The invention provides a visual text information discovery method based on a multilevel cooccurrence relationship word graph. The method comprises the following steps that: extracting the text contents of a document, and segmenting the text contents to obtain text segments; segmenting the text segments, extracting a keyword, and labeling a word category tag; according to the cooccurrence relationship of the keyword in the text segments, constructing the multilevel cooccurrence relationship word graph, wherein nodes in the graph correspond to the keyword, and an edge in the graph corresponds tokeyword co-occurrence constructing a word-document inverted index for each keyword in the graph for retrieving a document which contains the keyword; and through the cooccurrence relationship word graph, obtaining the visual text information. The invention also provides a visual text information discovery system based on the multilevel cooccurrence relationship word graph. The system comprises adocument preprocessing module, a keyword extraction module, a multilevel word graph construction module, a word-document index construction module and a visual information discovery module.