Form reading technology based on form-type identification and form-data recognition is proposed. This technology can solve difficulties in variety for reading different items on fairly large number of different types of forms. The form-type identification consists of two parts: (i) extraction of targets such as important keywords in a form by matching between recogised characters and word strings in a keyword dictionary, and (ii) analysis of positional or semantic relationship between the targets by constellation matching between these targets and word location information in the keyword dictionary. The formdata recognition consists of two parts: (i) extraction of a region of interest (ROI) contained a character string of the item by using a layout knowledge of the very form-type, and (ii) character string recognition of the item by using the linguistic constraint which can be obtained from a content knowledge of the form-type. A experiment using 642 sample forms with 107 different types in total confirmed that the form-type identification method can correctly identify 97% of 642 form samples at a rejection rate 3%. Another experiment confirmed that the form- data recognition method can correctly read 95% of the number of items on the form samples.