Menu

how can i get ranked list RankLib-v2.1

RankLib
soroush
2016-06-13
2016-06-13
  • soroush

    soroush - 2016-06-13

    hi guys

    I'm trying to do learning to rank with RankLib-v2.1
    i want to use AdaRabk Algorithm for Expedia (Rank hotels on Expedia.com to maximize purchases) data set.
    now i need a list of hotel as output.
    what commands should I use?
    i use this for modeled my ranking:
    Java -jar RankLib.jar -train train.dat -ranker 3 -metric2t NDCG@10 -save model.dat
    and this for test it:
    Java -jar RankLib.jar -load model.dat -test test.dat -rank rank.dat -metric2T NDCG@10 -score score.dat

    thank you

     
  • Lemur Project

    Lemur Project - 2016-06-13

    The current version of RankLib is 2.6 with version 2.7 due to be released in another week or so. You should consider using a more recent version.

    The -rank argument was used to produce a rankings list based on the scores produced by the -score argument.

    However, at some point, the -rank argument was removed, so now one should manually rank the documents by hand using the -score output (or obviously write a script to do so).

    I don't know if version 2.1 still has that argument active or not.

     
  • soroush

    soroush - 2016-06-13

    thank you so much for answer my question
    this version(2.1) also have -rank and -score
    sorry but I got confused
    how do i use -score in command line (what is the rank file? how can i produce it?)
    like this:
    Java -jar RankLib.jar -rank train.dat -ranker 3 -metric2t NDCG@10 -score score.dat (whit this i have a score.dat file that have a list of score corresponding to each record of train file!)

     
  • Lemur Project

    Lemur Project - 2016-06-13

    The score arg says score the results from my input data file. The rank file is the input data you want ranked.

    java -jar RankLib.jar -rank data/myTestData.dat -load myAdaRankModel.txt -score scores.out

    The scores.out file will look something like the following:

    1 0 0.004216452386274353
    1 1 0.004257460583300749
    1 2 0.004201425973661125
    1 3 0.00442558527711901
    1 4 0.005759673209865005
    1 5 0.004998650191299157
    ...

    The first column is the query ID but the second column is just a count iterator over the documents returned for the query and not a true rank. You'll need to sort the scores within queries to get the ranked lists.

     
  • soroush

    soroush - 2016-06-13

    Thank you for the answer
    i appreciate that

     

Log in to post a comment.

MongoDB Logo MongoDB