This paper examines some suitability aspects of existing web search engines in relation to the content and the stated learning objectives of language teaching in Cypriot secondary education, focusing on the language course of the third high school grade (G9). The end goal is to put the internet in the service of schooling; specifically to categorize the results returned by the search engine into genres in order to facilitate user (teacher or student) in choosing the most appropriate texts for their learning purposes. The tools for categorizing texts are being sought in the field of Semantic Web technology, such as metadata, ontologies, software agents, and, the techniques in the fields of Natural Language Processing (NLP), Information Retrieval (IR), Information Extraction (IE) and Text Mining. The paper proposes the categorization of texts into six major genre categories according to their external (structural) and internal (linguistic, stylistic) characteristics. For the purpose of this research, the MeDa13 metadata model was designed on the basis of the standard metadata model Dublin Core, and the Textual Genres Ontology (TeGO) was developed for describing the concepts mentioned in genres. In this work, we present the theoretical background for the development of the proposed models (MeDa13 and TeGO), and also the methodological plan to achieve the research objective, which is the categorization of texts into genres considering the content and learning objectives for language teaching.
The internet as a source of information and a means of communication, either
synchronous (chat) or asynchronous (email, blogs), has a dominant place in all fields of
human everyday life, including education. The fact that searching and information
gathering is the most common activity on the internet, brings up several questions
regarding the real usefulness of the internet in school education and the suitability of
the results returned by the existing classic web search engines in relation to the
context and the learning objectives as set out in the curriculum of Cyprus Ministry of
Education and Culture. Assuming that the internet is useful for the results returned to
the user, this study seeks answers regarding the suitability of the results returned
focusing on the learning purposes of language course of the third high school grade
(G9). The frequent return of unsuitable results, as noted in the international literature,
is related to the operation of existing web search engines, which are based on using
keywords and conducting searches that are related more to the word spelling than
with semantics [
The end goal is to design and develop a system that will distinguish textual genres, for example, a history text from a literary text, a text of the daily press from a sketch etc, and will organize internet data according to the textual characteristics of each genre. The first step is the definition of each (textual) genre and the clear wording of assessment criteria for categorizing texts. The long term vision is to build a search engine that returns the search results categorized into genres. This search engine will operate on the basis of specific assessment criteria for the texts, search the internet, filter all the data and sort them in the categories of genres. Therefore, the research looks for a tool that will 'help' the computer to process data, to understand the meaning e.g. of a historical text and the definition given for this genre, in order to filter the texts and subsequently to classify them into the genre categories that have been created. This means that the information stored in the computer should be semantically enriched. The research is oriented in the field of SW and seeks the tools with which the semantic enrichment of the information will be achieved, and, subsequently, the classification of internet data in the categories of genre created for the needs of this research. Additionally, the study looks for a tool that will select texts from all the available texts on the web and correctly register them in the appropriate categories.
