Forecasting in the financial domain is undoubtedly a challenging undertaking in data mining. While the majority of previous studies in this field utilize historical market data to predict future stock returns, we explore whether there is benefit in augmenting the prediction model with supplementary domain knowledge obtained from financial news reports. To this end, we empirically evaluate how the integration of these data sources helps to predict intraday stocks returns. We consider several types of integration methods: variable-based as well as bundling methods. To discern whether the integration methods are sensitive to the type of forecasting algorithm, we have implemented each integration method using three different data mining algorithms. The results show several scenarios in which appending market-based data with textual news-based data helps to improve forecasting performance. The successful integration strongly depends on which forecasting algorithm and variable representation method is utilized. The findings are promising enough to warrant further studies in this direction.