Natural language serves as important information source in all areas of human activity. The presence of a huge amount of texts on the Internet actualizes the problem of efficient information search; visual scanning of all the textual information is difficult and time-consuming. There is a need for efficient, high-quality systems that extract the relevant information from texts. The paper presents the architecture of an experimental system for automatic text understanding and information extraction, which has originally been developed for the domain of news reports. The possibility of adapting the methodology for the purpose of bioinformatics is discussed, and the similarities and differences between texts in the two different domains are discussed and exemplified.