This work proposes the automatic categorization of Greek texts that are available on the internet into genres, as presented and used in the school textbook for teaching Modern Greek language in the third high school grade (G9). To achieve this goal, an algorithm is developed and the relative techniques for categorizing texts are inquired in the fields of Natural Language Processing (NLP), Information Retrieval (IR), Information Extraction (IE) and Text Mining. The tools of SW technology, such as metadata and ontologies, are used to define the concepts or terms and the relationships between them, and to describe the related knowledge of a specific thematic domain, e.g. Medicine, Education etc., respectively. For the purposes of this study, a metadata model called MeDa13 is developed on the basis of the standard metadata model Dublin Core, and the Textual Genre Ontology (TeGO) is formed in order to represent the knowledge which is relevant to genres. 2
The main research questions of this study are summarized as follows: 1. Does the operation mode of the internet and existing search engines meet the learning objectives of schooling, particularly for language teaching? Are the results returned from the existing search engines suitable for the content and the learning objectives of language teaching? 2. What would be the ideal scenario for the operation of the internet in relation to the educational needs? What would make an 'ideal' search engine? (ideal = it would serve the learning objectives, return the most appropriate results in relation the query set by the user). 3. Is the appropriate technology to 'place' the internet in the service of schooling available? Is SW the solution for ensuring the suitability of the results returned, always in connection with the content and the intended learning purposes? Which tools are available (metadata, ontologies, software agents etc.) and, which of them can be used to better organize the material on the internet in order to make it suitable for use in language teaching? Are the existing models suitable for the purposes of this work? 4. How should the material on the internet be organized? Which techniques for categorizing texts are available and which of them serve the objectives of this work? 3
Search engines usually operate on the basis of keywords depending on the query entered by the user. The fact that these searches relate more to the spelling of the keyword than to semantics often results in unsuitable results. This is due to the construction and operation mode of the existing web search engines and their weakness to carry out the necessary semantic correlations between the query and the real objective of the user's search. Therefore, while the whole process of web search fascinates its users, at the same time, it is indifferent to computers while as machines fail to understand and interpret the information stored in them. The result is the creation of a communication gap between humans and machines related to human’s ability to read and interpret a word, phrase or sentence by associating it with the appropriate conceptual content resulting either in general or specific findings or reaching on reasonable inferences after the juxtaposition of two or more truthful sentences. Computers are unable to proceed to such automated correlations and (reasonable) inferences. For example, when humans read the sentence "O Μιχάλης είναι μεγαλύτερος από τον Αντρέα" (Michael is older than Andreas) easily conclude that "Ο Αντρέας είναι μικρότερος από τον Μιχάλη" (Andreas is younger than Michael). The machine cannot understand the relative age of the two subjects, which is expressed using the comparative degree of the adjective "μεγάλος" (old) and, also to comprehend the syntax used to compare two objects or subjects in Greek language. Furthermore, the phenomenon of multiplicity, i.e. the use of one word to describe different objects, for example, the word "language" that can be interpreted as "anatomical organ", "communication tool", "fish species" etc, and the phenomenon of synonym, i.e. the existence of two or more words that describe the same object or situation but differ in style, performance or expressive significance, for example, the word "clever" and its synonyms "smart", " intelligent", "very clever" etc, make it even more difficult for the search engine to understand the question (query) set by the user. SW based on the creation of a common semantic basis (framework) allows the automation of these functions with minimum human intervention aiming to produce meaning and retrieve the most suitable information. 4
The solution to overcome the structural and functional disadvantages of World Wide Web (WWW) seems to be SW, which is built upon the foundations of the existing web. SW, as the next step, structures (organizes) data correlating them with objects and entities of the real world, and, provides the necessary knowledge for any field of interest (e.g. Medicine, Education etc.). Using SW tools such as metadata and ontologies, computers are equipped with the knowledge of a specific subject area and the appropriate tools (data information, vocabularies, etc.) that enable them to 'read', process, interpret and understand the content of the information stored in them. SW bridges the communication gap between machines and humans based on the idea of having a commonly perceived, between computers and human beings, semantic framework of concepts and terms that refer to real objects and entities in the real world, expressing their mutual conceptual and semantic relations and representing the relevant knowledge. A semantic search engine returns the related to the search objective results after carrying out the essential conceptual correlations, linking concepts and objects with the most suitable semantic content.
