Abstract:

TF-IDF term weighting measurement is based on the knowledge of the frequency of words passing in texts. This measurement does not contain the meaningful relationship between the words. Based on artificial nerve networks, the Doc2Vec method hosts the meaningful relationship between words and words-containing documents and ensures the achievement of managed-dimensional document vectors. The positive effects of followed word group detection on text mining have been pointed out by many studies presented in literature. Identification of the consistent word group is important for ensuring the meaningful integrity within the document. This study compared the success values of both the traditional TF-IDF term weighting measurement and the basic machine learning classificers and community learning algorithms on the documents vectored by the use of the Doc2Vec method, a method based on YSA. The main classifiers in our study were Naive Bayes, K-Last Neighborhood, Logistics Regression, Decision Support Machines, Decision Tree, Multi-Last Detectors and Community Learning Methods using Rassal Forest, Torbalama and Adaboost algorithms. Also the last three most successful classification algorithms combined with the majority vote and the results obtained were shared. Classifiers have been applied on 4 different Turkish datasets containing news documents in different lengths. Before the contribution of our work to literature was passed to the classification phase, the realization of the sequential word group in the documents and after the document was vectored by the treatment of these word subjects as a single word, it was shown that the success values of the classifiers applied increased. In addition to a principle that is based on the principle of the frequency of passing together of words for the detection of the follow-up word group, the word links of the Turkish Wikipedia have also been used and the detection of the follow-up word subjects, which are meaningful despite the small number of passing in the documents, has been made possible. With consistent word group detection, higher success values have been achieved in almost all of the classification trials.

Keywords:

Abstract:

TF-IDF term weighting measure is based on frequency of words in texts. This measure doesn’t capture the semantic relationship between words. Doc2Vec which is based on artificial neural networks can capture the semantic relations between the words and it enables to yield document vectors of a more manageable size. Consecutive word detection has been reported to have important effects on text mining by many studies. Consecutive word phrases are important for expressing the semantic integrity within the texts. In this study, the performances of traditional machine learning classifiers and ensemble learning algorithms are compared on four different Turkish datasets which are vectorized with both traditional TF-IDF term weighting measurement and Doc2Vec method. The classifiers have been applied on 4 different Turkish datasets containing news documents of different lengths. The contributions of our study are “to apply consecutive word detection process to the documents before the classification phase” and “to show that the performances of the applied classifiers’ results have been increased after the consecutive word detection phase is applied”. In addition to the approach based on frequency of words for consecutive word detection, we also use the url links of Turkish Wikipedia. By using consecutive word detection, higher performance values are presented in almost all classification experiments.