Knowledge-driven text generation-based speech recognition field adaptive method and system

The embodiment of the invention provides a voice recognition field adaptive method and system based on knowledge-driven text generation. The method comprises the following steps: inputting target domain knowledge into a knowledge description framework for filling, and guiding a large language model...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: LIU SEN, YANG BAOCHEN, YU KAI, GUO YIWEI, ZHANG XIZHUO, LIANG ZHENG
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The embodiment of the invention provides a voice recognition field adaptive method and system based on knowledge-driven text generation. The method comprises the following steps: inputting target domain knowledge into a knowledge description framework for filling, and guiding a large language model to generate a target domain text conforming to the target domain knowledge; inputting the target domain text into a text-to-speech model to generate first training data; performing domain-adaptive first optimization training on the speech recognition model by using the first training data, and outputting a target audio hypothesis of the first training data by using the trained speech recognition model; inputting the target audio hypothesis into a large language model for knowledge-driven text generation iteration to obtain second training data; and performing domain-adaptive second optimization training on the speech recognition model by using the second training data to obtain a domain-adaptive speech recognition