Difference between revisions of "Lttoolbox-java"

From Apertium
Jump to navigation Jump to search
 
(56 intermediate revisions by 11 users not shown)
Line 1: Line 1:
[[Lttoolbox-java (français)]]
Notes


{{TOCD}}

== What is lttoolbox-java ==
lttoolbox-java is a Java port of the whole Apertium runtime system, ''including both lttoolbox and apertium''.

=== lttoolbox functions ===
[[lttoolbox]] can do the following:
* Compile: make binary files out of the .dix files (lt-comp),
* Proces: analysing or generating text (lt-proc) and
* Expand: Expand a dictionary .dix file (lt-expand).

The Java port of lttoolbox is also capable of
* [[Compounds]] (experimental)
* [[lttoolbox-java/Flag diacritics]] (highly experimental)
* Validate .dix files

=== apertium runtime functions ===

The Java port implements the typical functions used by Apertium during runtime.

* Read .mode files and execute the steps included in them
* Execute the tagger
* Execute transfer stages (all 3 of them)

The Java port needs the C++ binaries for preparing/developing a language pair, i.a. to compile transfer files and train the tagger.


The Java port of lttoolbox is also capable of
* Generate [[bytecode for transfer]] and execute it. The bytecode runs typically 10 times faster than the C++ version.



==Why==

A "Java port" of Apertium enables use on

* Windows,
* Android phones,
* Cross-platform desktop application,
* Java server applications.

The last 2 is relevant as, for example a LibreOffice plugin should be platform independent to be maintainable.

We havent seen anyone embedding Apertium in a desktop application. Currently Apertium is usable in a local subdir but installation isnt trivial to an end user.

Having a packaged easy-to-use version of Apertium ready for embedding MT in a larger program would be very cool.
Ideally should a self-contained Apertium JAR file, only dependent on JRE and an additional JAR file per language pair.

Another "embedding" approach is to use a client stub to one of our [[Apertium services]], but there can be reasons why people prefers to have things installed locally (we don't need to repeat them here).

==Features==
* Binary compatibility with lttoolbox. lttoolbox-java is able '''read''' and '''write''' the binary files lttoolbox and generates exactly the same output
* There is a comprehensive test suite that tests both lttoolbox (C++) and lttoolbox-java.

==Installation==
===Prerequisites: ===
* java-runtime
* apache-ant (for compilation)

Under Arch Linux, you can install the prerequisites with

pacman -S openjdk6 apache-ant

Under Debian/Ubuntu:

sudo apt-get install ant ant-optional # what else??
===Download, compile, install===
Download the newest [https://github.com/apertium/lttoolbox-java/releases release] or check out from github:
<pre>
<pre>
git clone https://github.com/apertium/lttoolbox-java/
<jimregan> Nic Cottrell contributed a Java port of lttoolbox
</pre>
<jimregan> but it needs work to finish it
<jimregan> and a test suite, in both C++ and Java
<jacobEo> Its in in apertium-tools/lttoolbox-java
<jacobEo> What is in apertium-tools/lttoolbox-java right now is NOT working.
<jacobEo> It's a line-for-line port of the C++ code of lttoolbox


Use Netbeans or Unix, whatever suits you best:
<pre>
sh autogen.sh
make
sudo make install
</pre>


You can also build and install using Maven 2 (http://maven.apache.org), by typing:
<jacobEo> and the great problem is the XML handling
<pre>
mvn install -DskipTests
</pre>


