Precision and Recall Lucene Java

[Gianluca Di Vincenzo]

Hi, I'm a new developer of Lucene inexperienced , I turn to you after a series of failed attempts .

My problem is to create a simple Java program that calculates the precision recall of a series of documents .

Of file system I have a collection of documents that I index using the command prompt with the command:

java -cp C:/lucene/jar/lucene-core-4.3.1.jar;C:/lucene/jar/lucene-analyzers-common-4.3.1.jar;C:/lucene/jar/lucene-demo-4.3.1.jar org.apache.lucene.demo.IndexFiles -index C:/lucene/indice -docs C:/lucene/confusion_track/original

Here's an example of one of the topics in the file C:/lucene/topics.txt :

<top>

<num> Number: CF11

<title> apache

<desc> Description:

I am looking for a document about the dismissal of a lawsuit Involving

Adventist Health Systems.

<narr> Narrative:

</ top>

Also place the Java code , in order to help you easily identify my error :

 public static void main(String[] args) throws IOException {

        File topicsFile = new File("C:/lucene/topics.txt");

        File qrelsFile = new File("C:/lucene/confusion.known_items");

        Directory fsDir = FSDirectory.open(new File("C:/lucene/indice"));

        IndexSearcher searcher = new IndexSearcher(IndexReader.open(fsDir));

        String docNameField = "Federal";

        PrintWriter logger = new PrintWriter(System.out, true);

        TrecTopicsReader qReader = new TrecTopicsReader();//#1

        QualityQuery qqs[] = qReader.readQueries(new BufferedReader(new FileReader(topicsFile)));

        Judge judge = new TrecJudge(new BufferedReader(new FileReader(qrelsFile)));//#2

        judge.validateData(qqs, logger);//#3

        QualityQueryParser qqParser = new SimpleQQParser("title", "description");//#4

        QualityBenchmark qrun = new QualityBenchmark(qqs, qqParser, searcher, docNameField);

        SubmissionReport submitLog = null;

        QualityStats[] stats = qrun.execute(judge, submitLog, logger);//#5

        QualityStats avg = QualityStats.average(stats);//#6

        avg.log("SUMMARY", 2, logger, "  ");

        fsDir.close();

    }

The result obtained is the following:

CF11  -  description:apache

CF11 Stats:

  Search Seconds:         0.014

  DocName Seconds:        0.000

  Num Points:             0.000

  Num Good Points:        0.000

  Max Good Points:        1.000

  Average Precision:      0.000

  MRR:                    0.000

  Recall:                 0.000

CF12  -  description:apache

CF12 Stats:

  Search Seconds:         0.000

  DocName Seconds:        0.000

  Num Points:             0.000

  Num Good Points:        0.000

  Max Good Points:        1.000

  Average Precision:      0.000

  MRR:                    0.000

  Recall:                 0.000

CF13  -  description:apache

CF13 Stats:

  Search Seconds:         0.001

  DocName Seconds:        0.000

  Num Points:             0.000

  Num Good Points:        0.000

  Max Good Points:        1.000

  Average Precision:      0.000

  MRR:                    0.000

  Recall:                 0.000

SUMMARY

Search Seconds:         0.003

DocName Seconds:        0.000

Num Points:             0.000

Num Good Points:        0.000

Max Good Points:        1.000

Average Precision:      0.000

MRR:                    0.000

Recall:                 0.000

Thank you all for the availability and the immense cooperation. 

[Fabio Grucci]

Dear Gianluca,

this group was born to answer on specific LAE questions/problems. I think that your problem should be posted in other places than this, above all to help growing the right open-source community. However there are many resources in the web that explains how to do that, for example StackOverflow Precision recall in lucene java otherwise you can try this or this.

Thanks for making me discover the org.apache.lucene.benchmark package :-)