掲示板 Forums - Grammar parser - please read before using!
Top > renshuu.org > Feature Requests/Improvements > Finished/Rejected Requests Getting the posts
Top > renshuu.org > Feature Requests/Improvements > Finished/Rejected Requests
(This is being written in advance of the official experimental release)
What is it: a matching system so single/multi-word grammar expressions can be referenced and examined within sentences.
Example: この 薬 を 飲んだ方がいい. (You should take this medicine / It'd be good to take this medicine).
Currently, you can mouse over words and get info on word definitions and verb conjugations. With this system, the words 飲んだ方がいい will have additional lookup info, binding them to the 方がいい grammar entry in renshuu's grammar library.
Things to note:
1. This is an automated system based on pattern matching. It's pretty good, and will do its best to suggest the most likely expression should several expressions overlap the same set of words. It is also flexible towards word order (when applicable) and conjugation of related elements.
2. Currently, it does not show basic particles, since there is not yet a good way to prioritize meanings per sentence (for example, で has several entries in renshuu's library). Some verb conjugations are not yet shown as well, and use the old system.
3. Grammar expressions that are embedded inside of words are not matched.
4. There is a temporary "Toggle Grammar" link that appears under sentences - this is used to see all matched grammar for a sentence at a glance.
If you see a grammar expression that you are pretty sure is in renshuu's grammar library, but isn't being in marked in a sentence, do the following:
A. Tap "Toggle Grammar"
B. Use the form contained to send a missing grammar report.
It is most likely not necessary to send multiple reports for the same grammar expression.
Aside from the amazing lookup feature, this will help us do amazing work with our 170,000+ sentence library. The following will be possible:
1. Easier mining of example sentences for grammar expressions
2. Better ordering of sentences based on user grammar knowledge
3. Expansion of grammar quizzing materials
4. Text Analyzer expansion - get grammar grading levels as well for bodies of text.
This is not yet available for any newly made user sentences. It will eventually be automatically applied, but this will only happen once the system is formally released.
Note: Typically, when a bug is reported, it can take several hours / a day for the sentences to get the new data. I am working on speeding up the system, but because it is so flexible, there is a *huge* number of word/conjugation/pattern combinations that need to be checked for each sentence.
For the identified grammar would it be possible to show the Japanese grammar page title in addition to the English description ?
For example:
The grammar comes from the "について" page, so I would prefer a display something like:
活動について
について
1. Concerning A; About A; Relating to A (details)
I find it easier/faster to understand what grammar it is talking about when that title is included.
Thanks to those of you testing it, the accuracy of the tool has gone up quite a bit. As I continue to work on that, I am implementing a new hinting system.
There are many expressions that are the same in terms of syntax, but have different meanings.
Example: まで
One meaning is "until, as far as, etc.", while another is "even". Both of them function with a noun in front of まで
renshuu will, in this case, list both of them, but only show one as the "suggested" meaning, with the second one listed as an alternate.
In order to better choose which one shows up, each expression can take one or more hints (these are being added slowly) to weight one expression over another.
As an example
学校まで走ってください。(Please run to / as far as the school)
In this case, the expression has "until" and "as far as" as English hints. If they are found in the English, it'll weight that expression over the "even" version of it.
It works with Japanese as well - some expressions are more likely to contain other words in a sentence. Later on, I'll be able to start classifying words, so we'll get even more fine-grained marking. (Ex: This expression has a noun that is usually a person)
A small change should make the grammar report tool somewhat easier to use.
The markup reporting tool is going to be moved to a more discrete location in the near future.
The markup system is fairly mature, and I doubt there will be many more structural improvements. I am currently integrating it into the text analyzer.
I expect a public release in December. That does not mean it'll pick up all copies of every grammar expression, though - this is an ongoing process as I continue to add new variations and patterns to pick up ones that fall through the cracks.
In other words, I still have a significant list in front of me for sentences that do not have all the grammar marked up inside of them, and I'm knocking them out, a few expressions a day.