Comparison to Existing Programs

The use of bigrams to understand and analyze different parts of a given language has been studied and implemented substantially. For example, there are programs that calculate bigram frequencies to evaluate a language's morphology. However, unlike our program, none to date have utilized the differences in bigram frequencies between two languages to distinguish one language from the next.

The system in the program that was used for determining the parts of speech in a sentence has rarely been attempted, and when it has been used, only partially and in conjunction with other methods. Most natural language processing programs have been designed to be as the efficient and effective as possible. As a result, many use large banks of initial data, which the program then analyzes and uses for subsequent input. As discussed previously, the most common and successful programs of this sort are statistical parsers. On the contrary, our program uses the recursive principle to acquire new vocabulary and forms of syntax for a given language, provided only a very small initial set of data. In practice, our model only required one sentence with verb and noun indicated to determine the parts of speech of all other words, although not denoting them in linguistic terms (article, preposition, etc.), as well as learn new words and, in theory, to learn new sentence patterns. Despite the extensive power of the recursive method, it has rarely been used in the history of natural language processing. The results of our program illustrate the potential abilities of the recursive method that have not been seen previously.