In the paper, we introduce an algorithm of reduced word forms restoration for automatic speech recognition which takes into account the results of psycholinguistic experiments on spoken word recognition and includes, besides the main block, the module that retrieves the morphological information about candidates for recognition and the rules for word form processing and syntactic grouping. The source for the list of word forms and their realizations used by the algorithm is the Corpus of Transcribed Oral Russian Texts (http://narusco.ru/search/trn-search.php).