OpenASR-CNidiom

This project is an education-oriented idiom recognition system, aiming at enabling idiom recognition direction through speech recognition technology. The key points of this project are low-resource artificial intelligence-speech recognition direction, data matching algorithm, and software engineering construction mode.

This project detailed code address: https://gitee.com/feeling-cool/Idiomrecognition

Hope you can star! If you have any questions, please send me an email. I’ll respond when I see it.

At present, our articles related to the project are in the state of repair and submission, and the specific papers will be published after receiving.

This project has developed a new recognition system for Chinese idioms for the benefit of speech recognition technology. The system is named OpenASR-CNidiom, which uses the acoustic model of the open source automatic Chinese speech recognition framework ASRT. We use three data retrieval algorithms to identify idioms. At the same time, the classical DTW algorithm of speech recognition is included in OpenASR-CNidiom for parallel prediction, so that openASR-CNidiom can achieve high recognition accuracy compared with the native ASRT.

OpenASR-CNidiom will use the acoustic model of the open source framework ASRT to predict sound sequences (idioms and pinyin sequences) when the identification begins after the input of the audio file. The pinyin sequence is given to the three data retrieval algorithms proposed in this paper, and the closest answer selected by the algorithm from the database is returned to the user. After that, identifying the correct speech sequence can be used as an audio template by the DTW algorithm, allowing the DTW algorithm to be selected for prediction the next time the same idiom is recognized.

We provided 100 audio files to test the accuracy of openASR-CNidiom. OpenASR-CNidiom achieves up to 88% accuracy, compared to the 8% accuracy of the open source framework ASRT. Collocation algorithm 1 FQ algorithm can achieve 48% accuracy collocation algorithm 2 SFQ algorithm can achieve 51% accuracy collocation algorithm 3 DFQ algorithm can achieve 88% accuracy

Usage of this project: Install the required library, set up the data table according to the data table in the SQL file, and insert the sql file in the Chinese idiom folder into the table. Then run GUI.py directly. Project Function Click the GUI button, according to the button to identify idioms and other operations.

I will continue to update the specific documents later.