2

I am attempting to use Duke to match records from one csv to another.First csv and second both has ID,Model,Price,CompanyName,Review,Url columns. I am trying to match to another csv to find duplicates records.

package no.priv.garshol.duke;

import no.priv.garshol.duke.matchers.PrintMatchListener;

public class RunDuke {

  public static void main(String[] argv) throws Exception {
    Configuration config =
        ConfigLoader
            .load("/home/kishore/Duke-master/doc/example-data/presonalCare.xml");
    Processor proc = new Processor(config);
    proc.addMatchListener(new PrintMatchListener(true, true, true, false, config.getProperties(),
        true));
    proc.link();
    proc.close();
  }

}

Here is an example of personalCare.xml:

<!-- language: xml -->
<!-- For more information, see https://github.com/larsga/Duke/wiki/ Improvements 
    needed: - some area numbers have spaces in them - not stripping accents from 
    names -->
<duke>
    <schema>
        <threshold>0.7</threshold>
        <property type="id">
            <name>ID</name>
        </property>
        <property>
            <name>Model</name>
            <comparator>no.priv.garshol.duke.comparators.Levenshtein</comparator>
            <low>0.4</low>
            <high>0.8</high>
        </property>
        <property>
            <name>Price</name>
            <comparator>no.priv.garshol.duke.comparators.ExactComparator</comparator>
            <low>0.04</low>
            <high>0.73</high>
        </property>
        <property>
            <name>CompanyName</name>
            <comparator>no.priv.garshol.duke.comparators.Levenshtein</comparator>
            <low>0.4</low>
            <high>0.8</high>
        </property>
        <property>
            <name>Review</name>
            <comparator>no.priv.garshol.duke.comparators.Levenshtein</comparator>
            <low>0.12</low>
            <high>0.93</high>
        </property>
        <property>
            <name>Url</name>
            <comparator>no.priv.garshol.duke.comparators.Levenshtein</comparator>
            <low>0.12</low>
            <high>0.93</high>
        </property>
    </schema>

    <database class="no.priv.garshol.duke.databases.InMemoryDatabase">
    </database>

    <group>
        <csv>
            <param name="input-file" value="personal_care_11.csv" />
            <param name="header-line" value="false" />
            <column name="1" property="ID" />
            <column name="2" property="Model" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
            <column name="3" property="Price" />
            <column name="4" property="CompanyName" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
            <column name="5" property="Review" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
            <column name="6" property="Url" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
        </csv>
    </group>

    <group>
        <csv>
            <param name="input-file" value="personal_care_11.csv" />
            <param name="header-line" value="false" />
            <column name="1" property="ID" />
            <column name="2" property="Model" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
            <column name="3" property="Price" />
            <column name="4" property="CompanyName" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
            <column name="5" property="Review" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
            <column name="6" property="Url" cleaner="no.priv.garshol.duke.cleaners.LowerCaseNormalizeCleaner"/>
        </csv>
    </group>


</duke>

The above code is working fine but it does not match the exact record example

STHDRNFKAQ4AFYE8,Littmann 3M Classic II S.E Acoustic
Stethoscope,6297,Littmann,,http://dl.flipkart.com/dl/littmann-3m-classic-ii-s-e-acoustic-stethoscope/p/itme3uhzbqxhzfda?pid=STHDRNFKAQFAFYE8&affid=3ba0de4902524e2b90e43b84b89ea0ef

which is in both csv files. I also want to know the work of low and high property value which is given in .xml file, how to decide the low and high value for column value.

Paolo Forgia
  • 6,572
  • 8
  • 46
  • 58
Kishore
  • 5,761
  • 5
  • 28
  • 53

1 Answers1

3

You are doing record linkage (two data sets) and not deduplication (single data set), so take out the .deduplicate() call.

Also, please don't use the 'no.priv.garshol.duke' package name. You should never use domain names you don't own yourself.

Anyway, the reason you can't find any matches is that the two records have the same ID. Duke verifies that it's not reporting records as matching themselves, and so the match gets filtered out. If you make a copy of the csv file and use that for group 2, then make a change to the ID then Duke finds the duplicate.

Here's what happens when I try that:

[lars.garshol@laptop tmp]$ java -cp ~/cvs-co/duke/duke-core/target/duke-core-1.3-SNAPSHOT.jar:. no.priv.garshol.duke.Duke --showmatches presonalCare.xml 

MATCH 0.9982630751840313
ID: 'SHDRNFKAQ4AFYE8', Model: 'littmann 3m classic ii s.e acoustic stethoscope', Price: '6297', CompanyName: 'littmann', Url: 'http://dl.flipkart.com/dl/littmann-3m-classic-ii-s-e-acoustic-stethoscope/p/itme3uhzbqxhzfda?pid=sthdrnfkaqfafye8&affid=3ba0de4902524e2b90e43b84b89ea0ef', 
ID: 'STHDRNFKAQ4AFYE8', Model: 'littmann 3m classic ii s.e acoustic stethoscope', Price: '6297', CompanyName: 'littmann', Url: 'http://dl.flipkart.com/dl/littmann-3m-classic-ii-s-e-acoustic-stethoscope/p/itme3uhzbqxhzfda?pid=sthdrnfkaqfafye8&affid=3ba0de4902524e2b90e43b84b89ea0ef',
larsga
  • 654
  • 6
  • 10
  • Thank you for reply. I want to know what is the meaning of low and high probability in property for all column. How can I decide. When I give high parameter as a less value like 0.5 it give more accurate result than 0.8. I an getting this. please help me. – Kishore Oct 08 '15 at 09:42
  • Low probability = probability when values don't match. High = probability when they do match. – larsga Oct 08 '15 at 16:02