Transcribing Mandarin as a learning method

[Flickserve]

Split off from discussion here.

 

On 04/03/2017 at 3:34 PM, Publius said:

Yup, that's what I did with English and Cantonese, transcribing is my method.

 

If I do this for Mandarin, is writing pinyin acceptable?

[imron]

The only way to find out it to try it! (because what works or not for you may be different from others).

 

I imagine it probably would work as long as you are understanding what the pinyin is and what it means, and not just trying to match the sound as closely as possible to pinyin without any regard to the meaning.

[Publius]

1 hour ago, Flickserve said:

If I do this for Mandarin, is writing pinyin acceptable?

I think it is, as long as you are comfortable reading pinyin (I don't... hanzi is too ingrained in my brain) because you probably want to review the transcripts at a later time.

[Flickserve]

13 hours ago, imron said:

 

I imagine it probably would work as long as you are understanding what the pinyin is and what it means, and not just trying to match the sound as closely as possible to pinyin without any regard to the meaning.

 

What I sometimes do just for fun is listen to one of my sentences on my mobile, open up pleco, type in the pinyin of words I don't know/forgotten/guessed meaning and then look at the Chinese characters plus translation and see if it fits. This is because my transcript is held on my computer at home. Is this roughly equivalent to transcribing? There is no way I could write out the Hanzi for each word that I might recognise.

 

Imron, i don't know if you want to split off some of these posts into another thread on 'How to transcribe for Chinese ' with both yourself and Publius' comments.

[imron]

8 hours ago, Flickserve said:

Is this roughly equivalent to transcribing?

I don't think so, because it seems like you'd be doing a word or two at a time rather than whole sections of continuous text.

 

8 hours ago, Flickserve said:

There is no way I could write out the Hanzi for each word that I might recognise

 

Not if you don't practise it, no.

 

If you think you can, or you think you can't, you're right.

 

[laurenth]

9 hours ago, Flickserve said:

my transcript is held on my computer at home. Is this roughly equivalent to transcribing? There is no way I could write out the Hanzi for each word that I might recognise

 

I've been doing such transcription exercices for a few months, in an attempt to improve my listening comprehension. At first I was writing everything on paper, in pinyin. I was using a notebook with wide margins on each page. When I had to look up a word or (if the transcript is available) where I got the transcription wrong, I jotted down the hanzi, some explanation and a translation in the margin. However, the pinyin somehow looked - how shall I put that? - unreal, like a long string of hieroglyphs that were hard to differentiate, though I'm convinced that the cliché of "Chinese having so many homonyms" is an exaggeration if you use tones with your pinyin and if you take account of the context - by  the way, would it be possible to *not* take the context into account anyway? 

 

Then I thought that if I was making the effort of transcribing anyway, I might as well type in hanzi directly, if only because I would be able to recycle my transcripts for some more exercises: reading and reviewing, or feeding the transcripts into WorkAudioBook to create subtitle files and practice repeated listening of difficult parts. So for now, it's the other way around. I have a word processor file containing a 3 column table: column 1 is for what I think I hear; column 2 (if I have a transcript) is a corrected version with difficult words in red; column 3 is for notes, pinyin of new words, etc.

 

[Edit] I'm reading this post again and it suddenly occurs to me that, in fact, feeding a pinyin transcript into WorkAudioBook to create pinyin-only subtitles could be useful as well.

[Publius]

What you do for fun is only part of the transcribing-as-a-learning-method process, namely, looking up unknown/unfamiliar words.

You may also need to use search engines for proper nouns -- names of individual persons, places, organizations.

Transcribing allows you to attack materials way above your current level. As a result, new words are inevitable. And believe me, words learned this way have a better retention rate than usual.

 

The cause of not understanding natural speech I think boils down to A) speed -- the input rate exceeds your brain's processing rate; B) content -- you wouldn't understand it even if it were written down; C) accent -- and other deviations from the "standard" such as reduced pronunciation, omission of certain words, etc. These all can be learned or trained for. And transcribing is a good way to go about doing it.

 

The immediate goal is to produce a transcript as accurately as possible -- to break down the flow of audio signals into words and phrases and reassemble them into grammatically correct and logically coherent sentences. So hanzi or pinyin doesn't really matter as long as they are meaningful to you as a text that carries meaning rather than a representation of sound syllables. (Speaking from my experience with Cantonese and Japanese, the relatively simple syllable structure and the extra layer of a phonetic IME make it easier to type characters/words you don't even know how to read -- I'm not sure whether it's a good thing or a bad thing though.)

