Incorporating Linguistic Model Adaptation into Whole-Book Recognition

Abstract

Whole-book recognition is a document image analysis strategy that operates on the complete set of a book's page images using automatic adaptation to improve accuracy. Our algorithm expects to be given approximate iconic and linguistic models-derived from (generally errorful) OCR results and (generally incomplete) dictionaries-and then, guided entirely by evidence internal to the test set, corrects the models yielding improved accuracy. The iconic model describes image formation and determines the behavior of a character-image classifier. The linguistic model describes word-occurrence probabilities. In previous work, we reported that adapting the iconic model alone (with a perfect linguistic model) was able to automatically reduce word error rate on a 180-page book by a large factor. In this paper, we propose an algorithm that adapts both the iconic model and the linguistic model alternately to improve both models on the fly. The linguistic model adaptation method, which we report here, identifies new words and adds them to the dictionary. With 64.6% words missing in the initial dictionary, our previous algorithm reduced word error rate from 40.2% to 23.2%. The new algorithm drives word error rate down further from 23.2% to 16.0%.

Topic Language Model Adaption for Recognition of Homologous Offline Handwritten Chinese Text Image

“…By substituting (8) and (9) into (4), the Bayesian model for topic determination is: (10) In fact, the best matched topic can be alternatively found by imposing logarithm operation on (10). (11) In this paper, the top largest -character transition probabilities are used as the topic features.…”

“…In speech recognition, the language model adaption explores contextual information and improves recognition accuracy. Xiu [10] integrated language model adaptation into whole book recognition and greatly decreased recognition errors. Wang [11] achieved apparent accuracy improvement on ancient domain text image recognition with language model adaption.…”

Multiple-agent adaptation in whole-book recognition

“…This agent implements the linguistic post-processing procedure in. 10 Parameters: (1) a set of words with their normal frequency ranges; (2) a default frequency range for new words (most are rare words).…”