The current paper suggests a new representation type of word collocations—the semantic sketches. It was first tested on one of the subcorpora of the General Internet-Corpus of Russian. The semantic sketches continue the idea of word sketches based on grammatical relations between words and expand it by adding the semantic information—word meanings and semantic relations between words. Moreover, the sketches can be additionally provided with metatextual characteristics. Certainly, building such sketches demands the semantic markup of the corpora. Therefore, we have used partial semantic analysis of the Compreno parser for our purposes. The paper demonstrates the examples of the sketches, provides the quality evaluation of the markup they are based on, and shows the advantages and disadvantages of the given approach.

The article summarizes the results of the long-term project "Languages of Russian Cities" (LoRC) of the regional vocabulary collecting and researching, which, unfortunately, was not depicted in any academic publications for a number of reasons. About 4 thousand pieces of regional materials were collected, systematized, and became the basis of the typology of regional differences consideration and the concept of a regional norm discussion. Reliability issues and methods of computer-based regional corpus research, including automatic text classification and author profiling, are paid attention to. Along with this article, the "reincarnation" of the LoRC project is also returning to the fund of open lexicographic resources basing on the joint portal for distinctive sociolinguistic research, which includes the General Web-corpus of Russian Language and the interactive dictionary "Languages of Cities and People" (LoC&P)

The paper presents the General Internet Corpus of the Russian Language (GICR) as a tool for linguistic research. Problems are identified that are common to any WEB-corpus that affect the reliability of such research. Among the problems considered: the importance of taking into account sociolinguistic variability, the influence of falsely attributed texts, thematic biases, the prospects and disadvantages of new methods for corpora output aggregation. A distinctive feature of our approach is the emphasis on linguistic significance, reliability, and interpretability of the results obtained.