Basic architecture, methods and algorithms of system for temporal information extraction from natural language texts

This article presents the basic architecture of the system for extracting temporal information from texts in natural language. The basic structural components of such an architecture are determined, as well as the methods and algorithms that are implemented in them. In particular, the stage of extraction of information about temporal elements - events and temporal references in the text is allocated. It emphasizes the need to use the syntactic dependencies between the words of the sentence being processed, and also the semantic roles of the word groups to establish the order relations between the temporal elements extracted from the text. A separate important component of the proposed architecture is the logical inference module that uses statistical information. Accordingly, the need to use machine learning methods and various cases of linguistic data (similar to WordNet, SemCor, TimeBank, etc.) is shown to successfully solve a common problem. Key words and phrases: Extraction of information, temporal elements, machine learning