Integrated Semantic Representation (ISR-Model): Syntax-Independent Model for Natural Language

Today, one of the most widely used natural language representation methods is meaning representation for text processing in different systems. Meaning representation methods have had many applications in the field of natural language processing in recent years, including automatic text summarization systems and question and answer systems, etc. Many text expressions may be different in terms of grammatical structure, but they are the same in terms of meaning, so how to apply a single and uniform meaning to them is one of the challenges of text processing. The main goal of this paper is to design an Integrated Semantic Representation (ISR) model for natural language. The proposed model, while maintaining its simplicity in annotation and understanding, does not depend on specific natural language features or on syntactic and lexical structure. In this regard, several examples in both English and Persian language, which have different in terms of written and grammatical structure, are presented in this paper. Moreover, by using the proposed model, the same representation is provided for texts with the same meaning and different grammatical structure. The proposed model is designed in graph and list format is recommended to annotate its corpuses. One of the main features of this model is that it can represent semantic relations at both sentence-level and document-level and is able to represent complex and important linguistic phenomena such as aspect, tense, and quantification. The simplicity of the proposed model helps to avoid making language processing slow or complicated in various applications, and the preparation of structures based on this model for different natural languages will not be too complicated, so that it can be used both for natural languages with low resources and for those with various resources. Further, features of the proposed model are compared with one of the most important related works.