To achieve the immediate goal you can use any tools available, loop as many times as you think is worthwhile, sentence by sentence, word by word, slow down if necessary, use dictionaries/thesauri, the internet, anything -- as long as it's not a speech recognition software -- yup, that would be cheating.

But sometimes perhaps you should limit the tools and resist the temptation of being able to understand a very challenging piece through transcribing. The best approach I think is to find something at or even below your level (try speeding up instead of slowing down hehe) and work up the ladder gradually.

 

I always say the most important thing in listening is not the ear. It's the brain. A language user at a higher level can spot a simple mistake and correct it by reading the transcript without even listening to the audio. That's the ultimate listener. If you don't believe me, consider this: If I write "a US piece core member" you will know immediately it's a mishearing of "Peace Corps" -- the listener either does not know the word, or has only seen it on paper without realizing that "ps" is silent. Similarly If you write "攔驢老少" I will know too that if it's not a word play then it must be a mishearing of "男女老少" -- the speaker has an accent that does not differentiate l and n, probably from Hubei, Hunan, Jiangxi.

There's a joke in the AI circle, "How to wreck a nice beach? You sing calm incense!" (How to recognize speech? Using common sense!) In successfully transcribing a passage you need to actively apply all the knowledge you possess about the language, the culture, the subject matter. So in a sense it's not merely a listening exercise and also why I consider it a very powerful technique.

 

And intensive studying should always be supplemented by extensive studying. Can't understand news broadcast because of its tortuous sentence structure? Then read more news articles. Find some news radio streams and put them on as background music.

 

The guy at HackingChinese also mentioned transcribing Chinese audio as an active form of listening practice. But contrary to his advice, my approach emphasizes *daily* drilling over a *sustained* period of time (in imron's words). It's a very demanding task, so don't bite off more than you can chew. Start from 5-minute audio clips. If it's still too much, try 2- or 3-minute ones. But you should do at least one hour of transcription exercise everyday for let's say three months. It worked for me, but it's not guaranteed to work for everybody. If it doesn't work for you then I presume you know better than to 在一棵樹上吊死.

 

[mlescano]

This is an inspiring thread. Doing daily transcription exercises over several months sounds like a good daily workout. Certainly a lot of words, structures and collocations will eventually stick this way. If you're already able to read and write a lot of characters, this gives you two options:

 

-Type

-Write by hand

 

If you type, you might be "cheating" in that maybe you don't really remember the correct character for some contexts (such as 做 instead of 作, 年纪 instead of 年级), and you're instead relying on the IME guessing it for you, thus depriving yourself of the chance of reinforcing which is the right character for the right context.

 

If you write by hand, you might be "cheating" in that maybe you don't quite remember the pinyin for some characters (was it cheng or chang? Did it finish with a NG or a N?), but you can still write it.

 

So... Maybe I'll try alternating "typing" days with "handwriting" days. What do you think?

[imron]

Prefer handwriting to typing, switching to writing pinyin if you don't know the character.

 

The automatic sentence correction with modern IMEs gives too much of a hint.  The whole point of this is to exercise active recall from passive input.  You remove a large amount of the benefit if you are using passive recall from passive input.

 

6 minutes ago, mlescano said:

maybe you don't quite remember the pinyin for some characters (was it cheng or chang? Did it finish with a NG or a N?), but you can still write it.

 

I expect these cases to be far less common than the alternative.

[mlescano]

Ok, so I started today, using video content with exactly timed SRT subtitles and LAMP player for automatic pause after each subtitle line. The video is of interest to me, but most people will find it boring. I first watched it in Spanish some days ago, so it's not completely new material.

 

For the first pomodoro (25min. study block) I used a Hanzi grids sheet, but then I decided to design my own. I wanted 25 squares per row, which was beyond Hanzigrids' maximum. I also left a space between rows for writing corrections, and some space on the right for counting corrections per row. At the end, I calculated my transcription rate (6.6 characters per minute for both exercises) and my accuracy rate (didn't record it for the first pomodoro, 92% for the second one.). For easier calculations, I'm not writing down spaces or punctuation. Only hanzi.

 

I'm sharing the blank A4 PDF in case anyone finds it useful. Having 25 characters per row might not be as comfortable, but makes for easier calculations once you finish your listening session. BTW, my handwriting is equally terrible in Spanish. 

