Lemmatizer: Akıllı Türkçe Kök Bulma Yöntemi

OpenAlex Türkçe

Recently, various studies have been conducted in the field of Turkish natural language processing.These studies require a smart system that has the capability of answering Turkish questions, translating articles into another language, summarizing the articles, and sending automatic replies to e-mails.The need to find the correct roots of Turkish words is the basis of the aforementioned capabilities.Although various methods of finding Turkish roots have been given in the literature, success rates are generally low due to the complex structures of Turkish words.In this study, a root finding system (Lemmatizer) has been developed using the agglutinating structure of Turkish words and its comparison with the Zemberek and Snowball methods is given.The Lemmatizer system is written in Python and is based on more than 130 most frequently used suffixes and TDK dictionaries in Turkish.In addition, statistical analysis was performed using the Knime platform.For this study, firstly the Lemmatizer system was trained with many Turkish articles and books and it has continuously improved its repertoire.At the same time, thanks to the feedback received using the Turkish suffix and root database called Kalbur, the accuracy rate has increased continuously.The results of the Lemmatizer system were compared with the Zemberek and Snowball methods previously made in terms of both number and accuracy.Turkish texts of different lengths were used for comparison.It has been shown that the Lemmatizer method gives results close to the Snowball and Zembere methods, and the success rate increases with each new text, thanks to its ability to learn using the TDK dictionary.