lojban / jbovlaste Goto Github PK
View Code? Open in Web Editor NEWHome Page: http://jbovlaste.lojban.org
Home Page: http://jbovlaste.lojban.org
From Nicholas Pinney via email:
The dictionary http://jbovlaste.lojban.org//lookup.pl performs case
sensitive queries. A search for "Green" finds no words, bit a search for
"green" does. I'm pretty sure this isn't intended behavior; it certainly
took me a few tries to work out. (I'm typing on a phone and the keyboard
automatically capitalises the first letter of a sentence)
Right now, the only way to find out about the newest words and discussions on jbovlaste is to go to http://jbovlaste.lojban.org/recent.html
The only emails I get are when someone edits my words or adds a comment.
I would like to have the option to receive email containing as much information as I like, including the full content of every new discussion (whether or not I'm involved in it) and even all new definitions added.
I remember discussing it more than a year ago. We need to use East Asia fonts in Chinese/Korean/Japanese pdf exports. See yourself. It shows squares. also it'd be better to embed fonts into the resulting pdf.
From gleki:
Since vlatai has a bug could you please add the following words
with any definition (even empty) in Test Language?
februari, martio, prilio, madjio, djunio, djulio, matriocka, patriarka, minstreli
Experimental gismu and cmavo are clearly denoted as such, so there’s no reason why experimental lujvo and fu'ivla aren’t.
Maybe this could be automated?
A lujvo is called “experimental” if at least one of its rafsi is derived from an experimental gismu or cmavo.
A fu'ivla is called “experimental” if it is in stage 3 and the used “topic rafsi” is derived from an experimental gismu or cmavo.
In future jvs 2.0 we should have a switch to change from viewing text in traditional chinese characters and in simplified chinese characters.
There no need in having them as separate languages as those are just two alternative orthographies (as if we had a switch for reading Lojban words in Tengwar).
tsani wrote:
Seeing as how important it is, could "lojban" be placed as the first language
in the list of languages in which one can define a word in jbovlaste?
https://github.com/lojban/jbovlaste/blob/master/export/latex-export.html#L163
\title{). $escapedlang . q( To lojban and lojban To ). $escapedlang .q( Dictionary}
As you can see English words are hardcoded. I suggest to change the relevant part to the following:
lo vlaste noi fanva fi lo escapeall($langrealname) la .lojban. gi'e fi la .lojban. escapeall($langrealname)
We also have
\chapter{lojban to ). escapeall($langrealname) .q(}).
This can be translated as
fanva fi la lojban. fo lo escapeall($langrealname)
\author{Lojbanic Community}
=> \author{lo lojbo cecmu}
\date{\today}
=> Should be in Lojban format!
What are your opinions?
Etymology field is awkward. There are two possible solutions:
From selpa'i:
I know some people wanted to be able to easily see who created a word
From rlpowell in #6:
It would be super nice to have the definition submission make a
micro latex file to test that you didn't screw up, and show the user errors as
necessary.
In the absence of fixing on a per-definition basis, the export should at least report the
LaTeX output to the user. :P:P:P But: The errors don't tell you where you are without
grepping teh latex. -_-
Clicking on natlang (all or preferred) in the menu yields a defective header:
Word listings for ""
From gleki:
word - definition id
bamju - 20732
bamju - 36750
balnu - 39121
balnu - 39123
balnu - 39122
balnu - 56024
vetli - 36700
tamsi - 36698
tamsi - 36699
LaTeX support seems to be broken. Here’s the example for the official definition of dekto (English, definition ID 1): “x1 mathend000# is ten [10; 1*101 mathend000#] of x2 mathend000# in dimension/aspect x3 mathend000# (default is units).”
From rlpowell:
Something to work on: Math, in particular ^, breaks PDF export if it's not wrapped
properly in $...$. It would be super nice to have the definition submission make a
micro latex file to test that you didn't screw up, and show the user errors as
necessary.
In the absence of fixing on a per-definition basis, the export should at least report the
LaTeX output to the user. :P:P:P But: The errors don't tell you where you are without
grepping teh latex. -_-
Valsi such as {denpa bu} and {slaka bu} are currently identified as "cmavo", although this is not entirely accurate. These should be reassigned a new category, "bu letteral", and the morphology check should assign similar compounds to this category.
