes the rank order.
For every training point in the set S, we want to find a mapping function that map the points to the rankings.






The error cost function is defined as 




lpref  is 1 if and only if the ranking of h(x1) and h(x2) is different from y1 and y2.
The expected cost can be computed as






However, that is hard to minimize.
Thus, we first convert it into another set S', which is a set representing the pairwise ranking.










And in theorem 1, it said










Which means that for every projection function we get in the set S', the 0-1 loss for S' is proportional to the expected cost for S. Then, the minimum cost projection function found on the set S' will be the same as found on set S.  Thus, now the ordinal problem is converted into the classification problem and can be solved by SVM.

●Solving by using SVM
We want to find a projection vector W, and some biases that project original data into ranks.







Then, we can give SVM some degree of violation, and solve W.




After finding W, we can find different thresholds Î¸(ri) for different ranking now by minimizing the following equation under the constraint of above equation.




After doing this, we can get the final results we want as the following figure.















When new data comes, it simply project it by using W, and judge its rank by check the thresholds Î¸(ri)

========================================================Comment:
    The authors are very smart, they convert the ordinal problem into classification problem. Then, they can directly using the methods developed well on classification to solve the ordinal problem. However, I think the most important thing is to tell us why the projection derive on set S' will be the same as S. They only provide a theorem to say that they will be the same but haven't explain it. I think that is one weak point.
    Another weak point is the notation of this paper. This paper use too much equation to define a simple concept, such as lpref.  I think I will become much better to provide some illustration, or state their meaning directly by using words.

Learning to Rank: From Pairwise Approach to Listwise Approach

To contact us Click HERE
In the previous paper, support vector learning for ordinal regression, it use pairwise method to solve the problem of ordinal learning. However, using pairwise approach to solve this problem will induce some problem.
First, the objective of learning is formalized as minimizing errors in classiï¬�cation of document pairs, rather than minimizing errors in ranking of documents.Second, the training process is computationally costly, as the number of document pairs is very large. Third, the assumption of that the document pairs are generated i.i.d. is also too strong. Fourth, the number of generated document pairs varies largely from query to query, which will result in training a model biased toward queries with more document pairs.
Thus, in this paper, they propose the list-wise approach.
The followings are the problem setting.
















In training, a set of queryies Q=(q(1),q(2), ... ,q(m)) is given.
Each query q(m) is associated with a list of documents d(i) = (d(i)1, d(i)2, ...,d(i)n(i)).
Each list of documents  d(i) is associated with a list of judgments (scores) y(i)= (y(i)1, y(i)2, ...,y(i)n(i)). The judgment y(i)j represents the relevance degree of d(i)j to q(i).
A feature vector x(i)j=ψ(q(i),d(i)j) is created from each query-document pair.
Note that there is a q(i)  in Ïˆ function. If it doesn't have this item, then the ranking list will be always the same no matter what query is.

We want to find a ranking function f, which will output the score by taking one feature vector.
For the list of feature vectors x(i), we can obtain a list of scores z(i) = (f(x(i)1), ... ,f(x(i)n(i))).

The objective of learning is formalized as minimization of the total losses with respect to the training data.






Where L is the loss function.

In order to measure the loss function of two different ranking, they first convert the ranking into probability distribution, and measure the loss function by using Cross Entropy.
They use the following definition to convert the ranking into probability distribution.

















However, if we want to derive this probability distribution precisely, we need lots of computation overhead.
Thus, instead of calculating this probability, they calculate top one probability.
Then, for top one probability, the value can be derived more easily.








After using this representation, then we can use Cross Entropy as metric.






And phi-function is defined as exponential function.







Then, they use gradient descent method to solve this optimization problem.
The ranking function fw is based on the Neural Network model w.
Given a feature vector x(i)jfw(x(i)j) assigns a score to it.
Thus, for a query q(i), the score list z(i)(fw) = (fw(x(i)1), fw(x(i)2), ..., fw(x(i)n(i))).
For simplicity, they use linear Neural Network model in the experiment, fw(x(i)j) = w × x(i)j.
That is, they tried to learn a linear combination that minimize the loss function measure by Cross Entropy.
And they show their list-wise method really performs better than other methods