xiaoling / figer Goto Github PK

View Code? Open in Web Editor NEW

126.0 8.0 39.0 11.88 MB

Fine-Grained Entity Recognizer

Home Page: http://xiaoling.github.io/figer/

License: Other

Scala 0.05% Shell 0.02% Java 99.93% HTML 0.01%

figer's Introduction

Fine-Grained Entity Recognizer (FIGER)

This distribution contains the source code for the experiments presented in the following research publication (PDF):

Xiao Ling and Daniel S. Weld (2012).
"Fine-Grained Entity Recognition",
in Proceedings OF THE TWENTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI), 2012.

Download the model file

One can test the trained model on the evaluation data or new data as they wish.

Run ./downloadModel.sh to download the Model and save it at the root directory. Run ./downloadModel.sh new for an updated model.

A better model has been trained and can be fetched by ./downloadModel.sh new. Change the config value accordingly.

Requirement

sbt >= 0.13.0

Replicate the experiments

To run the experiments in the AAAI-12 paper, you can proceed as follows:

$ ./run.sh "aaai/exp.conf" &> aaai/exp.log

Run FIGER on new data

To make predictions on new data, please see package edu.washington.cs.figer.FigerSystem for example code or run:

$ sbt "runMain edu.washington.cs.figer.FigerSystem <text_file>"

Alternatively, you can change the parameter values (e.g. the input file name) in config/figer.conf and get a more structured output by running:

$ ./run.sh "config/figer.conf"

Make a stand-alone jar

$ sbt assembly
# the actual path might be different
$ java -jar ./target/scala-2.10/figer-assembly-0.jar <text_file>

A simple web interface

Run

$ sbt ~container:start

and go to localhost:8080/index.html for a simple web demo.

Training Data

The training data train.data.gz (Download link) is gzipped and serialized in Protocol Buffer. Please see entity.proto in the code package for the definitions. Each Mention represents an entity mention defined by its token offsets together with the tokens in the sentence where the mention appears. The labels field shows the Freebase types of the underlying entity for the mention. Use the mapping in the next section to get the corresponding FIGER types.

In config/figer.conf, make the following changes:

useModel=false
modelFile=<the output model file>
# the training file has to follow the specs from `entity.proto`. See `train.data.gz` for example
trainFile=<training file>

Then run ./run.sh config/figer.conf to train a new model (It will need over 10G memory and about an hour to finish).

Mapping between Freebase MIDs and Wikipedia Titles

A mapping file from mids to titles is available here.

figer's People

Contributors

Stargazers

Watchers

figer's Issues

Some confusions about type (tag) taxonomy and mappings from entities to types

As i converted the train.tar.gz to text data, i've found that the types are a bit noisy, e.g.,
"LCD Soundsystem",
"/internet/social_network_user /broadcast/artist /music/artist /music/musical_group /common/topic".
There are some types not in the typeList in the research paper.

So, i wonder if there is a clear/specific mapping file, which maps entities exactly to some of the 112 types??

Anyway, i want to using the famous FIGER dataset in my experiments, so i have to pay attention to the details.

Thanks a lot!

java.lang.NullPointerException

After I ran the command ./run.sh "aaai/exp.conf" &> aaai/exp.log on my Mac High Sierra, I got the following response as recorded in the log. Any idea how to fix it?

