Analysis of statistical methods for stable combinations determination of keywords identification

• Main Authors: Vasyl Lytvyn, Victoria Vysotska, Dmytro Uhryn, Mariya Hrendus, Oleh Naum
• Format: Article
• Language: English
• Published: PC Technology Center 2018-03-01
• Series: Eastern-European Journal of Enterprise Technologies
• Subjects: stable word combination; NLP; Information Retrieval; SEO; Web-mining; statistical linguistic analysis; quantitative linguistics; heading
• Online Access: http://journals.uran.ua/eejet/article/view/126009

The study has solved the task of making comparative analysis and choosing an optimal statistical method to determine stable word combinations while identifying keywords to process English-language and Ukrainian-language Web-resources. The effectiveness of the method directly proportionally depends on the quality of linguistic analysis, of Ukrainian and English texts, respectively, based on the technology of Web Mining and NLP. A decomposition of methods of linguistic analysis was performed to determine the impact on the quality of forming stable word combinations as keywords. The features of the method are the adaptation of the morphological and syntactic analyses of lexical units to the peculiarities of Ukrainian-language words/texts. To determine stable word combinations effectively, it is essential to exclude functional words (stops or references), pronouns, numerals and verbs because they are not related to the subject and content of a published work. A set of stable word combinations as keywords is determined by qualitative morphological and syntactic analyses of relevant texts. The set of the identified stable word combinations is used further to compare and determine the degree of the text relevance to a specific topic or user request. The internal “dynamics” of forming a set of stable word combinations as keywords was investigated in the study depending on the statistical method applied to the texts. The obtained results have been verified. The study has produced results of the experimental testing of the proposed content-monitoring method for determining stable word combinations to identify keywords in the processing of English-language and Ukrainian-language web-resources of the technical content based on Web Mining technology. It has been determined that the authors of published works often identify the keywords that are far from being considered. It has also been proven that the quality of the result is influenced by the quality of linguistic analysis of texts and subsequent filtering. Further experimental research requires approbation of the proposed method for determining keywords for other categories of texts – scientific, humanitarian, belletristic, journalistic, etc.