Research behind this paper explores the relationship between social media sentiment signals and market prices, and the predictive strength of these signals for selected assets traded on Taiwanese and Hong Kong markets. The research utilizes data sourced from PTT Bulletin Board System (BBS) and ET Net forum, analyzing sentiment signals and their association with Return on Investment (ROI) through various analytical techniques including signal association analysis, causality analysis, and model back-testing. The findings demonstrate a significant association between high sparsity sentiment signals and ROI, with machine learning models showing improved performance over traditional Buy and Hold scenarios in both markets. Out of 14 assets represented on ET Net forum, machine learning models outperform the Buy and Hold scenarios in 9 cases. Out of 10 assets with data from PTT BBS, models trained with data have shown improved results for 6 assets. ML approach outperforms Buy and Hold scenarios in both markets, with more significant gains (20% in absolute ROI) on Taiwanese market, where also all benchmarks are exceeded.