How I've been using Bunpro to study for the JLPT N2

Bunpro
#1

Over the summer, I have been playing catch up from a year where I didn’t do very much immersion or study of Japanese. For the last month, I decided to finally try out Bunpro’s JLPT vocab decks in addition to revisiting the grammar reviews that I had left dormant since getting out of practice. I signed up for the JLPT N2 this year after underestimating the N1 exam the previous year, and definitely didn’t want to have a repeat of my initial testing experience, so I made sure to get a proper routine set up this go around.

After a little while, I realized how frustrated I was when I would try to get back into reading stuff for immersion because I kept getting tripped up by certain vocabulary and being unable to get into a good reading flow, as well as my abysmal reading speed despite how long I have been studying Japanese overall, so I decided to do an entire review of lower level JLPT vocabulary.

In a very extreme move, I did a very intensive SRS load on Bunpro by adding anything between 250 to 500 cards a day from the JLPT N5 and N4 decks, using the read and grade review modes (as opposed to typing in my answers, as that would have slowed me down significantly). Yes, I know what you are thinking. It was indeed quite the grind, and my reviews were as high as 1500 to 2000 on some days. But I made sure to be smart about what cards I was brute forcing. These were all words I either knew very well or encountered several times before. And for other cards, I would go ahead and mark as Mastered if I felt they were too obscure, or if I found out they were actually just a duplicate of a grammar card or of a vocab card in other level decks (of which where there were actually quite a few).

My overall goal was speeding up my reaction time: I wanted to have my brain pull up the meaning and reading of a card in a heartbeat, not 15 to 30 seconds later. For an exam like the JLPT, every second counts, so I knew that even just speeding up my reading ability would be a huge advantage. And actually, there were a few words in those N5 and N4 decks that were surprisingly new to me, so over time I gradually added those in too, and they have been very handy now as I have gotten back into immersion.

When the AWS outage happened the other day, like a lot of other people, I ended up using my time immersing when I couldn’t access my Bunpro reviews. Using vacation mode, I froze all my reviews at that point, which I had by then completed all of the N5 and N4 vocab decks, and had 1400 review items under N3 and N2 vocab decks. I then decided to watch the Chainsaw Man Compilation movie on Crunchyroll so that I had a nice refresher before watching the Reze Arc in theaters, watching with Japanese subs by way of the 字幕プレーヤー userscript to overlay them on Crunchyroll (which can be found here: GitHub - sheodox/jimaku-player: Use your own subtitles on VRV or Crunchyroll to learn Japanese!)

I can’t tell you how nice it has been, it feels like all of the friction is gone. I’m now at the point where I’m going back to using actual immersion as my “SRS”, and will probably keep all of my Bunpro reviews on vacation until I decide to unfreeze everything for when I take start studying for the N1 JLPT exam (since my focus this year is just on everything N2 and below, and haven’t finish the N1 grammar items in Bunpro).

Let me be clear that there is no way that I would do a study strategy like this for the long-term. In a sense, the AWS outage was the perfect timing for me to realize that, after a month of doing this, I was already in great shape to get back to immersing and was indeed getting exhausted by my self-imposed daily review load. This was just a short-term intensive study regimen to jumpstart my brain again when I felt like I kept hiccuping on stuff that I felt should have already mastered.

Thanks for adding vocab decks, Bunpro team! Keeping everything centralized to one website was a no brainer for me, and as things get closer to the exam, I will definitely be using the cram feature in addition to the new JLPT practice tests. For me, I’ll be spending the rest of my time from now until December just immersing and finishing up the Bunpou and Dokkai JLPT N2 books from Shin Kanzen Master, because I just need more practice in dealing with how they set up those particular exam questions.

12 Likes
#2

I’ve really been happy with the practice for grammar and vocab here and the way it’s set up. I’ve had more luck learning new vocab here than I have with any other method (including self-made flashcard decks) in the past.

I also commiserate as a slow reader. I’ve taken two JLPT. I didn’t finish either, though I did them backwards and so I only missed a couple of questions when I did N2 and managed to pass. When I took N1 I was well behind and had quite a few I didn’t even get to answer because I ran out of time. I’m slow in English, too, but that doesn’t help me when it comes to taking a timed test. I’m trying to improve my reading speed now so when I take N1 again I can at least get through the whole thing in the time limit.

1 Like
#3

When you say do the JLPT backwards, I’m wondering exactly what order you did the questions in.
I’ve heard it’s a good strategy to start with some of the more difficult reading questions, but I’m guessing you still read the texts in order (from top to bottom) and did the questions for each text in the original order? :thinking:

1 Like
#4

Just to add to this, I think one of the reasons it’s such a good idea to start from the back for the reading questions is because they’re likely weighted more. I’m not sure if there’s “proof” of this, but every practice test book I used had those questions worth more points when it came to figuring out if I would pass on a real test or not. Personally it also felt like the “back” reading questions required just as much reading as the “front” reading questions as well, so it seemed like the smartest move time-wise.