See also the README file


== Usage ==
<jimregan> it has to be binary compatible
<pre>
<jimregan> and the test suite has to be in both C++ and Java, to ensure that
$ java -jar dist/lttoolbox.jar
<jimregan> yeah, it's amost line for line identical to the C++, aside from Java/C++ differences
lttoolbox: is a toolbox for lexical processing, morphological analysis and generation of words
<jimregan> but, the binary stuff can be hard
USAGE: java -jar dist/lttoolbox.jar [task]
<jacobEo> therefore jimregan its not that hard.
Examples:
<jimregan> all you need is one bit in the wrong place, and it's useless
java -jar dist/lttoolbox.jar lt-expand dictionary.dix expands a dictionary
<jacobEo> jimregan: Binary stuff?
java -jar dist/lttoolbox.jar lt-comp lr dic.dix dic.bin compiles a dictionary
<jimregan> medium, then
java -jar dist/lttoolbox.jar lt-proc dic.bin morphological analysis
<jimregan> jacobEo, yeah
</pre>
<jimregan> well
<jimregan> the compression stuff
<cseong> if i dont know one of the required language, for example is C,C++ and XML are the requirements and i dont know XML, can i still choose it ?
<jimregan> and the transducer
<jacobEo> jimregan: The binary stuff is _probably_ easy, as you can debug the C++ and compare variables etc
<jimregan> cseong, XML is easy to pick up
<jimregan> there are plenty of APIs availabl
<jacobEo> cseong: Which project are you thinking of?
<jimregan> for C++, we use libxml2


or, using the a shell scripts:


<pre>
<jacobEo> Rah2: lttoolbox are making binary files out of the .dix files.
$ lt-comp-j
<jacobEo> Rah2: lttoolbox-java needs to at least be able to _read_ these binary files.
v3.2j: build a letter transducer from a dictionary
<jacobEo> Rah2: Pls compile lttoolbox and apertium and a language pair of your choice.
USAGE: LTComp lr | rl dictionary_file output_file [acx_file]
<jacobEo> Rah2: Then much more will be clear
Modes:
lr: left-to-right compilation
rl: right-to-left compilation
</pre>




<pre>
<jacobEo> Rah2: You don't need much knowlede of MT or NLP to do lttoolbox-java. But you need to know C++ and Java and be able to debug both
$ lt-proc-j
<Drew_> jacobEo: What was the location of lttoolbox again?
LTProc: process a stream with a letter transducer
<jacobEo> http://apertium.svn.sourceforge.net/viewvc/apertium/trunk/lttoolbox/
USAGE: LTProc [-c] [-a|-g|-n|-d|-b|-p|-s|-t] fst_file [input_file [output_file]]
<jacobEo> http://apertium.svn.sourceforge.net/viewvc/apertium/trunk/apertium-tools/lttoolbox-java/
Options:
<jacobEo> "Download GNU tarball" will give a compressed archive
-a: morphological analysis (default behavior)
-c: use the literal case of the incoming characters
-e: morphological analysis, with compound analysis on unknown words
-f: match flags (experimental)
-g: morphological generation
-n: morph. generation without unknown word marks
-d: morph. generation with all the stuff
-t: morph. generation, but retaining part-of-speech
-p: post-generation
-s: SAO annotation system input processing
-t: apply transliteration dictionary
-z: flush output on the null character
-v: version
-D: debug; print diagnostics to stderr
-h: show this help
</pre>


<jacobEo> The problem, I think, is the XML handling: The C code's library callback calls a method in the code both when it meets a START and an END tag.
<jimregan> avinesh, maybe you should say it to spectie because I'm not interested in unicode -> wx
<jimregan> not for any reason
<jacobEo> the Java's XML library only calls the callback method at the START tag.
<jimregan> jacobEo, that will be necessary for chunk merging
<jimregan> we don't have it yet, but it will be necessary
<Leftmost> jimregan, I guess I'm a bit unclear as to what form the regression tests should take. Simply translations between ga and gd?
<jimregan> because when chcontent in t2 is written in chunk mode, it will be without { or }, otherwise with
<avinesh> ok got it
<jimregan> to fit the current model, that has to be a bool set and unset on entry/exit
<Drew_> jacobEo: Is it a big job to make it work with the END tag?
<avinesh> no wx right :D
<jimregan> that's it
<jimregan> avinesh, noone told me anudev is a course supervisor :/
<jacobEo> Drew_: I don't know. Perhaps we could find another Java XML library that could be made also call for the end tags. Or some kind of wrapper-inbetween thing could be made. Or you could use SAX and make your own callback thing.
<jimregan> I think I would have expected more of his opinions if I knew he wasn't actually doing any of the work
<avinesh> umm he mainly working on anusaraka
<jacobEo> Drew_: There might be other problems. The project just got stranded on the XML parse part.
<Drew_> jacobEo: Ah, ok. I'm just compiling it now
<jacobEo> Drew_: You have to run the code to see. To do that you need to have at least one language pair runnning on your machine
* vaasu (n=yt@123.176.16.43) has joined #apertium
<Drew_> jacobEo: I can't find a main class in the source code, am I looking in the wrong place? :S
<jacobEo> Drew_: The Java code?
<cseong> uhm..i am interested in improving interoperability..but what formats are u refering to ?
<Drew_> jacobEo: Yeah, I loaded the java code into eclipse but it can't find a main method to compile the .java's
<jacobEo> Drew_: LTComp.java, LTExpand.java, LTProc.java
<jimregan> avinesh, yeah. So I was right when I thought he expected us to change all of apertium to suit the analyser :/
* abhiSri (i=AB-Alway@220.224.99.238) has joined #apertium
<jacobEo> Drew_: Use Netbeans if you can. It's kinda standard here in Apertium