[0m[�[0minfo�[0m] �[0mLoading project definition from /Users/chsuong/figer/project�[0m java.lang.NullPointerException at java.base/java.util.regex.Matcher.getTextLength(Matcher.java:1770) at java.base/java.util.regex.Matcher.reset(Matcher.java:416) at java.base/java.util.regex.Matcher.<init>(Matcher.java:253) at java.base/java.util.regex.Pattern.matcher(Pattern.java:1133) at java.base/java.util.regex.Pattern.split(Pattern.java:1261) at java.base/java.util.regex.Pattern.split(Pattern.java:1334) at sbt.IO$.pathSplit(IO.scala:723) at sbt.IO$.parseClasspath(IO.scala:821) at sbt.compiler.CompilerArguments.extClasspath(CompilerArguments.scala:64) at sbt.compiler.AggressiveCompile.withBootclasspath(AggressiveCompile.scala:50) at sbt.compiler.AggressiveCompile.compile2(AggressiveCompile.scala:83) at sbt.compiler.AggressiveCompile.compile1(AggressiveCompile.scala:70) at sbt.compiler.AggressiveCompile.apply(AggressiveCompile.scala:45) at sbt.Compiler$.apply(Compiler.scala:70) at sbt.Defaults$.sbt$Defaults$$compileTaskImpl(Defaults.scala:722) at sbt.Defaults$$anonfun$compileTask$1.apply(Defaults.scala:716) at sbt.Defaults$$anonfun$compileTask$1.apply(Defaults.scala:716) at scala.Function1$$anonfun$compose$1.apply(Function1.scala:47) at sbt.$tilde$greater$$anonfun$$u2219$1.apply(TypeFunctions.scala:42) at sbt.std.Transform$$anon$4.work(System.scala:64) at sbt.Execute$$anonfun$submit$1$$anonfun$apply$1.apply(Execute.scala:237) at sbt.Execute$$anonfun$submit$1$$anonfun$apply$1.apply(Execute.scala:237) at sbt.ErrorHandling$.wideConvert(ErrorHandling.scala:18) at sbt.Execute.work(Execute.scala:244) at sbt.Execute$$anonfun$submit$1.apply(Execute.scala:237) at sbt.Execute$$anonfun$submit$1.apply(Execute.scala:237) at sbt.ConcurrentRestrictions$$anon$4$$anonfun$1.apply(ConcurrentRestrictions.scala:160) at sbt.CompletionService$$anon$2.call(CompletionService.scala:30) at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264) at java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515) at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264) at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128) at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628) at java.base/java.lang.Thread.run(Thread.java:834)

Output for new predictions

Thanks for the useful tool! When running on new data, is there a way to write the predictions in BIO format to a file?

Absolute paths for testFile/outputFile cause problems

When I specify absolute paths for testFile and outputFile in the config file, each sentence with predicted mentions triggers a cryptic sid not found message and a NullPointerException, and is omitted from the output.

The workaround is to use a relative path. But a more informative error message, at least, would be helpful.

Buidling train data from protobuf and while modeling it from trained data. Unable to do that.

I want to build training data using protobuf. I explored entity.proto and reading the existing train.data file. I copied the single tagged sentence as it is and wrote a writer using protobuf and created train.data file. After that, I created model using that train.data file but I am unable to open. It gives Not a GZIP format.

I also want to know that how you create train.data using wikipedia data? Do you that code inside your git repository and where is it?

I am attaching writer file and figer.conf file both . PFA for that and remind me, if I am doing anything wrong.

To support attached, I have renamed the files.
Renamed figer.conf to figer.txt -> figer.txt
Renamed WriteEntityFile.java to WriteEntityFile.txt -> WriteEntityFile.txt

Wikipedia dump from where sentences were extracted

Greetings, Mr. Xiao

Could you please provide a link to download the WEX 20110513 Wiki dump from where you extracted your sentences?

Regards,
Luís

error: not found: value jetty

build.sbt:28: error: not found: value jetty
jetty()
^
[error] Type error in expression
Project loading failed: (r)etry, (q)uit, (l)ast, or (i)gnore? i
[warn] Ignoring load failure: no project loaded.
[error] Expected ID character
[error] Not a valid command: assembly
[error] assembly.sbt
[error] ^

The mapping from Wikipedia entities to Freebase Mids

Hi Xiao
I think for getting the types of Wikientities in the train.data, you had mapped them first to Freebase!
I wounder if you can provide the mapping file that you've used for that, because I need to get back the Freebase MIDs of entities! (I have a mapping file myself but a lot of entities are not covered there)
I would appreciate if you help me.
Thanks,
Yadollah

Unable to use generated jar from CLI

Hi,

I generated a jar of figer using this command from the readme:

sbt assembly

I now have 2 jar files ("figer_2.10-0.jar" and "figer-assembly-0.jar"), and am trying to use them to generate NE labels for a corpus of sentences.

