In the age of global communication, the need for fast, accurate and cheap translation from one language into another has become very pronounced. This situation can turn critical when considering translation and interpreting in high-risk technology domains. The discrepancy in codes, norms and standards in various countries as well as delayed information exchange gives rise to increasing disagreement in high-technology and dangerous fields of common engineering interest. The main subject of this paper is the correlation between the main problems and achievements of the present-day machine translation and translation memory systems and the possibilities of using such systems as real tools for a translator. This means investigating different kinds of linguistic and extra-linguistic knowledge and taking into consideration interference from the authors’ mother tongues when they generate an English text in a special domain. A solution is to consider the text structure and to use translation memory principles when a single text is processed in a machine translation system.