<pre>
$ lt-expand-j
v3.2j: expand the contents of a dictionary fileUSAGE: LTExpand dictionary_file [output_file]
</pre>




<pre>
$ lt-validate-j
v3.2j: validate an XML file according to a schema
USAGE : LTValidate -dix dictionary.xml
LTValidate -acx dictionary.acx
</pre>


===Examples===
<jacobEo_> Rah1: Great. The problem right now is that the XML parser in C++ and in Java behaves differently.
<jacobEo_> Rah1: C++ parser calls back when a tag BEGINS and when it ENDS
<jacobEo_> Rah1: Java parser calls back only when a tag BEGINS
<jacobEo_> Rah1: Therefore the code for Java parsing is not working.
<jacobEo_> Rah1: Apart from that there is probably some minor things.
<Rah1> hmm, ok
<jacobEo_> Rah1: You will have to debug the C++ and the Java version and compare executions.
<jacobEo_> Rah1: The Java version is a line-for-line more or less exact port of the C++.
<Rah1> yes, I think I can do that
<jacobEo_> Rah1: But the languages are different. C++ for example has some methods where some simple type variables are changed (the reference is passed)
<jacobEo_> Rah1: But in Java simple type variables can only be passed by value, and thus the caller's value is not changes.
<jacobEo_> Rah1: That sort of things needs to be sorted out.
<Rah1> Yeah, I see the problem
<Rah1> and I think I can take care of it
<jacobEo_> So Rah1 You don't have to know much about linguistics. You only have to understand what lt-expand, lt-comp and lt-proc does with a .dix file


Use the new compounding feature:
echo "lambakjöti" | java -jar dist/lttoolbox.jar lt-proc -e /home/j/esperanto/apertium/apertium-is-en/is.bin
^lambakjöti/lamb<n><nt><pl><gen><ind>+kjöt<n><nt><sg><dat><ind>$


===Encoding problems===
Try -Dfile.encoding=UTF-8, like


echo "lambakjöti" | java -Dfile.encoding=UTF-8 -jar dist/lttoolbox.jar lt-proc -e /home/j/esperanto/apertium/apertium-is-en/is.bin
^lambakjöti/lamb<n><nt><pl><gen><ind>+kjöt<n><nt><sg><dat><ind>$

===Mac users===
You need JDK1.6. Try

/System/Library/Frameworks/JavaVM.framework/Versions/1.6/Commands/java -jar dist/lttoolbox.jar


=== Windows usage ===

By default, the windows console uses UTF-16, whereas apertium's data
is encoded with utf-8. This command switches the dos box to utf-8:

<pre>
chcp 65001
</pre>
</pre>

Note: you also need to use an unicode-capable font for the windows console, like Lucida Console (Properties -> Font).

