restlesscoder.net

Name Generator

Language: Java Web Start

License: public domain

Launch

Source code

This name generator outputs random names using an algorithm based on Markov chains. It works by generating tokens of a fixed length based on an input dictionary—in our case, a list of names.

This Java Web Start application asks for full system privileges, but the only reason it needs them is so that you can specify a custom dictionary text file from your file system, rather than merely using the built-in male, female and surname presets. If granting this sort of access bothers you, feel free to use the Flex version instead.

I am not special. This program has been written many times in many languages for many purposes. Here are a few other random text generators that use Markov chains:

• Random Name Generator with the same input data as mine.
• Yet Another Fantasy NAme Generator, accessible via SOAP and XML-RPC.
• Peter's Random Name Generator, with Python source code.
• Markov Chain Random Name Generator project at freshmeat, with C source code.

I didn't find any other versions in Java or Flex, so maybe my implementations are not wholly redundant, but I didn't look very hard.

There are many other name generators out there that work on other principles, as well. These two just pick a random name from a dictionary. This one is template-driven. And here
are a few whose inner workings are not immediately evident.

Here is a detailed description of the algorithm:

1. First, we choose the token size. A good length for words that tend toward uniqueness while still sounding like names is 3. A choice of 2 will result in more "deviant" names, while 4 will be more similar (often identical) to actual names in the dictionary.
2. For each name in the dictionary, we add all three-letter substrings to our tokens list. For example, for the name "Curtis" we would add tokens "cur," "urt," "rti," and "tis." We also prepend a "beginning of name" character "^" and append an "end of name" character "$," meaning that "^cu" and "is$" also go into our list.

Once we have processed the entire input dictionary, we have a nice list of "name-like" substrings to use as the basis for generating our own names:

1. Pick a random token that begins with the "beginning of name" character "^," saving the characters to our string buffer.
2. Pick a random token beginning with the last two characters in our string buffer, appending its last character to the string buffer.
3. Repeat step #2 until we pick a token ending in the "end of name" character "$."

Note that the algorithm can generate names of arbitrary length, so if you are reimplementing the algorithm yourself, you may wish to impose a maximum length cut-off, or even discard results that get too long and start over.