Purpose

This package is used for evaluating and recording the performance of back-transliteratoin by using global edit distance and soundex with different parameters.

Content

The package contains following files:

• ApproximateMatching.jar for running
• names.txt and train.txt as data resources
• searchPatterns.py used for preprocessing
• percentage_matrix.json generated by searchPatterns.py and used as replacement matrix
• Uncompiled Java Class contained with three java packaged in ApproximateMatching.jar

Command

You may run the package by typing below command in bash, shell or equivalent:

java -jar .\ApproximateMatching.jar [choice parameter]

Command Parameters

The command contains choice parameters 1 to 8: 1 Evaluating Global Edit Distance with Levenshtein parameters 2 Evaluating Global Edit Distance with Replacement Matrix 3 Evaluating Global Edit Distance with different pairs of Additions and Deletions 4 Evaluating Global Edit Distance with different pairs of Additions and Deletions as wellas Replacement Matrix 5 Evaluate Soundex with traditional English table 6 Evaluate Soundex with adjusted table 7 Evaluate Soundex with further adjusted table 8 Evaluate all of the above

Example

If you just want to see the result of evaluating soundex, you may type:

java -jar .\ApproximateMatching.jar 5

Output

The output of each round of evaluation is a log file and a result file. The log file contains all the comparisons made between Persian and Latin names. The result files shows the number of total predictions and correct predictions.

Contact

Please send email to jiayu_fibby@hotmail.com if there is any problem with this program.