Adaptive and Interactive Approaches to Document Analysis

Abstract

Summary. This chapter explores three aspects of learning in document analysis: (1) field classification, (2) interactive recognition, and (3) portable and networked applications. Context in document classification conventionally refers to language context, i.e., deterministic or statistical constraints on the sequence of letters in syllables or words, and on the sequence of words in phrases or sentences. We show how to exploit other types of statistical dependence, specifically the dependence between the shape features of several patterns due to the common source of the patterns within a field or a document. This type of dependence leads to field classification, where the features of some patterns may reveal useful information about the features of other patterns from the same source but not necessarily from the same class. We explore the relationship between field classification and the older concepts of unsupervised learning and adaptation. Human interaction is often more effective interspersed with algorithmic processes than only before or after the automated parts of the process. We develop a taxonomy for interaction during training and testing, and show how either human-initiated and machine-initiated interaction can lead to human and machine learning. In a section on new technologies, we discuss how new cameras and displays, web-wide access, interoperability, and essentially unlimited storage provide fertile new approaches to document analysis.

New trends on digitisation of complex engineering drawings

“…This raises several interesting possibilities, for instance, the incorporation of human expert knowledge in a potential solution by means of human machine interaction. Interactive learning could be another possible direction [92].…”

Enforcing similarity constraints with integer programming for better scene text recognition

“…Again using four folds at a time for training, we estimated the parameter α from Eq. (18). We then solved the IP for the test fold and applied the error correction procedure of Section 4, often resulting in dramatic increases in word accuracy.…”

“…For example, in a long document, the most common symbol is usually the letter "e", but in a road sign, any letter could be most common. This makes it difficult to use cryptographic techniques [13,3,2,4,18]. Finally, many signs are meant to be eye-catching, and are for this reason designed with stylized, whimsical, or otherwise unusual fonts for which we are unlikely to have matching training data.…”

Image indexing in article component databases

“…Machine learning approaches require a very large document sample that we were unable to get, so the knowledge engineering approach was expedient 32008). This human interaction has been categorized as human‐initiated or machine‐initiated , and described as durable (when used to alter system parameters) or ephemeral (when used to label patterns or modify results, for instance) (Nagy & Veeramachaneni, 2008, p. 237).…”

“…It has been found that human interaction in the process is more effective when interspersed with the algorithmic process rather than when relegated to only before or after the automated parts (Nagy & Veeramachaneni, 2008). This human interaction has been categorized as human‐initiated or machine‐initiated , and described as durable (when used to alter system parameters) or ephemeral (when used to label patterns or modify results, for instance) (Nagy & Veeramachaneni, 2008, p. 237).…”