Also, don't forget to set these Runtime flags: -Xms64m -Xmx800 -Dfile.encoding=UTF-8

== Reasons for a Java port ==
* There are several devices (mobile phones, for example) which can run quite complicated software, but only if written in Java. lttoolbox is the first step to having Apertium run on these devices.
* Windows port. It won't be as powerfull as Unix based system, but it will be there
* Apertium will be the first MT system *ever* that can be demonstrated within a Java applets
* Transfer in bytecode has a promise of speedup factor 4 - compared to what we use now (interpreted XML). And transfer CPU usage is dominating when processing large amounts of text

== Performance of Java port ==

Compatibility and performance can be checked by invoking test_java_and_c.sh in testdata/regression.

=== Single-core processor (Jimmy O'Regan)===
<pre>
java version "1.6.0_18"
OpenJDK Runtime Environment (IcedTea6 1.8) (6b18~pre4-1ubuntu1)
OpenJDK Client VM (build 16.0-b13, mixed mode, sharing)
C analysis is... 0.59sec
OK
Java analysis is... 1.15sec
OK
C generator -g is ... 0.54sec
OK
Java generator -g is ... 1.13sec
OK
C generator -d is ... 0.56sec
OK
Java generator -d is ... 1.12sec
OK
C generator -n is ... 0.52sec
OK
Java generator -n is ... 1.12sec
OK
C postgenerator -p is ... 0.07sec
OK
Java postgenerator -p is ... 0.33sec
OK
All tests passed
</pre>

=== Dual-core processor (Jacob)===
<pre>
Java HotSpot(TM) Client VM (build 1.6.0-beta2-b86, mixed mode, sharing)
C analysis is... 0.39sec
OK
Java analysis is... 0.66sec
OK
C generator -g is ... 0.32sec
OK
Java generator -g is ... 0.62sec
OK
C generator -d is ... 0.33sec
OK
Java generator -d is ... 0.58sec
OK
C generator -n is ... 0.32sec
OK
Java generator -n is ... 0.64sec
OK
C postgenerator -p is ... 0.03sec
OK
Java postgenerator -p is ... 0.20sec
OK
All tests passed
</pre>

As you see Java version is currently (april 2010) a factor 2 slower than the C version. There are ways to remedy this (using simple types collection classes), but it hasnt been implemented, as no-one has requested it.

It still gives great performance, however, and Apertium running on Java is very fast, compared to other MT systems. The overhead of using the Java version instead of the C version is negligible, as transfer is the big ressource hog anyway.

The above test compares the basic lttoolbox functions. As Java transfer is much faster the result of performance test of a pure-Java and pure-C++ chain are comparable (and mostly in Java's favor). A hybrid can be made which beats performance of both systems.

== Known bugs ==

Oct 2019:
lttoolbox-java kan read files compiled with lttoolbox version 3.5.0
*note* lttoolbox-java currently lacks support for functionality added the last 5 years - and it doesent work in Java JDK 9+, as it uses BCEL classes that was embedded Java 8 but changed package name in Java 9.
Instead we should include BCEL as a library.

==Thanks==
* Nic Cottrell contributed an initial version of a Java port of [[lttoolbox]].
* During [[Google Summer of Code|GSOC2009]] [[User:Rah|Raphaël]] and [[User:Sortiz|Sergio]] worked on it, but processing still didnt work (compilation and expansion worked)
* November 2009 [[User:Jacob Nordfalk|Jacob Nordfalk]] finished it up and optimized some parts of it
* During GSOC 2010 Jacob mentored a full [[User:Kanmuri/GSoC 2010 Application/Java Runtime Port|Java Runtime Port of Apertium]] (i.e. not just lttoolbox) made by [[User:Kanmuri|Stephen Tigner]].

[[Category:Lttoolbox]]
[[Category:Documentation in English]]

Latest revision as of 09:49, 7 April 2020

Lttoolbox-java (français)

What is lttoolbox-java[edit]

lttoolbox-java is a Java port of the whole Apertium runtime system, including both lttoolbox and apertium.

lttoolbox functions[edit]

lttoolbox can do the following:

  • Compile: make binary files out of the .dix files (lt-comp),
  • Proces: analysing or generating text (lt-proc) and
  • Expand: Expand a dictionary .dix file (lt-expand).

The Java port of lttoolbox is also capable of

apertium runtime functions[edit]

The Java port implements the typical functions used by Apertium during runtime.