See also #38 and this discussion:
I think the votes should be public information, and possibly with required commentary. For instance, if your word gets downvoted, it would be helpful to know what is wrong and who thinks so.
Currently it exports all the words (even downvoted) but it should export only words that have the vote balance >=1. It seems that it worked exactly like this until recently. Something got broken.
dag reports that his XML exports in haskell are dramatically quicker than jbovlaste.
Currently, it is possible to make exactly one vote only for one definition per language, either positive or negative. If there are multiple definitions, it is therefore not possible to cast votes for each defininition.
I understand this makes sense for positive votes, to indicate a preference.
But what if you think all definitions for a given language are bad? You have as well to choose one, while being neutral to all other definitions.
Plus, you can’t vote positive if you voted negative and vice-versa.
This voting system is weird.
Maybe the voting system should be rethought.
"cmavo clusters" are referred to as "compound cmavo" in CLL. xorxes points out that this is misleading, since the compounds do not form a single cmavo, so "cmavo compound" is more accurate.
CLL should probably be updated to use "cmavo compound", but even without such an update, if jbovlaste uses "cmavo compound" instead of "cmavo cluster", it's more likely that users of jbovlaste and CLL will correctly identify them as the same thing.
See also #38 and this discussion:
Per #26 and #37, there are issues with vlatai (part of jbofi'e) which are preventing some words from being entered. As part of a process of evaluating camxes as a replacement, I'm auditing the classification of words that are already entered into jbovlaste, to ensure that the classification performed by camxes conforms to expectations.
This issue identifies words that are marked as "fu'ivla" in jbovlaste, which are not identified as such under camxes.
Non-lojban words ("nonLojbanWord" per camxes)
balnkpi
bliardo
blueta
cidjrbentou
cidjrsurstrmi
cipnrbuteo
ckoala
cmiai
driomedeida
etreo
fiprkti'omizo
fiprpsefuru
fiprpseudoskafirinku
fiprtetrapleurodo
freskeo
infarkta
kriofla
ku'urpicea
melpsita
mudrfselia
mudrpterokarpu
pervu'ui
ricrbeaukarne'a
ricrpterokarpu
ridrdverga
rozrbanksi
rutrskuamosa
samcrarkti
skientia
sparanakampti
spararktantemu
spararkti
spareukomi
sparipeastru
sparksitropi
sparnimfea
sparpersea
sparpsikopsi
sparpsofokarpu
spatrleoxari
stagrleoxari
tanxebentou
tarksako
tirxpardu
trueno
xarjrngiri
Other non-fu'ivla per camxes
aierne: "cmavo + fu'ivla" per camxes
selda'ergau: "lujvo" per camxes
siukrida: "cmavo + gismu" per camxes
per gleki:
we have "experimental gismu" in jbovlaste. in pdf export they are marked with a
triangle. I think we should have a similar class of gismu and cmavo called "obsolete".
However, placing this or that word into this category would be done only by directly
editing the database and on behalf of Robin and nobody else. in pdf export they
should be marked as [obsolete] adter the word (or something similar )
definitions.definitionnum seem to be intended as a serial number for ordering the definitions for a particular valsi and language. But it appears that sometimes they are set to the same value as definitionid, e.g. one of the chinese definitions for finti
Either jbovlaste should always display the definitionid, or definitionnum as displayed should be set to a different, meaningful value.
A global score threshold would filter out any words or definitions that fail to meet a certain scoring threshold. It could be deactivated on a per user or per session basis by users who wish to browse all content, even low-scoring content.
acolotl: Sometimes it seems people downvote words for no reason whatsoever.
acolotl: Even with completely normal words.
durka42: this happens anywhere a downvote button is provided :p
acolotl: Yeah, but here it causes the words not to show up.
durka42: a word could start with more than 1 vote initially
durka42: like, I dunno, 5
acolotl: Yeah, at this point even 3 would make a big difference.
