ryanheise's experiment log - Experiment 5 (Japanese)

Continue or start your personal language log here, including logs for challenge participants
User avatar
MorkTheFiddle
Black Belt - 2nd Dan
Posts: 2114
Joined: Sat Jul 18, 2015 8:59 pm
Location: North Texas USA
Languages: English (N). Read (only) French and Spanish. Studying Ancient Greek. Studying a bit of Latin. Once studied Old Norse. Dabbled in Catalan, Provençal and Italian.
Language Log: https://forum.language-learners.org/vie ... 11#p133911
x 4824

Re: ryanheise's "experiment" log - Experiment 2 (Japanese)

Postby MorkTheFiddle » Wed Sep 11, 2019 10:39 pm

ryanheise wrote:Experiment 2

  • I have so far had a lot of practice listening to short phrases or sentences and trying to imitate what I hear. But my brain is so used to stopping at the end of the short segment that when a real person is speaking to me in longer strings of sentences, my brain just stops at about the the length I've practiced.
This is a very perceptive observation. It applies, I think, to both listening and reading. Keeping everything in your head from a long passage of Proust or Plato in the original can be challenging. Even less lengthy passages take some working up to.
2 x
Many things which are false are transmitted from book to book, and gain credit in the world. -- attributed to Samuel Johnson

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's "experiment" log - Experiment 2 (Japanese)

Postby ryanheise » Wed Sep 18, 2019 4:08 am

MorkTheFiddle wrote:This is a very perceptive observation. It applies, I think, to both listening and reading. Keeping everything in your head from a long passage of Proust or Plato in the original can be challenging. Even less lengthy passages take some working up to.


I think what this exercise has taught me so far is that to progress beyond short segments, you need to develop the skill of chunking. When short segments is all I had to deal with, I found that my brain was "getting by" with a more limited set of processing techniques that don't scale well. But when extending that to longer segments, my brain has no choice but to learn how to take in meaning in bigger chunks at a time.

I should post an update on this experiment soon. I've just been trying to fiddle with the settings in WorkAudioBook to get the optimal segment lengths which can be a bit tricky, and also figure out a base line. I'll update soon.
2 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's "experiment" log - Experiment 2 (Japanese)

Postby ryanheise » Wed Sep 18, 2019 5:07 am

Just a really quick update. The settings I'm using for WorkAudioBook are:

  1. Short segments. Preferred phrase length: "I don't care", Max phrase length: 3 seconds, Min phrase length: 0.5 seconds.
  2. Extended segments. Preferred phrase length: "I don't care", Max phrase length: 8 seconds, Min phrase length: 0.5 seconds.

When I choose 8 seconds for extended segments, that's just a maximum and the average is going to be a bit lower. For the shorter segment settings, it has difficulty finding the best place to cut the audio mid-sentence, as you would expect, so I'm still trying to figure out the best settings there. Best results usually come from choosing "I don't care" for the preferred phrase length which I think allows it to choose what it thinks is the best cut point (although it still does get challenged when you try to make the segments small).
3 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's "experiment" log - Experiment 2 (Japanese)

Postby ryanheise » Sun Sep 22, 2019 4:14 pm

Some news! Today something clicked and I was able to comfortably extend the segment length up to a maximum length of 25 seconds.

  • First, I listened to segments in the range of 0.5 seconds to 3 seconds, speaking it after hearing, and repeating several times until comfortable before moving onto the next segment. I did this for about 20 seconds of audio. I repeated some harder segments about 7-8 times to let them sink in, while others I only needed a couple of repetitions.
  • Second, I went back to the beginning to cover the same ground again, but increased the max segment length to 6 seconds, again speaking it after hearing and repeating until comfortable before moving on. The number of repetitions per segment decreased over the previous round.
  • Third, I went back to the beginning but this time I increased the minimum from 0.5 to 6, and increased the max to 12 (basically doubling the max each time). This time I focused on listening/comprehension and did not speak after hearing. This is the level I was really trying to build to in this experiment, where I've got the output practice out of the way with the smaller more manageable segment sizes, and now I can just relax and focus on comprehensions without the stress of trying to repeat the whole thing. I had never really done targeted practice on this before, but I could immediately feel the payoff as it didn't take many re-listens to keep my listening attention span alive for the whole duration.
  • Fourth, I increased the min to 12, and the max to 25 (there's no setting for 24), and again from the beginning just focused on listening/comprehension. By this stage I had already become comfortable with the previous iteration of segments which were already substantial in length so joining them together (with fewer joins this time) was no problem.

After that, I repeated the same exercise on the next (roughly) 20 seconds of audio. And then finally I went back to the very beginning, and listened to the entire 40 seconds.

So far, I have to say I'm REALLY liking this approach. I can see improvement in the number of repetitions I required for the second 20 seconds compared to the first. I plan to keep on doing this with new and fresh material to monitor it a bit longer term.

It is a bit fidgety sometimes getting the segment lengths right, but when it works, it works really well.

