Seventy-seven preschool hearing children (NH) aged 2;11~6;3 (median 5;0) and thirty preschool children with hearing impairment aged 3;3~6;11 (median 5;3) were recorded. Among them, 20 children wore traditional hearing aids (HA, mild to profound), and 10 children were fitted with a cochlear implant (CI, severe to profound). For further details about the subjects, please refer to Appendix A. The children were instructed to read the story The Tortoise and the Hare with the help of six picture cards that were presented to them in a fixed order. The HA and CI children were recorded during their regular auditory-verbal therapy (AVT) session in sound-proof classrooms (Dornan et al. 2007), using the built-in video equipment of the Children’s Hearing Foundation. The NH children were either recorded at Academia Sinica in sound-proof studios or in quiet classrooms at their kindergarten by the second author’s research team. The data were digitized at a sampling rate of 44.1 kHz with 16 bit quantization. The speech content was transcribed in traditional Chinese characters with annotations of paralinguistic sounds and pauses. The ILAS phone aligner, an automatic phone aligner developed at Academia Sinica, trained by adult conversational speech data, was used to obtain timestamps for syllable and phoneme boundaries (Tseng 2019). Human verification was performed on the alignment results to ensure high-quality boundary precision. Signal-aligned word boundary information was then generated by integrating information derived from results of the CKIP automatic word segmentation and POS tagging system (Chen et al. 1996, Tseng 2013) with post-edited syllable boundary annotations. For the complete list of the CKIP POS tagset, please refer to Tseng (2013, p.7).