Here’s an usabilty suggestion:
If an user searches for a lujvo, but it is not in its lowest-scoring form, jbovlaste should still be able to recognize the lujvo and redirect the user to the lowest-scoring form. If there is no entry, the request fails, but if there is, the user finds the lujvo he/she/it requested. This would be helpful because it is unlikely an user does the score calculation in the head. ;-)
Currently, jbovlaste simply would not find the entry, although the user entered a perfectly valid lujvo.
Adding this feature would be also be closer to Lojban, because according to the CLL, all valid forms of a lujvo are equal.
It may be neccessary to add support for manually added redirects for the extremely rare cases a lujvo has two or more lowest-scoring forms.
Lujvo formed with {zei} are currently split by jbovlaste, with the various parts parsed by vlatai. If the parsing is approved, the word is entered as a lujvo. This issue proposes to reclassify those lujvo as "zei lujvo" and to change the morphology check so that it assigns this new category as appropriate.
See also #41 and this discussion:
Currently in order to place links to similar words one has to use "Notes" field. However, I propose splitting it into two fields:
Per #26 and #37, there are issues with vlatai (part of jbofi'e) which are preventing some words from being entered. As part of a process of evaluating camxes as a replacement, I'm auditing the classification of words that are already entered into jbovlaste, to ensure that the classification performed by camxes conforms to expectations.
This issue identifies words that are marked as "lujvo" in jbovlaste, which are not identified as such under camxes. They are all {zei} lujvo. Jbovlaste currently partially bypasses vlatai when a word contains {zei}.
cirla zei burgere (gismu, cmavo, fu'ivla)
cu'a zei fancu (cmavo, cmavo, gismu)
dropanra zei ionti (lujvo, cmavo, fu'ivla)
ga'e zei lerfu (cmavo, cmavo, gismu)
gregori zei nanca (fu'ivla, cmavo, gismu)
jgitrviolino zei konceto (fu'ivla, cmavo, fu'ivla)
jgitrviolono zei konceto (fu'ivla, cmavo, fu'ivla)
konceto zei pagbu (fu'ivla, cmavo, gismu)
ma'u zei ionti (cmavo, cmavo, fu'ivla)
ni'u zei ionti (cmavo, cmavo, fu'ivla)
pipnrpiano zei konceto (fu'ivla, cmavo, fu'ivla)
simfoni zei pagbu (fu'ivla, cmavo, fu'ivla)
tarbi zei asna (gismu, cmavo, fu'ivla)
te'o zei dugri (cmavo, cmavo, gismu)
to'a zei lerfu (cmavo, cmavo, gismu)
ubutycys zei cacra (cmene, cmavo, gismu)
ze'i zei seldejni (cmavo, cmavo, lujvo)
ze'u zei seldejni (cmavo, cmavo, lujvo)
zo si si zei fa'o (cmavo x 5)
From gleki:
we need a new field for cmavo in jbovlaste 2.0: a field for the terminator for each
cmavo (if applicable). Just like we have now with selma'o.
ofc. they also must be clickable but that's too easy. just dont forget
In a subdirectory such as ...
http://jbovlaste.lojban.org/needed/
... the menulinks render as relative, and thus are broken, e.g
http://jbovlaste.lojban.org/needed/export/latex.html
They also don't include port number, for those who wish to run jbovlaste on a non-standard port.
From guskant via gleki:
btw guskant worte that xml export doesnt export all the db in simple english direction
and probably in chinese
Per #26 and #37, there are issues with vlatai (part of jbofi'e) which are preventing some words from being entered. As part of a process of evaluating camxes as a replacement, I'm auditing the classification of words that are already entered into jbovlaste, to ensure that the classification performed by camxes conforms to expectations.
This issue identifies words that are marked as "cmene" in jbovlaste, which are not considered as lojban words ("nonLojbanWord") under camxes:
altfor
arktik
bivmast
cibmast
daumast
djordj
feimast
gaimast (previously entered as "geimast" but may be a transcription error)
guanJOUS
maskFAS
mumymast
pavmast
relmast
slovensk
sozmast
tai,UAN
tcesk
vonmast
xavmast
xrvatsk
zelmast
There are currently more than 350 natural language votes which attribute natural language words (glosswords or place keywords) from one language to definitions from another language:
select count(*)
from natlangwordvotes v
join natlangwords w on w.wordid = v.natlangwordid
join definitions d on d.definitionid = v.definitionid
where w.langid <> d.langid;
It seems likely this is at least largely due to "dict/votebits" looking up natlangwords without including langid in the query.
jvs xml export should include "the number of upvotes minus the number of downvotes" for each word.