  • Read .mode files and execute the steps included in them
  • Execute the tagger
  • Execute transfer stages (all 3 of them)

The Java port needs the C++ binaries for preparing/developing a language pair, i.a. to compile transfer files and train the tagger.


The Java port of lttoolbox is also capable of

  • Generate bytecode for transfer and execute it. The bytecode runs typically 10 times faster than the C++ version.


Why[edit]

A "Java port" of Apertium enables use on

  • Windows,
  • Android phones,
  • Cross-platform desktop application,
  • Java server applications.

The last 2 is relevant as, for example a LibreOffice plugin should be platform independent to be maintainable.

We havent seen anyone embedding Apertium in a desktop application. Currently Apertium is usable in a local subdir but installation isnt trivial to an end user.

Having a packaged easy-to-use version of Apertium ready for embedding MT in a larger program would be very cool. Ideally should a self-contained Apertium JAR file, only dependent on JRE and an additional JAR file per language pair.

Another "embedding" approach is to use a client stub to one of our Apertium services, but there can be reasons why people prefers to have things installed locally (we don't need to repeat them here).

Features[edit]

  • Binary compatibility with lttoolbox. lttoolbox-java is able read and write the binary files lttoolbox and generates exactly the same output
  • There is a comprehensive test suite that tests both lttoolbox (C++) and lttoolbox-java.

Installation[edit]

Prerequisites:[edit]

  • java-runtime
  • apache-ant (for compilation)

Under Arch Linux, you can install the prerequisites with 

pacman -S openjdk6 apache-ant

Under Debian/Ubuntu: 

sudo apt-get install ant ant-optional      # what else??

Download, compile, install[edit]

Download the newest release or check out from github: 

git clone https://github.com/apertium/lttoolbox-java/

Use Netbeans or Unix, whatever suits you best: 

sh autogen.sh
make
sudo make install

You can also build and install using Maven 2 (http://maven.apache.org), by typing: 

mvn install -DskipTests

See also the README file

Usage[edit]

$ java -jar dist/lttoolbox.jar
lttoolbox: is a toolbox for lexical processing, morphological analysis and generation of words
USAGE: java -jar dist/lttoolbox.jar [task]
Examples:
 java -jar dist/lttoolbox.jar lt-expand dictionary.dix     expands a dictionary
 java -jar dist/lttoolbox.jar lt-comp lr dic.dix dic.bin   compiles a dictionary
 java -jar dist/lttoolbox.jar lt-proc dic.bin              morphological analysis

or, using the a shell scripts: 

$ lt-comp-j 
 v3.2j: build a letter transducer from a dictionary
USAGE: LTComp lr | rl dictionary_file output_file [acx_file]
Modes:
  lr:     left-to-right compilation
  rl:     right-to-left compilation



$ lt-proc-j
LTProc: process a stream with a letter transducer
USAGE: LTProc [-c] [-a|-g|-n|-d|-b|-p|-s|-t] fst_file [input_file [output_file]]
Options:
  -a:   morphological analysis (default behavior)
  -c:   use the literal case of the incoming characters
  -e:   morphological analysis, with compound analysis on unknown words
  -f:   match flags (experimental)
  -g:   morphological generation
  -n:   morph. generation without unknown word marks
  -d:   morph. generation with all the stuff
  -t:   morph. generation, but retaining part-of-speech
  -p:   post-generation
  -s:   SAO annotation system input processing
  -t:   apply transliteration dictionary
  -z:   flush output on the null character 
  -v:   version
  -D:   debug; print diagnostics to stderr
  -h:   show this help



$ lt-expand-j 
 v3.2j: expand the contents of a dictionary fileUSAGE: LTExpand dictionary_file [output_file]



$ lt-validate-j 
 v3.2j: validate an XML file according to a schema
USAGE : LTValidate -dix dictionary.xml
        LTValidate -acx dictionary.acx

Examples[edit]

Use the new compounding feature: 

echo "lambakjöti" | java -jar dist/lttoolbox.jar lt-proc -e /home/j/esperanto/apertium/apertium-is-en/is.bin