#5

There’s no evidence to suggest difficult questions are weighted more. In fact, the weighting is different from person to person because they use a programme that takes your raw score and adjusts it based on your “answering patterns.” Which, from my understanding, means you could get adjusted to a low score even if you score high on a hard part if you otherwise do not show proficiency in the exam. However, I do agree it’s probably good to do more tiring reading questions when your brain is fresh.

"Calculation of test scores not affected by the difficulty of exams (scaled scores) is based on a statistical test theory called Item Response Theory (IRT). This is completely different from calculation of raw test scores based on the number of correct answers.
Scaled scores are determined mathematically based on “answering patterns” of how examinees answer particular questions (correctly or incorrectly). "

Interestingly some people seem to interpret this as “difficult questions are weighted more” but that’s not what it seems to imply at all. The purpose of using IRT is to equalise grading over changes in difficulty among exams. One of IRT’s purposes also seems to be to account for guessing.

Anyway, it takes the raw scores and adjusts the final scores based on a statistical model, not on the difficulty of the questions themselves.

from https://www.jlpt.jp/e/about/pdf/scaledscore_e.pdf
more info about IRT Item Response Theory (IRT): Better assessment with ML | Assessment Systems (ASC)

2 Likes
#6

To give an example of how IRT could work:

Say person A gets 3 easy questions right and 2 hard questions wrong. Person B gets 3 easy questions wrong but the 2 hard ones right. Person B shows an inconsistent item response, which indicates guessing. Thus, person B will likely receive a lower adjusted score than person A.

#7

I can’t speak to weighting of the questions (I think no one really knows how it’s scored lol), but usually the last readings have more questions associated with them and they’re longer, so you might have four or five questions with one page-long reading, while the first readings might be a single question with a paragraph.

Advice back when I took it was that you should do the readings backwards because if you don’t finish, it’s better to not finish the ones that are one question per reading than the ones where you have four or five per reading. It also means if you’re crunched for time, you’re trying to finish with shorter readings with just one question, so it’s easier to skim it or try to find the answer in a rush, compared to trying to search a whole page for multiple answers when you’re just trying to get the last ones in.

I’m not sure I think they’d be adjusting scores based on thinking people are guessing. Generally speaking, if you’re guessing, you’re statistically not going to pass. If you have four question choices, you’re only 25% likely to get it. Maybe someone right on the edge of passing might be able to guess on a couple and get lucky and bump themselves up to a pass, but how do you determine if it’s really a guess, or someone subconsciously remembering or going with their gut and their gut is right?

I also think it’s not really fair to try to judge a “hard” question versus an easy one. Sure, some grammar points might be more difficult, but I might do a reading on a scientific topic that I don’t know much vocab for, so that reading for me might be really hard, but maybe someone else is good at science vocab, but not art, so they do great on a scientific reading, but not one about art.

Also, “hard” and “easy” is going to be highly dependent on what language base you’re coming from, and the JLPT is a national test. What’s hard for someone in America might be easy for someone in China, or vice versa. I’ve heard speculation that it’s weighted, and I could see there being weight to certain questions, but I am not sure I’d believe that they’re weighting it based on perceived inconsistencies. To me, that’s too subjective.

#8

Part of IRT (which is used to calculate your final score) is definitely about accounting for guessing on multiple choice exams. This isn’t quite the same as adjusting your score based on guessing, but rather accounting for it in their calculations of your ability.

These are the three parameters of IRT:

The foundation of IRT is a mathematical model defined by item parameters . A parameter is an aspect of a mathematical model that can change its shape or other aspects. For dichotomous items (those scored correct/incorrect), each item has three parameters:

a: the discrimination parameter, an index of how well the item differentiates low from top examinees; typically ranges from 0 to 2, where higher is better, though not many items are above 1.0.

b: the difficulty parameter, an index of what level of examinees for which the item is appropriate; typically ranges from -3 to +3, with 0 being an average examinee level.

c: the pseudo-guessing parameter, which is a lower asymptote; typically is focused on 1/k where k is the number of options.

Examinees with higher ability are much more likely to respond correctly. Look at the graph above. Someone at +2.0 (97th percentile) has about a 94% chance of getting the item correct. Meanwhile, someone at -2.0 has only a 25% chance – barely above the 1 in 5 guessing rate of 20%. An average person (0.0) has a 60% chance. Why 60? Because we are accounting for guessing. If the curve went from 0% to 100% probability, then yes, the middle would be 50% change. But here, we assume 20% as a baseline due to guessing, so halfway up is 60%

This is from the earlier link about IRT.

2 Likes
#9

This is all very confusing. :sweat_smile:

2 Likes