Generation of phyton syntax of Turkish verbal expressions with machine translation

In this study, the goal is automatically to create the syntax term in the Python programming language of Turkish verbal expressions is targeted. The study aims to simplify the learning of Python code, easiness for software teachers in educational environments, as well as providing a remarkably swift to learn code for students who take the software course. A new approach has been taken to bring this about. First of all, sentences pronounced in Turkish are translated into text. The transcription is carried out with the help of the "speech recognition" library in Python. Secondly, the Turkish sentences translated into the text are translated into English using deep learning algorithms such as LSTM, BiLSTM and GRU. During machine translation, the Encoder-Decoder template, which includes the Seq2seq architecture, is utilized. The Encoder-Decoder model translates sentences in Turkish at the entrance into English at the exit. The English translated sentences are then transferred as input to the GPT-3 template. The GPT-3 template generates the Python code based on the English text sent to itself. Both the web-based and Windows-based applications have been generated using the libraries "Django" and "PyQt5" to allow the study to be used by end-users. The study results have appraised by BLEU, METEOR and TER metrics. As the study results, the BiLSTM template have succeeded in generating higher ratings on all metrics than others.