A detailed survey of Turkish automatic speech recognition

Abstract

Significant improvements have been made in automatic speech recognition (ASR) systems in terms of both the general technology and the software used. Despite these advancements, however, there is still an important difference between the recognition performance of humans and machines. This work focuses on the studies conducted in the field of Turkish speech recognition, the progress made in such studies in recent years, the language-specific constraints, the performance results achieved in the applications developed to date, and the development of a general scheme for researchers wishing to develop an ASR system for the Turkish language. A comprehensive study on the Turkish language, including all literature and speech recognition systems, was prepared. There has been significant progress in the languages of developed countries in terms of ASR systems, but it was observed in this work that the opposite is true for the Turkish language. The lack of labor force and material resources remains the biggest obstacle to developing better systems. As a result, concrete studies are needed to achieve an ASR system with performance equivalent to the human level for the Turkish language.

Customized deep learning based Turkish automatic speech recognition system supported by language model

“…It has been observed that these studies commonly incorporate layers such as GRU, LSTM, and Transformer, resulting in the most successful outcomes among the methods utilized ( Chiu et al, 2018 ; Liu et al, 2018 ; Hsu et al, 2021 ; Yu et al, 2022 ). It has been noted that a majority of ASR studies in existing literature primarily concentrate on languages such as English, Chinese, Spanish, German, and French, whereas there is a scarcity of research focusing on languages within the agglutinative language group, such as Turkish ( Arslan & Barışcı, 2020 ).…”

“…The spectrograms obtained at this stage served as inputs for the proposed deep learning models, which predicted the character sequences present in the audio data. A high WER in the prediction scores obtained from deep learning models is a common issue, particularly for agglutinative languages like Turkish ( Arslan & Barışcı, 2020 ). Although CER is lower than WER, significant improvements in WER can be achieved by using a good language model.…”

“…Considering all these situations, within the study’s scope, a language model was developed to enhance the performance of the proposed models. Upon reviewing the literature, it becomes evident that there are few studies dedicated to the Turkish language, with the majority focusing on languages like English, Chinese, Spanish, German, and French ( Arslan & Barışcı, 2020 ). In addition to the complexity of the language, the limited availability of studies also results in a lack of accessible resources.…”