Çavusoglu, I.I.ÇavusogluPielka, MarenMarenPielkaSifa, RafetRafetSifa2022-03-152022-03-152020https://publica.fraunhofer.de/handle/publica/41197210.1109/DSAA49011.2020.00100Natural Language Processing, and specifically Sentiment Analysis are still unexplored topics with respect to Turkish text. A key challenge is to extract meaningful word and paragraph representations. We provide a comprehensive overview on pre-processing and featurization methods for this problem. Our focus is on the inherent difficulties that come with analyzing Turkish real-world data from the e-commerce domain, such as inconsistent spelling or complicated morphological and grammatical structures.en005006629conference paper