Abstract: The present invention has an architecture capable of eliminating a necessity for previously designating in a document, when reading aloud the document, attributes for reading aloud. Focusing on utilizing the document with attributes, a document reading system analyzes contents of the attributes and reads aloud texts in the document with a voice synthesizing module. The attributes are set irrespective of reading aloud conditions. A basic reading condition setting module sets the reading aloud conditions for the entire document. An individual reading condition setting module sets the reading aloud condition for every attribute. A selective reading module, when reading aloud the document, in principle, reads aloud the text with reference to the basic reading aloud condition set by the basic reading condition setting module, and selectively reads aloud the text having the individual reading aloud condition given a higher priority than the basic reading aloud condition.