The article discusses the role of a widely understood context, including frames and scripts, and utterances in text automatic translation. On the basis of several polysemic nouns of the French language analysed as an example, the author shows that in many cases of a lexicographic description following the object-oriented class presented by W. Banyś operators and attributes accompanying the above-mentioned nouns not always point to their proper equivalent in the target language. Therefore, it is the frame of an utterance that often happens to be the only criterion conditioning a correct interpretation and translation.