Abstract: A text information extraction device extracts analysis networks from texts and stores them in a database. The analysis networks consist of lines each including elements and relations extracted from the texts. The analysis networks are complemented via synonym/near synonym/thesaurus process and via complementary template and the lines thereof are weighted via concept template. A text similarity matching device judges similarity of input and database analysis networks on the basis of agreements of words, word pairs, and lines. A text search system stores texts and complementary term lists prepared therefrom in respective databases. Queries are inputted in the form of analysis networks from which sets of keywords and relations are extracted. After searching the texts and complementary term lists stored in databases with respect to the keywords extracted from each input query, agreements of the sets of keywords and relations are determined.