In this paper, we develop a keyword spotting system using vocabulary-independent speech recognition technique, and investigate several non-keyword modeling methods to improve its performance. In order to overcome the weakness of conventional syllable model, we propose the syllable filler based on syllable information of keywords and syllable-like filler model. The former prohibits syllable filler network from taking the common syllables that keyword network has for better descrimination between keywords and filler. According to our experiments, syllable filler model using syllable information of keyword yields error reduction rate of 52%-54%. The latter constructs syllable filler network by concatenating the clustered CI phonemes classes. It leads to a 75 times faster decoding than conventional syllable filler while not requiring a large size of text corpus. Keywords: vocabulary-independent keyword spotting, non-keyword modeling.