Open data is drawing attention for the creation of innovative services in recent years. For promoting a greater number of consumer services on the web, a search function that can reveal what kinds of data are available would be helpful. However, if we assume that the open data is described in a triple language like Resource Description Framework (RDF) in future, full-text search is not suitable for data fragments, and a formal query language is difficult for ordinary users. Therefore, we propose a question answering service based on Linked Open Data. As the problems of using the open data as a knowledge source, we then focus on mapping of question sentence to data schema, and data acquisition. And we propose improvement of accuracy based on user feedback and acquisition of new data by user context information. We also present `Flower Voice' which is an application of the service for assisting with fieldwork and confirm the effectiveness.