Ontology of Graphemes for Latin Character Recognition

Available online at www.sciencedirect.com

Procedia Engineering

ProcediaProcedia Engineering 00 (2011) Engineering 24 000–000 (2011) 579 – 584 www.elsevier.com/locate/procedia

2011 International Conference on Advances in Engineering

Ontology of graphemes for Latin character recognition Hacene Belhadefa, Mohamed Khireddine Kholladib, Aicha Eutamenec a* a,b,c

Department of Computer Sciences, Mentouri University of Constantine, Road of Ain ElBey 25017, Algeria

Abstract In this paper, we present a general description of an ontology, describing the different character of Latin language, which is semantically annotated by a domain expert. This ontology represents the typographical description of the analyzed and processed document, by a set of concepts where each concept, represents a grapheme of character and must be designated by a unique number denoting its position or its localization in the document. The relationship between, two graphemes is of spatial type; it represents the connectivity between graphemes in the document to facilitate the recognizing process of characters. Our ontology is generic and can support other languages by enriching it by new specific spatial relationships.

© 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of ICAE2011. Keywords: Character recognition; grapheme; ontology; spatial relationship; typographical features

1. Introduction The general goal of the recognition of documents; whether printed or manuscript, is to transform them into understandable and usable representation by machine. The process of recognition is not always easy as long as the content of documents can have multiple representations. In the case of printed documents, size, style (Bold, Italic... etc.), the cast of characters and other factors play a crucial role in this process. Images are the basic means for information representation of document content, they are used in many areas in scientific research, medicine, ecology, chemistry, engineering, military science ... Etc. The development of automated systems for image processing, analysis, evaluation, and understanding is an important branch of oriented fundamental and applied research. Over the last years, ontologies have demonstrated to be an important and useful instrument for representing, sharing and reusing knowledge. Evidences for that are: ontology languages allow for expressing rich semantics and provide reasoning capabilities. Ontological knowledge representation has the following advantages: (1) it provides the opportunity to

* Corresponding author. Tel.: +213-31-818494; fax: +213-31-818494. E-mail address: [email protected]

1877-7058 © 2011 Published by Elsevier Ltd. doi:10.1016/j.proeng.2011.11.2699

2580

Hacene Belhadef al. / Procedia Engineering 24 (2011) 579 – 584 Hacene Belhadef et al./ et Procedia Engineering 00 (2011) 000–000

establish a common understanding of the considered field of knowledge, (2) it enables us to represent knowledge in a convenient form for processing by automated information processing and analysis systems, and (3) it provides the opportunity for acquisition and accumulation of new knowledge and for multiple use of knowledge [5]. So using ontologies in pattern (document, character, Etc) recognition, we bridge the gap between the needs of new trends of scientific research and proposed solutions in this area. 2. Related work There has been much research using ontologies to improve image search, but only Belhadef and al., have used the ontologies in character recognition area[1][9], they have described a new process based on ontology, this last is annotated by a human expert and contains concepts representing the graphemes, where each grapheme represents a part of character in the document. In this section we will cite some works but we will not exhaustive, e.g. Hyvonen, et al., use ontologies to aid annotation, user search and browsing [10]. Their technique requires changes to the way the images need to be annotated and the way the users can find the images. Schreiber et al., explore the use of ontologies to index and search collections of photographs [11]. They developed a tool for annotation and searching for specific images. Windson Viana and al., developed a mobile and location based system, called PhotoMap, for the automatic annotation and the spatial-temporal visualization of photo collections using ontology [12]. Juliusz L. Kulikowski developed an approach to computer-assisted image content understanding which is based on using domain ontology as a source of knowledge concerning the external world whose objects are visualized by the image [13]. And the works of Gurevicha I. B. [14][15][5], that are classified very interesting in this domain, we can cite, An Ontological Framework for Media Analysis and Mining, Cell Image Analysis Ontology, Fundamental concepts and elements of image analysis ontology. 3. Character recognition Character recognition is a synonym for the automatic conversion of printed characters into editable text files. Optical Character Recognition (OCR) is a process converting scanned printed images, mark, character, illustration or handwritten text into a format that is identifiable by computer, such as ASCII. OCR scanners can read envelopes, full pages of document text including broken characters and also common fonts. OCR scanner converts printed object into bitmap image, while OCR software converts bitmap image into ASCII characters. The ASCII characters then can be imported into a variety of word document, page layout, spreadsheet, database, or graphical processing application. This greatly helps the process of filing millions of paper documents [6]. 4. Ontology The term "ontology" comes from the field of philosophy that is concerned with the study of being or existence. In philosophy, one can talk about ontology as a theory of the nature of existence. In the context of computer and information sciences, ontology defines a set of representational primitives with which to model a domain of knowledge or discourse. The representational primitives are typically classes (or sets), attributes (or properties), and relationships (or relations among class members). [7] The preceding definition leads to a set of definitions that can be used as a basis for algebraic formulation of the term Ontology and its components [4]: Definition 1. A term is a triple τ = [η, δ ,Α] , τ ∈ Τ , where η is a string of characters containing the name of the term, δ is a string of characters containing its definition and Α is a set of attribute domains A 1 , A 2 , ..., A n , each associated to a value set V i . Definition 2. relation φ : Τ → Τ , φ ∈Φ , : is a function from Τ to Τ such that for every term τ 1 ∈Τ,

