Multi-tool AI integration (TTS, LLM, SRS) for the initial stage of RCT: a 14-day analysis of vocabulary retention and pronunciation accuracy

Abstract

The work is devoted to substantiating and empirically verifying the effectiveness of the systemic integration of three AI tools (Anki/SRS, NaturalReaders/TTS, and ChatGPT/LLM) into the methodology of teaching Russian as a Foreign Language (RFL) at the zero level (A0). The relevance of the research is determined by the critical shortage of vocabulary and phonetic inaccuracy in beginners, which hinders the formation of stable oral output. The purpose of the study is to create and test an accelerated learning model that ensures the stable formation of basic vocabulary and multi-component phonetic accuracy (stress, intonation, rhythm) during a short 14-day introductory period. A 14-day empirical experiment was conducted based on the self-study of a single subject (N=1) beginning language acquisition. The methodology involved an immersive digital environment combining daily standardized input (TTS), algorithmic repetition (SRS), and corrective dialogue practice (LLM). Assessment included daily self-evaluation of TTS input, LLM ratings across four phonetic parameters, and analysis of objective SRS metrics. Key results showed: accelerated improvement in phonetic parameters was noted by the 5th–10th day, followed by a plateau (11th–14th days). High repetition accuracy was recorded at intervals greater than one day, along with a steady increase in vocabulary retention (approximately 60% to 88%) despite increasing intervals and ease factor. The data confirm the positive synergistic chain of simulated input – corrected output – sustained repetition . The conclusion is made about the advantages of the integrative model and the viability of its psycholinguistic orientation at the preparatory stage. Methodological recommendations are provided for transitioning from accuracy to fluency and pragmatic transfer .