Yes, you could do that with both the binary or the source codes. If you want to use the source codes, see the Evaluator.java class. It contains all the APIs you need (e.g. training/loading a model, specifying a measure, etc.).
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear sir,
thank you for your kind and quick reply :)
how can i make k-fold cross validation by providing training, testing and validation data.
yours sincerely,
pwint
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The current released version of RankLib is 2.5. I'd recommend you use that version.
Evaluator.java is the main code for accessing all the ranking algorithms. It resides in the RankLib src directory under ./ciir/umass/edu/eval/Evaluator.java.
SMH
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Thank you for your reply.
Do I have to install Indri 5.9 to be able to use the source code of RankLib 2.5. Please direct me to the link to install it fully (i.e. not only the RankLib 2.5 jar file).
Also, is there an API that I consult for using RankLib java classes?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear sir, could you help me? Can you publish the running results of some algorithms (For example, Rankboost, AdaRank and LambdaMART) in some datasets (For example, MQ2008 and OHSUMED) on web site? And it is more convenient to verify and compare my results with the online results.
Yours sincerely!
LEEZHONG
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Do I have to install Indri 5.9 to be able to use the source code of RankLib 2.5. Please direct me to
the link to install it fully (i.e. not only the RankLib 2.5 jar file).
The Lemur project is divided into three main groupings: Indri (a C++ based API/application set), Galago (a Java based API/application set that is supposed to supplant Indri) and RankLib, which is a stand alone set of ranking algorithms written in Java. They are independent of one another.
You might choose to use Indri or Galago to come up with features to use with RankLib, but they aren't required for using RankLib.
We are attempting to move on from Indri. No further development is going to be put into Indri. Galago is now the applicaton/API of choice for search engine processing.
As such, if you wish to build indexes or perform searches, I would move on to Galago. Indri is still very solid and widely used. If you wish to build Indri indexes, feel free, but galago and Indri are not interchangeable, and largely irrelevant as far as RankLib use is concerned.
Also, is there an API that I consult for using RankLib java classes?
I'm working at producing additional documentation for galago and RankLib. If you are looking to integrate RankLib functionality (its various learning to rank algorithms), I'd use Evaluator.java as a starting point.
If learning to rank is what you are interested in (I presume this must be the case if using RankLib), look at the sources in ./ciir/umass/edu/learning/ for the classes that represent various learning to rank algorithms.
SMH
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear sir,
I would like to know can I use ranklib in eclipse and how can I make k-fold
cross validation using the folds provided by Letor.
Yours sincerely,
pwint
Yes, you could do that with both the binary or the source codes. If you want to use the source codes, see the Evaluator.java class. It contains all the APIs you need (e.g. training/loading a model, specifying a measure, etc.).
Dear sir,
thank you for your kind and quick reply :)
how can i make k-fold cross validation by providing training, testing and validation data.
yours sincerely,
pwint
Hello,
Could you elaborate from where to get the Evaluator.java file. I'm interested in using RankLib in Java. I have downloaded the RankLib-2.1-patched.jar file from this link https://sourceforge.net/projects/lemur/files/lemur/RankLib-2.1/
and I have executed this jar file from the command line as instructed in this link https://sourceforge.net/p/lemur/wiki/RankLib%20How%20to%20use/
Now I need to incorporate RankLib in my Java code. Would appreciate your help and guidance.
The current released version of RankLib is 2.5. I'd recommend you use that version.
Evaluator.java is the main code for accessing all the ranking algorithms. It resides in the RankLib src directory under ./ciir/umass/edu/eval/Evaluator.java.
SMH
Thank you for your reply.
Do I have to install Indri 5.9 to be able to use the source code of RankLib 2.5. Please direct me to the link to install it fully (i.e. not only the RankLib 2.5 jar file).
Also, is there an API that I consult for using RankLib java classes?
Instructions can be found in the RankLib wiki: https://sourceforge.net/p/lemur/wiki/RankLib%20How%20to%20use#kcv
Last edit: Van Dang 2013-08-20
thank you so much sir :)
Dear sir, could you help me? Can you publish the running results of some algorithms (For example, Rankboost, AdaRank and LambdaMART) in some datasets (For example, MQ2008 and OHSUMED) on web site? And it is more convenient to verify and compare my results with the online results.
Yours sincerely!
LEEZHONG
Same Leezhong's question! Could you public result in MQ2008 dataset, because i'm have not experience in machine learning.
Thank you!
The Lemur project is divided into three main groupings: Indri (a C++ based API/application set), Galago (a Java based API/application set that is supposed to supplant Indri) and RankLib, which is a stand alone set of ranking algorithms written in Java. They are independent of one another.
You might choose to use Indri or Galago to come up with features to use with RankLib, but they aren't required for using RankLib.
We are attempting to move on from Indri. No further development is going to be put into Indri. Galago is now the applicaton/API of choice for search engine processing.
As such, if you wish to build indexes or perform searches, I would move on to Galago. Indri is still very solid and widely used. If you wish to build Indri indexes, feel free, but galago and Indri are not interchangeable, and largely irrelevant as far as RankLib use is concerned.
I'm working at producing additional documentation for galago and RankLib. If you are looking to integrate RankLib functionality (its various learning to rank algorithms), I'd use Evaluator.java as a starting point.
If learning to rank is what you are interested in (I presume this must be the case if using RankLib), look at the sources in ./ciir/umass/edu/learning/ for the classes that represent various learning to rank algorithms.
SMH