Specifically, I've tried these commands:

java -jar figer_2.10-0.jar edu.washington.cs.figer.FigerSystem sentences.txt 
java -jar figer-assembly-0.jar edu.washington.cs.figer.FigerSystem sentences.txt

In both cases, I get this error:

no main manifest attribute

My questions are:

Which of these jars should I use?
Is there a manifest file in the jar? (Based on the error, I believe the error is a missing manifest file.)

Note that I'm able to execute sbt "runMain edu.washington.cs.figer.FigerSystem sentences.txt" successfully, so the issue is not the source code.

Thanks in advance for your help!

How to download Training data in Text Format

I want to download Traning data used in the FIGER system for another NER task to see if this data can improve the accuracy. The traning data mentioned has Proto Buffer format which can be accessed using Java class.

I need the data in format like OntoNotes train/test data where tokens are labeled individually. I can take care of BIO format. Can you please provide the link to download the data. This will help.

We _ _ O
respectfully _ _ O
invite _ _ O
you _ _ O
to _ _ O
watch _ _ O
a _ _ O
special _ _ O
edition _ _ O
of _ _ O
Across _ _ B-ORGANIZATION
China _ _ I-ORGANIZATION

. _ _ O

Thank you.

After creating a model I am getting error

Download link
With train.tar.gz data I created model which I am unable to load due to this error.

