The real estate price index serves as a crucial indicator reflecting the operational status of the real estate market in China. However, it often lags until mid-next month, hindering stakeholders from grasping market trends in real time. Moreover, the real estate market has an extremely complex operating mechanism, which makes it difficult to accurately assess the impact of various policy and economic factors on the real estate price index. Therefore, we hope, from the perspective of data science, to explore the emotional fluctuations of the public towards the real estate market and to reveal the dynamic relationship between the real estate price index and online news sentiment. Leveraging massive online news data, we propose a forecasting scheme for the real estate price index that abandons complex policy and economic data dependence and is solely based on common and easily obtainable online news data. This scheme involves crawling historical online real estate news data in China, employing a BERT-based sentiment analysis model to identify news sentiment, and subsequently aggregating the monthly Real Estate Sentiment (RES) index for Chinese cities. Furthermore, we construct a Vector Autoregression (VAR) model using the historical RES index and housing price index to forecast future housing price indices. Extensive empirical research has been conducted in Beijing, Shanghai, Guangzhou, and Shenzhen, China, to explore the dynamic interaction between the RES index and both the new housing price index and the second-hand housing price index. Experimental results showcase the unique features of the proposed RES index in various cities and demonstrate the effectiveness and utility of our proposed forecasting scheme for the real estate price index.