LearnPos: a new tool for interactive learning positioning

Carton, Cérès; Lemaitre, Aurélie; Coüasnon, Bertrand

doi:10.1117/12.2042379

Cited by 2 publications

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To do so, we need to define 8 positions operators, one for each different keyword. The position operators are inferred following the method presented in [11] with ground truth. It consists in the inference of the 6 parameters that compose them (definition of the zone boundaries and definition of the order of analysis of the elements in the zone).…”

Section: Rule Inferencementioning

confidence: 99%

Automatic and interactive rule inference without ground truth

Carton

Lemaitre

Coüasnon

2015

2015 13th International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

Dealing with non annotated documents for the design of a document recognition system is not an easy task. In general, statistical methods cannot learn without an annotated ground truth, unlike syntactical methods. However their ability to deal with non annotated data comes from the fact that the description is manually made by a user. The adaptation to a new kind of document is then tedious as the whole manual process of extraction of knowledge has to be redone. In this paper, we propose a method to extract knowledge and generate rules without any ground truth. Using large volume of non annotated documents, it is possible to study redundancies of some extracted elements in the document images. The redundancy is exploited through an automatic clustering algorithm. An interaction with the user brings semantic to the detected clusters. In this work, the extracted elements are some keywords extracted with word spotting. This approach has been applied to old marriage record field detection on the FamilySearch HIP2013 competition database. The results demonstrate that we successfully automatically infer rules from non annotated documents using the redundancy of extracted elements of the documents.

show abstract

Section: Rule Inferencementioning

confidence: 99%

Automatic and interactive rule inference without ground truth

Carton

Lemaitre

Coüasnon

2015

2015 13th International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

show abstract

Eyes Wide Open: an interactive learning method for the design of rule-based systems

Carton

Lemaitre

Coüasnon

2017

IJDAR

View full text Add to dashboard Cite

International audienceWe present in this paper a new general method, the Eyes Wide Open method (EWO) for the design of rule-based document recognition systems. Our contribution is to introduce a learning procedure, through machine learning techniques, in interaction with the user to design the recognition system. Therefore, and unlike many approaches that are manually designed, ours can easily adapt to a new type of documents while taking advantage of the expressiveness of rule-based systems and their ability to convey the hierarchical structure of a document. The EWO method is independent of any existing recognition system. An automatic analysis of an annotated corpus, guided by the user, is made to help the adaption of the recognition system to a new kind of document. The user will then bring sense to the automatically extracted information. In this paper, we validate EWO by producing two rule-based systems: one for the Mau-rdor international competition, on a heterogeneous corpus of documents, containing handwritten and printed documents, written in different languages and another one for the RIMES competition corpus, a homogeneous corpus of French handwritten business letters. On the RIMES corpus, our method allows an assisted design of a grammatical description that gives better results than all the previously proposed statistical systems

show abstract

LearnPos: a new tool for interactive learning positioning

Cited by 2 publications

References 10 publications

Automatic and interactive rule inference without ground truth

Automatic and interactive rule inference without ground truth

Eyes Wide Open: an interactive learning method for the design of rule-based systems

Contact Info

Product

Resources

About