^lambakjöti/lamb<n><nt><pl><gen><ind>+kjöt<n><nt><sg><dat><ind>$

Encoding problems[edit]

Try -Dfile.encoding=UTF-8, like 

echo "lambakjöti" | java -Dfile.encoding=UTF-8 -jar dist/lttoolbox.jar lt-proc -e /home/j/esperanto/apertium/apertium-is-en/is.bin

^lambakjöti/lamb<n><nt><pl><gen><ind>+kjöt<n><nt><sg><dat><ind>$

Mac users[edit]

You need JDK1.6. Try 

/System/Library/Frameworks/JavaVM.framework/Versions/1.6/Commands/java -jar dist/lttoolbox.jar


Windows usage[edit]

By default, the windows console uses UTF-16, whereas apertium's data is encoded with utf-8. This command switches the dos box to utf-8: 

chcp 65001

Note: you also need to use an unicode-capable font for the windows console, like Lucida Console (Properties -> Font).

Also, don't forget to set these Runtime flags: -Xms64m -Xmx800 -Dfile.encoding=UTF-8

Reasons for a Java port[edit]

  • There are several devices (mobile phones, for example) which can run quite complicated software, but only if written in Java. lttoolbox is the first step to having Apertium run on these devices.
  • Windows port. It won't be as powerfull as Unix based system, but it will be there
  • Apertium will be the first MT system *ever* that can be demonstrated within a Java applets
  • Transfer in bytecode has a promise of speedup factor 4 - compared to what we use now (interpreted XML). And transfer CPU usage is dominating when processing large amounts of text

Performance of Java port[edit]

Compatibility and performance can be checked by invoking test_java_and_c.sh in testdata/regression.

Single-core processor (Jimmy O'Regan)[edit]

java version "1.6.0_18"
OpenJDK Runtime Environment (IcedTea6 1.8) (6b18~pre4-1ubuntu1)
OpenJDK Client VM (build 16.0-b13, mixed mode, sharing)
C analysis is... 0.59sec
OK
Java analysis is... 1.15sec
OK
C generator -g is ... 0.54sec
OK
Java generator -g is ... 1.13sec
OK
C generator -d is ... 0.56sec
OK
Java generator -d is ... 1.12sec
OK
C generator -n is ... 0.52sec
OK
Java generator -n is ... 1.12sec
OK
C postgenerator -p is ... 0.07sec
OK
Java postgenerator -p is ... 0.33sec
OK
All tests passed

Dual-core processor (Jacob)[edit]

Java HotSpot(TM) Client VM (build 1.6.0-beta2-b86, mixed mode, sharing)
C analysis is... 0.39sec
OK
Java analysis is... 0.66sec
OK
C generator -g is ... 0.32sec
OK
Java generator -g is ... 0.62sec
OK
C generator -d is ... 0.33sec
OK
Java generator -d is ... 0.58sec
OK
C generator -n is ... 0.32sec
OK
Java generator -n is ... 0.64sec
OK
C postgenerator -p is ... 0.03sec
OK
Java postgenerator -p is ... 0.20sec
OK
All tests passed

As you see Java version is currently (april 2010) a factor 2 slower than the C version. There are ways to remedy this (using simple types collection classes), but it hasnt been implemented, as no-one has requested it.

It still gives great performance, however, and Apertium running on Java is very fast, compared to other MT systems. The overhead of using the Java version instead of the C version is negligible, as transfer is the big ressource hog anyway.

The above test compares the basic lttoolbox functions. As Java transfer is much faster the result of performance test of a pure-Java and pure-C++ chain are comparable (and mostly in Java's favor). A hybrid can be made which beats performance of both systems.

Known bugs[edit]

Oct 2019: lttoolbox-java kan read files compiled with lttoolbox version 3.5.0

  • note* lttoolbox-java currently lacks support for functionality added the last 5 years - and it doesent work in Java JDK 9+, as it uses BCEL classes that was embedded Java 8 but changed package name in Java 9.

Instead we should include BCEL as a library.

Thanks[edit]

Retrieved from "https://wiki.apertium.org/w/index.php?title=Lttoolbox-java&oldid=72010"