Publications

Advanced search

Abstract

Alejandro H. Toselli, Verónica Romero, Enrique Vidal. Viterbi Based alignment between Text Images and their Transcripts. Language Technology for Cultural Heritage Data (LaTeCH 2007), 2007. pp. 9-16.

An alignment method based on the Viterbi algorithm is proposed to find mappings between word images of a given handwritten document and their respective (ASCII) words on its transcription. The approach takes advantage of the underlying segmentation made by Viterbi decoding in handwritten text recognition based on Hidden Markov Models (HMMs). Two HMMs modelling schemes are evaluated: one using 78-HMMs (one HMM per character class) and other using a unique HMM to model all the characters and another to model blank spaces. According to various metrics used to measure the quality of the aligmnents, encouraging results are obtained.