Recognition of sequential data using finite state sequence models organized in a tree structure

摘要: In a method of automatically recognizing data which comprises sequential units represented as tokens grouped into one or more items, known items are stored respective finite state sequence models. Each corresponds to token and the models have common prefix states organized in tree structure such that suffix comprise branches from there plurality structures each having different state. unit is compared with reference identified by generate scores indicating similarity units. An accumulated score for final determined steps (a) sequentially calculating model reach comprising leaf tree, (b) identifying closest branch corresponding next an stage has not yet been calculated, (c) accumulating These repeated trees. The item highest recognized best matching data.