Some names has first name, middle name, and last name. The last name is a family name, except some country like Chine. Many literatures for simplicity only show the last name with abbreviation to the first and middle ones. For example Rahmadya Trias Handayanto will be written as R. T. Handayanto or Handayanto, R. T. (see e.g in IEEE: http://ieeexplore.ieee.org/search/searchresult.jsp?newsearch=true&queryText=handayanto) so we want to predict that R. T. Handayanto is similar to Rahmadya Trias Handayanto. For experiment try to type in Matlab:

As usual, I use ‘{‘ instead of ‘(‘ because I made a cell type that fit the record on database. We want to collect other name that has similar to it like r. t. handayanto. For collecting the last name you can see my other post using regular expression: https://rahmadya.com/2014/04/18/how-to-search-last-name-using-matlab/. Therefore, we just add the first and middle, or only the last name if only a single name.

The logic is we find first the index of space that indicate the first or middle name with last name. After we count the total number of name we represent only first character with point for first name and middle name. I use if-else to detect whether the name only has one name (last name).

Try using a command window to convert the name become short name. You see that now rahmadya trias handayanto became r. t. handayanto that will compare to other that similar to it using format character (see the bottom line).

For example, we have 10 data:

You can see that Ahmed Abdul-hamid and A. Abdul-hamid that has similar name, but A. Cain may be different from A. J. Cain.

The result state that there is two names with similarity of last name and abbreviation of firs name that id=3 and id=10. With function lower we change all author name into lower case in order to easily compare. Of course, to get this code we have to understand how to manipulate matrices in Matlab.

It is a hard day, because I always sleep over the mid night. Ok, it’s about text mining. We want search from the data that has a particular family name, for example, may name, rahmadya trias handayanto. Ok, because our data usually cell type (for database), use this kind of data type:

name={‘rahmadya trias handayanto’}

We use symbol { for creating a cell type that we always find in table format. The logic for searching my lastname is by finding the last space. Try using the regular expression.

index=regexp(name,’ ‘)

Don’t forget to type space between ‘ because we want to search the index of space for name. Type celldisp(index) to see the content of index, because it is cell type. Many people have more than two names, with middle name. So we have to choose the biggest index using function max.

bigindex=max(index{1})

We have to convert the name into character if we want to present the last name using char function.

charname=char(name);

key=charname(bigindex+1:end)

We use ‘+1’ because we don’t include space to our last name. You can see the instruction below, it is very easy.

Untuk memproses 2000-an record terhadap 2 juta record jika dijalankan dengan satu proses akan memakan waktu cukup lama karena karakternya yang serial. Untuk menjalankan proses yang parallel secara bersama maka kita dapat menjalankan lebih dari satu matlab dalam satu computer. Biasanya ini dilakukan jika kita ingin memproses data berukuran besar (big data).

First, you have to change current directory of your first matlab. If we want to create with two machines, you can open your Matlab again form start menu or from your shortcut wall. For example I want to use three Matlab for computing my algorithm at the same time for faster result.

Second, things that we have to consider is the location of the result, if you have the similar name to the output, you have to separate current directory because of the conflict of output (the last output will replace the previous one).

Last, you have to count the power of your machine. I use i5 processor, and sometimes I use four processing. The picture below show my Matlab that process 3 text processing of 2 thousand of author to other 2 million author.

Kalau pada tulisan yang dulu (https://rahmadya.com/2014/04/07/mencari-kata-dalam-text-dengan-matlab/) ketika mencari kata kita terlebih dahulu menentukan keyword (atau diistilahkan dengan term), untuk mencari nama yang sama mau tidak mau Matlab harus menentukan sendiri term yang akan dicari kesamaan dengan kata pada record yang lain. Sebagai contoh di sini saya menentukan 13 nama yang memiliki 6 orang dengan nama yang sama.

Cara kerjanya adalah pertama-tama Matlab mengambil kata kunci record pertama yaitu “ayman kaheel” sebagai term yang akan dibandingkan dengan record kedua, ketiga, dan seterusnya hingga panjang maksimal record (dalam hal ini 13 record). Jika ditemukan nama yang sama, dengan instruksi isempty, baik record term maupun yang mirip dengannya dipisahkan dalam satu tabel result. Ceritanya panjang, lebih baik langsung saya tampilkan saja kode program:

Seperti biasa, fungsi strfind bermaksud membandingkan suatu string pada vektor. Cellstr berfungsi merubah string menjadi cell karena kalau tidak dirubah, tidak bisa digabungkan menjadi satu matriks karena string berbeda-beda ukurannya, sedangkan cell selalu tetap, berapapun panjang string, tetap dianggap satu sel. Pindahkan script itu ke command window, anda akan memperoleh hasil berikut ini:

Ketika hasil diketik (result) akan memunculkan deretan nama yang sama dengan no urutnya. Berikutnya adalah bagaimana cara untuk mencari nama yang nama keluarganya sama, karena terkadang dalam literatur ilmiah, misalnya ayman kaheel, terkadang ditulis a. kaheel, sehingga jika digunakan cara di atas akan tidak berhasil. Selamat mencoba !