FAQ
Hello, taking a look now at our benchmarking package in java lucene, I
think its currently too difficult to use, mainly because for accurate
results the output must be sent thru trec_eval.

At the same time users have brought up things like producing graphs
and other similar improvements, and I'd like to look at making the
whole process simpler and more automated.

Can you provide more information on your trec_eval JNI wrapper?
What modifications to trec_eval's build system (makefile, etc) must be
done to support it?
I'm assuming it has to build a shared library instead of a static
executable, but maybe I am wrong...

--
Robert Muir
rcmuir@gmail.com

Search Discussions

  • Nicola Ferro at May 26, 2010 at 5:10 pm
    Dear Robert,

    I was very pragmatic: I kept the files:
    trvec_teval.c
    measures.c
    form_trvec.c
    buf_util.c
    trec_eval.h
    tr_vec.h
    sysfunc.h
    common.h
    buf.h

    unmodified as they come in the trec_eval distribution.

    Then I copied the part of trec_eval.c which does the computation (after input parameter parsing and before printing output) into another c file which contains also the code called by the java library.

    The file is compiled as a shared library.

    If you are interested, we can go into further details also with the code. Actually, I'm not working on it since a couple of years so I need to "refresh" the details a little bit.

    By the way, a couple of weeks ago, I talked with TREC people and they are planning for a next release of trec_eval, with some new metrics (mainly (n)DCG ones) and with some further changes.

    All the best,
    Nicola



    ---------------------------------------------------------------------------------
    Nicola Ferro - Ph.D. in Computer Science
    Assistant Professor

    Department of Information Engineering (DEI)
    University of Padua
    Via Gradenigo, 6/A - 35131 Padova - Italy
    Tel +39 049 827 7939 Fax: +39 049 827 7799

    skype: nicola.ferro
    e-mail: ferro@dei.unipd.it
    home page: http://ims.dei.unipd.it/members/ferro/
    ----------------------------------------------------------------------------------



    Il giorno 26/mag/2010, alle ore 14.51, Robert Muir ha scritto:
    Hello, taking a look now at our benchmarking package in java lucene, I
    think its currently too difficult to use, mainly because for accurate
    results the output must be sent thru trec_eval.

    At the same time users have brought up things like producing graphs
    and other similar improvements, and I'd like to look at making the
    whole process simpler and more automated.

    Can you provide more information on your trec_eval JNI wrapper?
    What modifications to trec_eval's build system (makefile, etc) must be
    done to support it?
    I'm assuming it has to build a shared library instead of a static
    executable, but maybe I am wrong...

    --
    Robert Muir
    rcmuir@gmail.com

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupopenrelevance-dev @
categorieslucene
postedMay 26, '10 at 12:52p
activeMay 26, '10 at 5:10p
posts2
users2
websitelucene.apache.org...

2 users in discussion

Robert Muir: 1 post Nicola Ferro: 1 post

People

Translate

site design / logo © 2018 Grokbase