Semantic Web is an initiative of the WWW Consortium (W3C) that was inspired
by the creator of WWW, Tim Berners-Lee. Its goal is to structure information and to
improve the current web, placing a semantic layer that allows machines to understand
and process the (human) information effectively [
Concerning the existing text categorization techniques, NLP moves in the field of
linguistic analysis of texts at different levels. These are the morphological, lexical,
syntactic, semantic, discourse and pragmatic level [
Preliminary ideas 1. Concentration of text genres to be included in the list of categories and formulation of general definitions for each genre. Recording the characteristics for each genre that will be used as assessment criteria, and grouping the texts in the following six major categories: Informative, Scientific, Literary, Artistic, Multimodal and Foreign_ Language. 2. Development of the algorithm for categorizing texts into genres and formulation of rules and axioms (assumptions) that lead to reasonable conclusions regarding the distinction of one genre from another (classification criteria). 3. Design and development of the metadata model MeDa13 based on the Dublin Core, for the enrichment of digital recordings (text) information on the (text) genre. MetaDa13 includes the following elements (13): Title, Creator, Subject, Publisher, Rights, Date, Date Modified, Source, Type, Description, Language, Data Writing, Authorial Intention, and, has two applications. The simple application is used to categorize texts into the six major categories of genre (e.g. Informative) and includes only five elements: type, description, language, data writing and authorial intention. The composite application is used for more specified categorizing specific (sub-categorization, e.g. Journalistic) and includes all thirteen elements. 4. Design and development of Textual Genre Ontology (TeGO) for the description of concepts (terms) relating to genres (definitions and attributes) and their relations. 6
The course of the research process in this paper is as follows: 1. Literature review to gather information on how World Wide Web and Semantic
Web, and the web search engines operate (descriptive approach). 2. Conducting an experiment to search and gather information from the internet, for the purposes of an educational activity designed according to the content and learning objectives of a randomly selected topic from the school textbook for teaching Modern Greek language in the third high school grade (G9). The aim is to investigate the suitability of the results returned from the existing web search engines in relation to the content and the learning objectives of the language course (experimental approach). 3. Study of the curriculum of Cyprus Ministry of Education and Culture and the school textbook for teaching Modern Greek language in the third high school grade (G9). Wording of general conclusions concerning the suitability or unsuitability of the results returned from the experiment. 4. Study of the available Semantic Web tools (metadata, ontologies, software agents) and examination of their suitability for the purpose of this work. 5. Study of the existing techniques for categorizing texts in the fields of NLP, IR, IE and Text Mining, and search for the most appropriate for the needs of this study. 6. List all textual genres included in the school textbook and categorization into six categories created for the purposes of this study: Informative, Scientific, Literary, Artistic, Multimodal (texts using many modes i.e. language, sound, picture) and Foreign_ Language. Wording of general definitions and recording of the characteristics for each category and subcategory of genre. Categories and subcategories are defined according to the external and internal characteristics of each genre. 7. Gathering sampling (text) material from the internet in two phases: random and semi-directed selection of texts, using Google search engine and the keyword 'language'. Registration of the results (sample texts). 8. Categorizing a hundred selected results (texts) according to the general definitions and characteristics of genres (step 5). 9. Set two control groups consisted of ten primary school teachers and ten secondary school teachers respectively, to test the validity of definitions and assessment criteria by distributing questionnaires. 10. Gathering the results from the questionnaires and comparison of teachers’ answers between them and with the results of the initial categorization (step 7). Wording of conclusions of the process. 11. Development of an algorithm on the basis of the assessment criteria for automated text categorization in the proposed categories of genres. 12. Design and development of the metadata model MeDa13, consisting of thirteen elements, on the basis of the standard metadata model Dublin Core, in order to describe the relationships between concepts (or terms) and objects (or entities). 13. Design and development of the Textual Genre Ontology (TeGO) to represent the knowledge related to the field "language teaching" and to describe the concepts relating to genres and their relationships. 7
The originality of this work lies in the fact that it is carried around the axis of the content and the stated learning objectives for the language course for the third high school grade (G9), focusing on the textual material used in accordance to the lines of the curriculum of Cyprus Ministry of Education and Culture. The target group is teachers and students that use the internet for gathering information in the context of an educational activity and, the body of texts examined and categorized into genres is mainly written in the Greek language. Regarding the examination of the unsuitability of the results returned from the existing web search engines this study suggests the categorization of internet data into genres using Semantic Web tools, i.e. metadata and ontologies, in order to achieve the learning objectives of language teaching. The importance of this work lies in the proposal for the construction of a semantic search engine that will serve the purposes of language teaching and schooling, in general, as it will return search results after carrying out the necessary semantic associations and, finally, categorizing them into genres, underlining at the same time the need of multiclassification according to the external and internal characteristics of each text.