 

Tomorrow I might start keeping track of my stats on Excel so I see how my rate and accuracy evolve with time. This section was easy for me, but others will not be.

 

transcriptionpaper.pdf

[Shelley]

If you put Hanzi grid into landscape mode you can get more per line. I do this to make me do more writing of characters I have problems with.

 

I suppose the whole point of this exercise is to write what you hear. I find I understand less as I struggle to keep up with writing as I listen. I prefer to listen till I understand completely then write it out.

 

I prefer to use characters but will resort to Pinyin if I can't recall the character, then fill it in later.

Is the goal perfect/or at least good, writing of characters or getting the meaning correct and just adequately writing the character?

 

[mlescano]

Well, today I found this article: http://www.languagesurfer.com/2012/08/25/how-to-transcribe-an-essential-skill-in-foreign-language-learning/

 

Step 2 recommends listening to the whole part you will be transcribing once before starting to transcribe. This way you get the proper context.

 

I'm not aiming for perfect handwriting in Chinese: I was almost refused entry to elementary school because of my poor handwriting in Spanish! 

 

And I'm advancing very slowly: Today in my two sessions of 25 min each, I only managed to transcribe a minute and a half of video! (165 characters in each session).

 

I think I'll start a blog to keep you guys posted about how I'm doing with Publius' "one hour for 3 months transcribing challenge"

[imron]

5 minutes ago, mlescano said:

And I'm advancing very slowly: Today in my two sessions of 25 min each, I only managed to transcribe a minute and a half of video!

Keep track of this value every day.

 

You won't notice your improvement on a day-to-day basis, but when you look back after a month, or after 3 months, you'll have hard data that shows you have improved.

[imron]

On 3/23/2017 at 5:30 AM, mlescano said:

I wanted 25 squares per row, which was beyond Hanzigrids' maximum

Ask and you shall receive.

 

You should now be able to have up to 27 squares per row (you might need to reload Hanzi Grids if you currently have it open)

[Publius]

Yeah, I failed to mention Step 2. :/

It's always best to listen through the material once and get a general idea of what's going on before you start writing.

 

I also want to add: Use dictionaries wisely.

Sometimes you can deduce the meaning from the context, you just don't know what exactly is the word. A thesaurus can be very useful. Or try some words in the language you know with bilingual dictionaries and see what they give you in the target language.

 

And I can't emphasize this too much: Don't forget how to wreck a nice beach! Cheers.

 

[imron]

1 hour ago, Publius said:

Don't forget how to wreck a nice beach!

Even modern voice recognition systems still fail at this.

[Flickserve]

On 05/03/2017 at 4:05 PM, imron said:

you think you can't, you're right.

 

You are right. I can't. 

I can't write Chinese. 

[imron]

...yet

[mlescano]

@Publius, you mentioned transcribing being "your method". I'm most interested in how you learned English. I'm about to tackle an European language along with Mandarin.

[Publius]

There really isn't much to talk about. My English was rubbish as you would expect from an average college graduate in China who isn't an English major. I started to transcribe newscasts (VOA), documentaries (NatGeo/Discovery), movies (The Matrix), hoping to improve my listening skills around 2000. That was the dial-up, pre-Google era and before transcribing became known as a "method" among English-learning Chinese students (鐘道隆逆向英語學習法). I did it again in 2007-08, this time on a language-learning website built around the idea of "crowdsourced transcription." I stopped when I felt I had reached a level where transcribing became too time-consuming and less efficient. Overall in my experience, materials for transcription can be arranged in order of difficulty like this: educational stuff (which I skipped) < Special English < standard news report/public speech < documentary/cartoon < interview/talk show < TV/movie.

During the same period I read a lot of books, starting from 30 or so of Agatha Christie's detective novels. I also did some intensive reading: Lolita by Vladimir Nabokov and The Thorn Birds by Colleen McCullough (which turned out to be too long I couldn't finish it). It was before SRS went mainstream so I just kept my vocab list on paper flashcards. Reading enlarged my vocabulary like no way before. And a large vocabulary makes everything easier. Bam! virtuous circle. And that's about it.

If I'm to learn a new language today, this will be how I plan to do it: Get the script and pronunciation down quickly. Grab a grammar book, read a couple of times, not to remember everything, but to know where to find the answers. Find a 5000-ish vocab list with sample sentence audios and cram it like mad using SRS. Then just read, read, and read some more. Do the transcribing at the same time until it feels like a waste of time. As for the production side? Well, output will come naturally when you have enough input -- hopefully.