Review of Research on Mongolian Fixed Phrases Recognition

Mongolian fixed phrase recognition is one of the most fundamental tasks in Mongolian natural language processing, and its main purpose is to identify the fixed phrase boundaries and types with specific meanings in Mongolian. We introduce the concept, classification, and evaluation of Mongolian fixed phrases at first. Secondly, according to the research process of Mongolian fixed phrase recognition, methods are classified into two categories: dictionary-based and rule-based methods in the past, which are summarized. A detailed comparative analysis reveals the effectiveness and limitations of these methods in practical applications. Then, due to the lack of literature and research, we implement three Mongolian fixed phrase recognition models based on sequences labeling, and briefly introduce the development of Mongolian named entity recognition based on deep learning. Finally, the research trends of Mongolian fixed phrase recognition are discussed to provide some reference for the proposal of new methods and future research directions. Due to the lack of reference studies, we implement a sequences labeling-based model using the available corpus and conduct a discussion and analysis.