The aim of this study is to present an analysis of continued word association responses in Mandarin through the selection and validation of CKIP CoreNLP in Academia Sinica for grammatical category, word frequency and concreteness of lexical items. Semantic similarity in consecutive responses is analyzed by calculating the semantic distance using word embeddings whereas phonological similarity is measured by Levenshtein distance. Results from the three elicited responses in the continued word association task in Mandarin confirm type-token ratios and entropy found in English. Semantic and phonological distances are found to be significantly influenced by response positions. Semantic factors are more dominant to aid information retrieval. In contrast, the influence of phonological relatedness diminishes with consecutive elicited responses, suggesting that phonological similarity does not strongly facilitate the utilization of distinct levels of Mandarin lexicon. The implementation of the continued word association task provides valuable insights into the dynamics of lexical retrieval.