Per #26 and #37, there are issues with vlatai (part of jbofi'e) which are preventing some words from being entered. As part of a process of evaluating camxes as a replacement, I'm auditing the classification of words that are already entered into jbovlaste, to ensure that the classification performed by camxes conforms to expectations.
This issue identifies words that are marked as "cmavo" or "cmavo clusters" (CLL: "compound cmavo") in jbovlaste, which are not identified as such under camxes.
cmavo
y: "initialSpaces" per camxes
cmavo cluster
denpa bu: "gismu + cmavo" per camxes
slaka bu: "gismu + cmavo" per camxes
ybu: single cmavo per camxes
{.ua .uu} can't be entered into JVS. But this poses an interesting question: if compound cmavo can be entered why any other compound like tanru or even complete bridi can't be entered?
In fact I can enter {do du} but can't enter {do klama}!
I am not actually sure about this one, but I heard jbovlaste uses vlatai (an (almost?) undocumented tool which comes with jbofi'e) to check the morphology of words. The problem with that is that vlatai is buggy and does not correctly check the morphology.
I am not sure if I am correct here. Somebody please verifiy.
I can't find a link for http://jbovlaste.lojban.org/languages.html from the frontend.
It must be hidden somewhere which is a bit silly. May be just place it to the main (side) menu?
It is not possible to skip a place key word. If you want to specify play keyword 2 only, you also have to specify place keyword 1. If you just specify keyword 2 anyways, jbovlaste will actually put this keyword to place 1.
A current ugly workaround exists. Simply fill the places to be skipped with some nonsense character, like “-”.
Natural language words that are entered with both a "default" sense as well as at least one explicit sense are displayed with an error:
Data Inconsistency
There appears to be both a default meaning for this word, and one or more
specialized meanings. This makes little sense, and should be corrected.
e.g. http://jbovlaste.lojban.org/natlang/en/extreme
There are currently more than 800 language/word pairs in the database which exhibit this problem, as indicated by this query:
select w.langid, w.word, count(*)
from natlangwords w
where exists
(select 1 from natlangwords ww
where ww.langid = w.langid and ww.word = w.word
and ww.meaning is null)
and exists
(select 1 from natlangwords www
where www.langid = w.langid and www.word = w.word
and www.meaning is not null)
group by w.langid, w.word;
Either this rule should be enforced and the problem cases cleaned up, or the rule should be scrapped along with this error message.
We should think of implementing camxes.js in JVS or JVS 2.0 Obviously it shouldn't be just a link to http://ilmen.tk/lojban/camxes.html
May be you enter a sentence, it parses it, the result gives links to entries in the database, a switch would allow you to changes the output language in the glossing.
we have "experimental gismu". in pdf export
they are marked with a triangle. we should have a
similar class of gismu and cmavo called "obsolete". However,
placing this or that word into this category would be done only
by directly editing the database and on behalf of Robin and
nobody else. in pdf export they should be marked as [obsolete]
after the word (or something similar )
Vlasisku has all the interface in English. It's completely unacceptable. Vlasisku.ru is running with the same interface although it is to be used by speakers of Russian.
I can't generate gua\spi pdf.
Looks like the name of gua\spi needs to be escaped for TeX
Per #39 and #40, some words entered into jbovlaste as fu'ivla and cmene are not valid according to camxes. In anticipation of replacing vlatai (jbofi'e) as the morphological classifier in jbovlaste (per #26), these words will be assigned new types, "obsolete fu'ivla" and "obsolete cmene". These types may be the first steps to address #9.
There's no need to mix test data with production data. Let's make a test instance of jbovlaste, then remove all of the test data from the production instance.
It would be nice to illustrate some words with images taken from Wikimedia and/or properly licensed for us (free modification and copying even commerially would be the best option). I guess any of us can upload an image illustrating {plise}.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.