S. Sheremetyeva

South Ural State University (RUSSIAN FEDERATION)
In the paper a strong need for effective computer support in teaching professional writing is put into focus. Barriers to human understanding (readability) and machine translation (translatability) of technical texts are discussed. Technical documentation is an indispensible means of scientific and technical progress in the human society. Being an important communication media in the dissemination and assimilation of domain specific knowledge technical texts should be highly comprehensible for the interested audience both in the native and foreign languages. This is directly related to such parameters as text readability and translatability. Readability is related to the level of the clarity of a text for human understanding. Technical texts are often extremely difficult to understand (low readable) which also creates problems in human and machine translation which is nowadays getting more and more popular, not withstanding its quality failures. The efficiency of machine translation depends on the so called translatability indicators, text phenomena that lower the quality of machine translation. Translatability correlates (though does not coincide) with readability.

The mainstream of the research on improving text readability is associated with certain text simplification techniques for particular types of audience, e.g., poor literacy readers, readers with mild cognitive impairment, elderly people, language learners of different levels or just “regular” readers. As for text translatability, one of the most popular approaches to its improvement is to rewrite the source text into a controlled language to ensure that the machine translation input conforms to the desired vocabulary and grammar constraints.

In this paper we attempt to contribute to the solution of the mentions problems by suggesting an authoring tool that can be used to teach technical writing while addressing the issues readability and translatability. The tool makes the authors (students) aware of the typical areas of concern in the texts and provides an authoring environment to improve problematic fragments.

The target of our effort was thus defined by the intersection of the following criteria:
(i) detection of readability indicators (for humans);
(ii) detection of translatability indicators (for machine translation)
(iii) development of a controlled language imposing restrictions on the use of readability/translatability indicators
(iv) development of an authoring tool with the control language embedded into the tool knowledge base and included into a comprehensive, self-paced training material.

The authoring environment is interwoven with the hybrid analysis components and completely automatic generation module. The tool is illustrated on the Russian texts on engineering, but the methodology of both developing and using the tool is portable between domains and languages.