The underlying assumption of the paper is that the text is the result of information transfer and the source (starting point) o of information mining and extraction. Thereafter the scientific text content  is universal and can be extracted in case of minimum coincidence of the author’s and recipient’s thesauri. In text translation the proper information extraction is determined with correct interpretation of terms/objects names. The paper focuses that adequacy of text perception and understanding on the lexical level is dictated by the text saturation with noun phrases, the degree of their compression and / or completeness of object nominations and their proper translations based on adequate bilingual dictionaries. Text translation in such a hazardous domain as aseismic construction is a crucial point in information exchange and mining and a safety challenge. For terminology analysis of this domain text a special research text corpus (1 mln tokens) had been created. The analysis results had been used in the specialized automatic dictionary creation for machine translation system WORD⁺.