?
Towards Creating a Semantic Platform for News Article Adaptation
In this paper we present an approach to lexical adaptation of news articles written in Russian language. Such adapted texts could then be used in learning/teaching Russian as a foreign language. To this end, we propose an algorithm of finding substitutes for low-frequency words by ranking all hypothetic synonyms. The ranking is based on whether the substitute candidate is included in the dictionaries of synonyms, hypernyms and hyponyms, and in lexical minimum for the target proficiency level in Russian, the frequency of the word in question, and its semantic proximity to the word that is to be replaced. We performed empiric analysis of two methods for measuring semantic proximity. The first relies on the vector of normalized frequencies of word use in the nearest context. The second is based on the Distributional Semantic Model. We have found that in both cases contextual proximity yields useful results for synonym ranking.