Forecasting models of the financial market generally use time series data. However, external factors can influence time series, such as political events, economic crises, government macroeconomic policy, and the foreign exchange market. This information is not explicit in the time series and can influence the prediction of the variable values. Textual data can provide knowledge about external factors for time series forecasting models. Representations models of time series that include data extracted from texts have been proposed and pointed out as a research trend. The combination of different data sources for forecasting tasks is known in the literature as information fusion. Thus, this study aims to systematically map the main information fusion techniques used to forecast the financial market. More than three hundred works were done with the search strings. However, the complete reading was carried out in only fourteen works. Analyzes were performed to identify the vector representations of texts, information fusion techniques, and predictive models used in the works. As a result, we identified that 78\% of the included works are techniques based on Early Fusion, and textual representations that consider semantic features have been more used in recent works.