3581

Hacene Procedia Engineering Engineering 00 24 (2011) (2011) 000–000 579 – 584 Hacene Belhadef Belhadef et et al. al.//Procedia

there is a term τ 1 = φ(τ 1 ),τ 2 ∈Τ. Definition 3. A semantic relation σ between two terms is a relation that belongs to the set of semantic relations Σ = {Hypernymy, Hyponymy (is-a), Mereonomy (part-of), Synonymy }, Σ ⊂ Φ . Definition 4. A spatial relation ρ between two terms is a relation that belongs to the set of spatial relations Ρ = {adjacency, spatial containment, proximity, connectedness}, Ρ⊂ Φ. Definition 5. An ontology is a Θ=[ Τ, Φ] , where Τ= {τ 1, τ 1 , ....,τ n } is a set of terms, and Φ= {Φ 1, Φ 2,......, Φ n }, and ∃φ i ∈( ∑∪ Κ ) . 5. Grapheme The grapheme is the fundamental unit of a writing data; it is the smallest unit of meaning graph whose variation changes the value of the sign in writing b. For ideographic scripts, it can represent a concept. In phonographic writing, it represents an element of achieving sound (syllable, consonant, and letter). So in alphabetic writing, the grapheme is commonly referred letter c. In our work, we show the interest of using the graphemes as features for describing the individual properties of Handwriting (Part of character). Each grapheme is produced by the segmentation module [2] of a recognition system. At the end of a segmentation step, the document can be viewed as the concatenation of some consecutive graphemes (Fig. 2). The handwritten document D is thus described by the set of graphemes X i D={ Xi , i:1 to n}

(1)

A subset of successive graphemes, may construct a word W j , or a single character C k : Wj={X i , i:1 to m, m
(2)

C k ={X i , i:1 to p, p
(3)

6. Our approach The main goal of image analysis ontology (IAO) development, especially character recognition in document is formal description of knowledge on the processing, analysis, and recognition of images accumulated and used by experts [5]. Our approach is based on the architecture of a new system of character recognition manuscript, which was described in [1]. Our contribution is located between the step of feature extraction and recognition step, it is expressed by instantiating the ontology already created by a domain expert (see Fig. 1). In this figure we show just an excerpt of the global ontology representing all the Latin alphabet. In this ontology, we implemented all possible relationships, which can be found between the different graphemes that build the characters; these relationships are of spatial type (Just-below, Next-left…Etc). For example, the grapheme Tal-trunk can have multiple relationships with graphemes: Bar, Arc, etc.... to form respectively the characters L and P.

b c

wikipedia http://alis.isoc.org/glossaire/grapheme.fr.htm

4582

Hacene Belhadef al. / Procedia Engineering 24 (2011) 579 – 584 Hacene Belhadef et al./ et Procedia Engineering 00 (2011) 000–000

Fig. 1. ontology of graphemes represents the alphabet of Latin language

Fig. 1 presents our ontology that created under protégé editor d . "Protégé" is the best known editor and most widely used, it is open source, developed by Stanford University, it has evolved since its first versions (Protégé-2000) to integrate from the 2003 web standards including OWL semantics. It offers many optional components: reasoners and graphical interfaces. In the Fig. 2, we show an example of the different graphemes that form the word “LIRE”, specially the character “L” that is composed by two graphemes such as Tall-trunk and Bar, respectively numbered by number 1 and 2. The relationship between the grapheme number 1 and 2, is clearly depicted in Fig. 3, and the corresponding code of this relationship is shown below.