java.util.zip.ZipException: Not in GZIP format
at java.util.zip.GZIPInputStream.readHeader(GZIPInputStream.java:164)
at java.util.zip.GZIPInputStream.(GZIPInputStream.java:78)
at java.util.zip.GZIPInputStream.(GZIPInputStream.java:90)
at edu.washington.cs.figer.util.Serializer.deserialize(Serializer.java:40)
at edu.washington.cs.figer.ml.LogisticRegression.readModel(LogisticRegression.java:57)
at edu.washington.cs.figer.FigerSystem.(FigerSystem.java:87)
at edu.washington.cs.figer.FigerSystem.instance(FigerSystem.java:64)
at edu.washington.cs.figer.web.WebDemoServlet.doGet(WebDemoServlet.java:56)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:687)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:790)
at org.eclipse.jetty.servlet.ServletHolder.handle(ServletHolder.java:751)
at org.eclipse.jetty.servlet.ServletHandler.doHandle(ServletHandler.java:566)
at org.eclipse.jetty.server.handler.ScopedHandler.handle(ScopedHandler.java:143)
at org.eclipse.jetty.security.SecurityHandler.handle(SecurityHandler.java:578)
at org.eclipse.jetty.server.session.SessionHandler.doHandle(SessionHandler.java:221)
at org.eclipse.jetty.server.handler.ContextHandler.doHandle(ContextHandler.java:1111)
at org.eclipse.jetty.servlet.ServletHandler.doScope(ServletHandler.java:498)
at org.eclipse.jetty.server.session.SessionHandler.doScope(SessionHandler.java:183)
at org.eclipse.jetty.server.handler.ContextHandler.doScope(ContextHandler.java:1045)
at org.eclipse.jetty.server.handler.ScopedHandler.handle(ScopedHandler.java:141)
at org.eclipse.jetty.server.handler.ContextHandlerCollection.handle(ContextHandlerCollection.java:199)
at org.eclipse.jetty.server.handler.HandlerCollection.handle(HandlerCollection.java:109)
at org.eclipse.jetty.server.handler.HandlerWrapper.handle(HandlerWrapper.java:98)
at org.eclipse.jetty.server.Server.handle(Server.java:461)
at org.eclipse.jetty.server.HttpChannel.handle(HttpChannel.java:284)
at org.eclipse.jetty.server.HttpConnection.onFillable(HttpConnection.java:244)
at org.eclipse.jetty.io.AbstractConnection$2.run(AbstractConnection.java:534)
at org.eclipse.jetty.util.thread.QueuedThreadPool.runJob(QueuedThreadPool.java:607)
at org.eclipse.jetty.util.thread.QueuedThreadPool$3.run(QueuedThreadPool.java:536)
at java.lang.Thread.run(Thread.java:745)
2015-11-17 15:24:40.642:WARN:oejs.ServletHandler:qtp200539426-16: /figer
java.lang.NullPointerException
at edu.washington.cs.figer.ml.LogisticRegression.readModel(LogisticRegression.java:58)
at edu.washington.cs.figer.FigerSystem.(FigerSystem.java:87)
at edu.washington.cs.figer.FigerSystem.instance(FigerSystem.java:64)
at edu.washington.cs.figer.web.WebDemoServlet.doGet(WebDemoServlet.java:56)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:687)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:790)
at org.eclipse.jetty.servlet.ServletHolder.handle(ServletHolder.java:751)
at org.eclipse.jetty.servlet.ServletHandler.doHandle(ServletHandler.java:566)
at org.eclipse.jetty.server.handler.ScopedHandler.handle(ScopedHandler.java:143)
at org.eclipse.jetty.security.SecurityHandler.handle(SecurityHandler.java:578)
at org.eclipse.jetty.server.session.SessionHandler.doHandle(SessionHandler.java:221)
at org.eclipse.jetty.server.handler.ContextHandler.doHandle(ContextHandler.java:1111)
at org.eclipse.jetty.servlet.ServletHandler.doScope(ServletHandler.java:498)
at org.eclipse.jetty.server.session.SessionHandler.doScope(SessionHandler.java:183)
at org.eclipse.jetty.server.handler.ContextHandler.doScope(ContextHandler.java:1045)
at org.eclipse.jetty.server.handler.ScopedHandler.handle(ScopedHandler.java:141)
at org.eclipse.jetty.server.handler.ContextHandlerCollection.handle(ContextHandlerCollection.java:199)
at org.eclipse.jetty.server.handler.HandlerCollection.handle(HandlerCollection.java:109)
at org.eclipse.jetty.server.handler.HandlerWrapper.handle(HandlerWrapper.java:98)
at org.eclipse.jetty.server.Server.handle(Server.java:461)
at org.eclipse.jetty.server.HttpChannel.handle(HttpChannel.java:284)
at org.eclipse.jetty.server.HttpConnection.onFillable(HttpConnection.java:244)
at org.eclipse.jetty.io.AbstractConnection$2.run(AbstractConnection.java:534)
at org.eclipse.jetty.util.thread.QueuedThreadPool.runJob(QueuedThreadPool.java:607)
at org.eclipse.jetty.util.thread.QueuedThreadPool$3.run(QueuedThreadPool.java:536)
at java.lang.Thread.run(Thread.java:745)

tagged phrase in training data not tagged by model

I can see Tame Impala in the training data, tagged:

Tame ImpalaB^M/music/artistB^T/music/musical_groupB^]/internet/social_network_userB^M/common/topicH

Also in types.map 2 of the tags are mapped:

/music/artist /person/artist
/music/musical_group /person/musician

But Tame Impala is not tagged by the model. Is there a reason why?

Research Publication GOLDEN Set

Is the file exp.label containing the 434 sentences of the golden set used for the evaluation in the research paper?

What is the file exp.out containing? The predictions of the FIGER model? If yes, why the sentences are a lot less than 434?

Regarding retraining figer

Hi xiaoling,

I want to re-train Figer with the addition of some data to existing training data but training data available in git hub is serialized using google protobuf. Can you please guide me to modify training data, so that I can add some training data and corresponding label in types.map of it?

Training FIGER on new data

Hi,

I am able to train your system on new training data and able to run it on new test data. But in test output some confidence scores are negative. How to interpret it? Should it be considered as valid classification with negative score or it should be ignored as 'O'?

Regards
Tapas

Is 'aaai/exp.txt file' the test data?

Thanks.