GATE: Classification and clustering of text for semi-vowel/j/-morphophonemic approach

In recent years, many successful machine learning applications have been developed. Classification & Clustering is one such. This application is cross-disciplinary, now that it is based on data mining algorithms on the technical side and on graphemes and morphophonemic on the linguistic side. It will thus map the correspondence between grapheme 〈y〉 and related phonemes via morphemes in a given context. The grapheme 〈y〉 is often realized as an approximant /j/ (a consonant phoneme), as in 〈yacht〉 or as vowel / i / as in 〈racy〉, or a diphthong / ai / as in 〈sky〉, etc. The objective, that is to say, of this text analysis is to map the various articulators//phonemic realizations of the grapheme 〈y〉. This experiment will thus help the study the occurrence of 〈y〉 in different positions in words, word initially, finally or elsewhere, as in these examples. The training data (or corpus) chosen for this experimentation is a set of English literary texts [Harry Potter part 1, part 2, The complete works of William Shakespeare By William Shakespeare, The Adventures of Sherlock Holmes By Arthur Conan Doyle, (A Christmas carol) By Charles Dickens's. As for the tools, the alphabet recognizer is used for retrieving information and categorizing the data as noun, adjective, etc. Then, GATE Developer, the Text engineering tool is used to analyze the dataset and to derive the output with statistical data of global distribution 〈y〉 as in all the input documents. This project is particularly relevant because it offers, as will be demonstrated, a solution to problems due to lack of one-to-one correspondence between spelling and pronunciation in English, as for instance, in the context of language pedagogy. Glossary of technical terms (of both linguistics & computing) is appended, now that the paper is inter-disciplinary.