LIRE Grapheme 1 Tall trunk

Grapheme 2 Bar

Grapheme 2 Bar

Fig. 2. Segmentation of the word “LIRE” by graphemes features

d

http://protege.stanford.edu/

Hacene Procedia Engineering Engineering 00 24 (2011) (2011) 000–000 579 – 584 Hacene Belhadef Belhadef et et al. al.//Procedia

RDF e description of data type property “ next_left “ betwen the classes: “tall_trunk" and” bar”

"is a" relationship "instances" relationship

Fig 3. Spatial relationship between graphemes

7. Conclusion The work presented in this paper presents a new and original vision, to solve the problem of character recognition, our idea is expressed through the creation of an ontology that represents the contents of a document written manually or automatically (by seizure). This ontology transforms the image-version of document to a representation of concepts and spatial relationships. The idea here is not limited to the representation of the content but rather the exploitation of side semantic of this content, taking advantage of all the benefits of ontologies, including the formulation of local queries which are necessary for making intelligent decisions, or creating the web services that can work above. The development of ontologies in this area can be used to provide image analysis automation support and efficient use of modern methods and techniques for image analysis and pattern recognition. In a future work, we plan to use the matching operations with external resources such as WordNet f, using dedicated similarity measures.

e f

Resource Description Framework WordNet : A lexical database of English language: http://wordnet.princeton.edu/

5583

6584

Hacene Belhadef al. / Procedia Engineering 24 (2011) 579 – 584 Hacene Belhadef et al./ et Procedia Engineering 00 (2011) 000–000

References [1] Belhadef H, Eutamene A, Kholladi M K, A General Description of a New Character Recognition Process Based on Ontology, International Conference, ICDIPC 2011. Proceedings, Part II, Published in the CCIS Series, Vol. 189 of Springer LNCS, ISBN 9783-642-22409-6, Ostrava, Czech Republic, July 7-9, (2011) [2] Bensefia A, Paquet T, Heutte L, Information retrieval based writer identification. 7th International Conference on Document Analysis and Recognition (ICDAR2003). 946-950 ( 2003) [3] Belaïd A, Reconnaissance automatique de l'écriture. Pour la science (2001) [4] Fonseca F, Davis C, Câmara G, Bridging Ontologies and Conceptual Schemas Geographic Information Integration. Geoinformatica, Volume 7 Issue 4, 355-378. (2003) [5] Gurevicha I B, Salvettib O, Trusovaa Yu O, Fundamental Concepts and Elements of Image Analysis Ontology, Pattern Recognition and Image Analysis. Vol. 19, No. 4. 603–611 (2009) [6] Material Handling and Industrial Distribution Lab : Optical Character Recognition, http://cobweb.ecn.purdue.edu/~tanchoco/MHE/ADC-is/OCR/main.shtml [7] Thomas R. Gruber, A Translation Approach to Portable Ontology Specifications Knowledge Acquisition, 5(2):199-220 (1993) [8] Vers la communication sur Internet dans toutes les langues, http://alis.isoc.org/glossaire/grapheme.fr.htm [9] Eutamene A, Belhadef H, Kholladi M K, Ontology based New Approach for Character Recognition International Journal of Computer Applications 21(10):40–44. Published by Foundation of Computer Science. DOI: 10.5120/2623-3648, (2011) [10] Hyvonen E, Saarela S, Styrman A, Viljanen K,Ontology-based image retrieval. In: WWW (Posters). (2003) [11] Schreiber A T G, Dubbeldam B, Wielemaker J, Wielinga B, Ontology-based photo annotation. IEEE Intelligent Systems 16(3) 66-74 (2001) [12] Viana W, Bringel J, Gensel J, Oliver M V, Martin H, PhotoMap: annotations spatio-temporelles automatiques de photos personnelles pour les utilisateurs nomades.Actes de la Conférence Québéco-Française pour le Développement de la Géomatique, SAGEO’2007 [13] Kulikowski J L, Ontological Models as Tools for Image Content Understanding, L. Bolc et al. (Eds.): ICCVG 2010, Part I, LNCS 6374, pp. 43–58, 2010. Springer-Verlag Berlin Heidelberg, 2010 [14] Colantonio S, Salvetti O, Gurevich I B, Trusova Yu, An Ontological Frame work for Media Analysis and Mining, ISSN 1054-6618, Pattern Recognition and Image Analysis, 2009, Vol. 19, No. 2, pp. 221–230. Pleiades Publishing, Ltd., 2009. [15] Colantonio S, Martinelli M, Salvetti O, Gurevich I B, Trusova Y O, Cell Image Analysis Ontology, ISSN 1054-6618, Pattern Recognition and Image Analysis, 2008, Vol. 18, No. 2, pp. 332–341. Pleiades Publishing, Ltd., 2008