Next, the main thing I need to do is just devote more time to my language learning. Last week I took a lot of time off my language learning schedule to work on a new project to build a sort of a "search engine" for i+1 foreign language audio content. It was fun to build, but I did neglect my language learning so from here on it's time for me to catch up on my language learning. The search engine project now at least works, so it should be fine without my full attention for the time being!
6 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's "experiment" log - Experiment 2 SCORECARD (Japanese)

Postby ryanheise » Fri Sep 27, 2019 11:23 am

I've been working on a way to track my listening progress. I came up with this scorecard which I fill in while doing a listening session to record how many lapses of concentration I had, or in other words, how many times I had to listen to a segment again:

Image

This is for a session dated 27/09/2019 in which I listened to a portion of 消滅 from 02:40 to 03:37 (57 seconds). The first row of numbers break down how many times I had to repeat any segment of audio, where I used AudioWorkBook to cut the audio up into segments of between 0.5-3 seconds in length. There were 7 segments that required a repeat hearing. Of those 7, there were 2 that required a second hearing. Of those, there was 1 that required a 3rd hearing, and that one also required a 4th hearing. Breaking it down this way strikes a balance between the amount of useful data I can extract out of a listening session, and how much data is actually practical to write down without being too disruptive. (Although I'd welcome any other ideas on what I should be tracking!)

Next, since I like to practice listening outdoors and not in front of my computer, I also needed a way to easily record this information. After searching on the Google Play Store, I found an app called Thing Counter which allowed me to create 7 separate counter buttons, one for each of these 7 columns in my table above. Using the "multitasking" feature of Android, I am able to create a split screen where I can run AudioWorkBook in one half of the screen, and Thing Counter in the other half:

Image

Whenever I repeat a segment, I can easily press the appropriate counter button, and take screenshots after completing the run for a particular row which I can copy into my spreadsheet when I get home.

With this experiment, I should also mention that I am listening to material where I already understand all of the words, and where therefore my need to stop and repeat a segment is purely due to a lack of concentration or a limited attention span.
4 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's experiment log - Experiment 2 RESULTS (Japanese)

Postby ryanheise » Mon Oct 14, 2019 4:51 am

Preliminary results

Each chart shows the progress in my listening attention span over time for a different story. The X-axis shows each day. Each day I listened to roughly the next minute of the story segmented into 4 different sizes tracked by the 4 lines. The Y-axis shows how many reps per minute, where I repeated only when I didn't "get it" and needed to hear the segment again.

Image

One of the difficulties of measuring progress is that the different stories have different difficulties, so comparing reps from one story to another is not going to mean much. However, assuming that the difficulty within a story is roughly the same, these charts do show an improvement in the number of needed reps over time within the same story, particularly for the shortest segment sizes (0.5-3sec in the first story and 0.5-4sec in the second).

In the second story, the 0.5-6sec progress shows me actually getting worse in the middle of the story and then getting better later on. This could be either because the language was more difficult in the middle or because around that time I missed my usual opportunity to do my listening at daytime, and instead did it shortly after midnight. I did feel at the time that I was losing concentration, and at other times I felt like I had moments of perfect focus, so there will be some variability here. In any case, the overall trend is downwards, so I'm happy with that. And most of all, this sort of practice is quite enjoyable.

The data recording process itself is not enjoyable, so I stopped recording data after this, but I might revisit this periodically to see if I've improved.

Next up, I would like to do some more work with language exchange and would like to try poly-glot-a-lot's language exchange technique. I'll write up a new post soon.
3 x

User avatar
MorkTheFiddle
Black Belt - 2nd Dan
Posts: 2114
Joined: Sat Jul 18, 2015 8:59 pm
Location: North Texas USA
Languages: English (N). Read (only) French and Spanish. Studying Ancient Greek. Studying a bit of Latin. Once studied Old Norse. Dabbled in Catalan, Provençal and Italian.
Language Log: https://forum.language-learners.org/vie ... 11#p133911
x 4824

Re: ryanheise's experiment log - Experiment 2 RESULTS (Japanese)

Postby MorkTheFiddle » Mon Oct 14, 2019 10:26 pm

ryanheise wrote:One of the difficulties of measuring progress is that the different stories have different difficulties, so comparing reps from one story to another is not going to mean much .... In the second story, the 0.5-6sec progress shows me actually getting worse in the middle of the story and then getting better later on.

"Different stories have different difficulties" is a very good point. Even with the same author, I have found.
As far as flagging in the middle, that happens to me a lot. In the beginning, the freshness keeps me alert, and towards the end the prospect of ending keeps me alert. But in the middle, there is less impetus.
An interesting experiment you have conducted, getting some honest data instead of using mere intuition.
2 x
Many things which are false are transmitted from book to book, and gain credit in the world. -- attributed to Samuel Johnson

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's experiment log - Experiment 3 (Japanese)

Postby ryanheise » Fri Nov 08, 2019 2:05 pm

Experiment 3

A few weeks ago I wrote some notes on Poly-glot-a-lot's interesting language exchange method, and I thought it would be fun to experiment with. The key idea is to use children's picture books as subject matter for language practice and to converse 100% in the target language without any grammar explanations and without any language corrections. Your partner describes what's happening in the pictures in his/her own words and may ask you yes/no questions to confirm your understanding of language acquired so far. You will learn the yes/no responses up front, in addition to learning how to ask your own set of simple questions: what's that? what's he/she doing? why? You will also learn up front a set of expressions to use in situations where you're stuck. Another aspect of the language exchange session is TPR practice where your partner asks you to do something that requires a physical response rather than a verbal response. As your skill progresses, you ask your partner to give i+1 language.

