ETAP-3 is an outgrowth of the ETAP-2 MT system and, in its present state, has evolved into a multipurpose linguistic processor. It is implemented on a VAX-6340 computer. The programming languages used are PL/| and C. The processor’s two basic options are communication with relational databases in unconstrained natural language and bidirectional (English-to-Russian and Russianto- English) machine translation (MT). The latter system can joined to the former, so that the database can be accessed in either of the two working languages. The formats of linguistic knowledge representation are highly standardized and completely language independent, so that the number of working languages can be easily enlarged. Linguistic knowledge itself (morphological, syntactic and transfer rules, morphological, combinatory and semantic dictionaries) is not domain-oriented and does not require any re-adaptation (apart from the domain-specific part of the dictionaries) when transferred to a new object domain. The algorithms are geared exclusively to the formats of linguistic knowledge representation and do not require any re-adaptation either.