Development of information models with the hidden Markov model and Bayes-based classifiers in information systems

Along with the developments in information technologies, a large number of texts accumulated in the computer and Internet environments. The users are faced with difficulties in accessing the information they seek in the face of such dense accumulation of texts. Mostly, the users cannot find the time to go through such data. The existing information systems remain inadequate in providing information from such dense masses of data. The present study aims to develop a model that will automatically provide information from text content in the computer and Internet environments, thereby enabling automatic access to information from large-scale masses of texts in the computer environment. The study resulted in the development of a complex multiphase model. The model was prepared by taking into consideration the agglutinative structure of Turkish and through the development of an algorithm for the separation of words from suffixes. The Hidden Markov Model and Bayes-based classifiers, that constituted the theoretical structure of the model, were used to classify the words with respect to word types. Words and their associated data and fields were examined through the entity-relationship model and subsequently transferred to the database. This was followed by the extraction of information through the Structured Query Language (SQL). Once the model was applied to the content of a Turkish text of 584 words retrieved from the Internet environment, a success rate of 90.8% was achieved in the classification of words by type. In addition, the model enabled a body of meaningful and representative information of 13 words to be obtained automatically from the textual mass of 584 words. Developed with the aim of providing information from textual content in Turkish, the model was observed to offer successful results in classifying words by type and obtaining information automatically.