Data annotation is the cornerstone of sentiment extraction processes within practically all technological approaches: rule-based, statistical, machine learning, neural or hybrid. The article presents a methodology for automated creation of a gold-standard corpus for multi-aspect sentiment analysis as well as results of the methodology application to a collection of general public evaluative opinions about the activities of medical institutions found on online platforms. The specificity of such online reviews as compared to literate media published texts and its influence on the annotation procedure is investigated. Described are the stages of a) review corpus acquisition and lexicon extraction b) conceptualization of lexical knowledge into sentiment and aspect related classes, whose label codes are used as a tagset, c) developing program tools to digitalize and store the linguistic knowledge, as well as to automate knowledge acquisition and annotation processes. The tools reused, updated and newly developed include a lexical extractor, grammar checker, annotation platform which, in turn, consists of an e-lexicon and a tagger that can produce coarse and fine levels of tagging. The research results, both the gold-standard corpus and tools are to be further used for aspect sentiment analysis at the following stages of our research. They may also deserve the attention of both computational linguists and sociologists.