My goal is to simply experiment with the elements of his method that I haven't already adopted. In particular, the use of children's stories as subject matter, and the prescription of a specific style of question, answer, description and command to practice in language exchange sessions.

What are the problems?

  • My current diet of intermediate input alone is not transferring into intermediate speaking ability.
  • My ability to understand intermediate spoken language leads to my language exchange partner to ask questions that would require an intermediate speaking ability to answer.

What will I try?

I will try Poly-glot-a-lot's method in full. My theory is that the reason my intermediate input diet is not being effectively converted into active speaking ability is that I simply need to go back to more "basic" input where I have a better chance to listen to basic and useful sentence patterns and then apply them when speaking, with a fast turn around time that will help to cement those patterns in my active vocabulary and grammar. i+1 input may have helped me to get from zero to the intermediate level for listening, but to get to that same level for speaking, I need to rewind 'i' back to zero and gradually build up my speaking competency from the start. Poly-glot-a-lot's method provides a nice framework in which I think I can do that.

This is probably going to a long-term experiment, but I have my first language exchange session scheduled for tomorrow, so we'll see how it goes!
6 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's experiment log - Experiment 3 (Japanese)

Postby ryanheise » Mon Nov 11, 2019 1:53 pm

So how did it go?

First, I just want to recognise how fortunate I was to have an excellent language exchange partner who very quickly grokked the rules of the game and skillfully stayed within the framework to help me acquire new expressions and vocabulary while keeping me engaged. I could imagine it being less effective otherwise, so YMMV, and perhaps you may wish to try out several partners. Second, at the moment, I only have one session scheduled per week, and I'll need to realistically find more exchange partners to fill in the other days if I want to fully benefit from this immersive method.

Here are my observations and reflections after the first session:

  • Going into this, I was a bit skeptical that I could be as creative as Jeff Brown clearly is in his demo video, but now that I've had my first session, with a good partner, I can say that the framework Jeff has provided makes it easier for creativity to thrive. If there's word or expression you don't understand, there are many creative ways for your partner to compare and contrast what's going on in this picture to what went on in other pictures where you already comprehended his description.
  • The fact that there are lots of pictures means that pointing becomes a very efficient way to link words to comprehension.
  • The format of descriptions followed by engaging questions and answers is brilliant. The fact that you have these rules set in place where the questions you ask are short but may have long answers, whereas the questions he asks may be long but have short answers, essentially means that the conversation flows, covers more ground, allowing you to ultimately hear and acquire more words and expressions.
  • The fact that you're covering simple material means that you are much more able to notice sentence patterns when they keep getting used over and over again, and because of this repetition effect, before long, you feel like you know the pattern so well that it's natural for you to output a sentence with that pattern. When I did output sentences based on a new pattern I had heard repeatedly, I felt I could do it because of that repetition and the fast turnaround time between hearing it and being able to apply it. This framework makes it really easy to do that.

One thing I did not do in the first session was TPR so I'll have to remember to do that next time.
6 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: ryanheise's experiment log - Experiment 3 (Japanese)

Postby ryanheise » Wed Nov 20, 2019 4:59 pm

I had my second session last Friday (yep, a bit late to add this to my log!), here is my summary.

This time I made a point to try TPR, and we spent the first 10 minutes doing so. We had to come up with a list of commands and write them down, and the plan is to keep adding to this list until we build up a few hundred over the coming months. It would be helpful to have Jeff's staged lists but we'll try to make do building our own. During the activity, he would command me to do something and I would respond physically. If I was stumped, he would demonstrate and then jot a mark on paper next to that action so that he could try it again later to help reinforce the memory. I think we'll need a system or an app to help keep track of which ones I was stumped on to make it easier for him to know which command to use next.

Was TPR helpful? My initial feeling was that it might have potential, but was not as helpful as the main event. That could be because the list so far is too short, so I will see how things evolve week to week. However, once moving onto the main event, it turned out to be beneficial in the sense that he used some of the commands I struggled with in the context of the stories. I do appreciate that responding to language in a physical way brings another dimension to the connections I can make with the words to help reinforce memories.

I wouldn't say there is anything new to say about the main event, i.e. using the picture books. It continues to be fun.

One interesting thought I had came during his telling of the Three Little Pigs story, and that is, what an excellent story this is to help young children pick up a language, due to the repetition built into the story (i.e. the first house, the second house, the third house).
3 x


Return to “Language logs”

Who is online

Users browsing this forum: emk, Sizen and 2 guests