A digital audio-video information reproducing apparatus and a reproducing method thereof are provided, in which a predetermined time point or a keyword from users is used as a basic for controlling the reproduction of a file-based AV file. If the subtitle is in a bitmap format, the bitmap subtitle is transformed to text contents. If the subtitle is in a text format, the subtitle is directly interpreted. Then, the interpreted/transformed text content is compared with the keyword, so as to interpret a matched time point corresponding to the subtitle, and the matched time point is set as a start or end point for reproducing the AV file. Alternatively, the comparison result may be applied in analyzing a correlation between the AV file and the keyword.