49fb6c3cbeaaf1b0e33489ebe16d38a5.ppt
- Количество слайдов: 39
LTe. L - Language Technology for e. Learning Paola Monachesi, Lothar Lemnitzer, Kiril Simov, Alex Killing, Diane Evans, Cristina Vertan
LT 4 e. L - Language Technology for e. Learning -1 • EU-IST-FP 6 Project 2005 - 2008 • The LT 4 e. L project uses multilingual language technology tools and semantic web techniques for improving the retrieval of learning material. The developed technology will facilitate personalized access to knowledge within learning management systems and support decentralisation and co-operation in content management.
LT 4 e. L - Language Technology for e. Learning -2 • • • Start date: ; 1 December 2005 Duration: 30 months EU finacing: 1. 5 milion Euro Type project: STREP IST-4 Coordination: Paola Monachesi (Utrecht university) • Contact for information: Paola. Monachesi@let. uu. nl
LT 4 e. L - Partners • • • Utrecht University (UU), The Netherlands University of Hamburg (UHH), Germany University “Al. I. Cuza” of Iasi (UAIC), Romania University of Lisbon (FFCUL), Portugal Charles University Prague (CUP), Czech Republic Institute for Parallel Processing, Bulgarian Academy of Sciences (IPP-BAS), Bulgaria University of Tübingen (UTU), Germany Institute of Computer Science, Polish Academy of Sciences (ICSPAS), Poland Zürich University of Applied Sciences Winterthur (ZHW), Switzerland University of Malta (UOM), Malta University of Cologne (UCO), Germany Open University (OU), United Kingdom
LT 4 e. L - Languages • • • Bulgarian Czech Dutch German Maltese Polish Portugese Romanian English
LT 4 e. L -Aims • Improve retrieval of learning material • Facilitate construction of user specific courses • Improve creation of personalized content • Support decentralization of content management • Allow for multilingual retrieval of content
LT 4 e. L- Objectives -1 • Scientific and Technological Objectives – Creation of an archive of learning objects and linguistic resources – Integration of language technology resources in e. Learning – Integration of semantic Knowledge in e. Learning – Integration of functionalities in open source LMS – Validation of enhanced LMS
LT 4 e. L- Objectives -2 • Political objectives – – – Support multilinguality Knowledge transfer Awareness raising Exploitation of resources Facilitate access to education
LT 4 e. L - Workpackages • ▪ WP 1 - Setting the scene - WP leader: University AI. I. Cuza of • • • Iasi ▪ WP 2 - Semi-automatic metadata generation driven by Language Technology resources - WP leader: University of T歟 ingen ▪ WP 4 - Integration of the new functionalities in the ILIAS Learning Management System - WP leader: University of Cologne ▪ WP 3 - Enhancing e. Learning with semantic knowledge - WP leader: IPP, Bulgarian Academy of Science ▪ WP 5 - Validation of new functionalities in the ILIAS Learning Management System - WP leader: Open University (England) ▪ Multilinguality - Leader: University Hamburg
User Profile LMS Documents SCORM Pseudo-Struct. Basic XML LING. PROCESSOR Lemmatizer, POS, Partial Parser CROSSLINGUAL RETRIEVAL CONVERTOR 2 Documents SCORM Pseudo-Struct Documents HTML Ontology Glossary CONVERTOR 1 EN CZ MT PL GE DT RO Documents User (PDF, DOC, HTML, SCORM, XML) EN Lexikon EN CZ Lexicon Lexikon PT Lexikon RO Lexikon PL Lexikon GE Lexicon Metadata (Keywords) Ling. Annot XML BG PT MT BG Lexikon DT REPOSITORY
Collection of Learning materials • collection of the learning material (uploads & updates at http: //consilr. info. uaic. ro/uploads_lt 4 el/ - passwd protected) • ▪ IST domains for the LOs: ▪ – 1. Use of computers in education, with sub-domains: ▪ • 1. 1 Teaching academic skills, with sub-domains: ▪ • 1. 1. 1 Academic skills▪ • 1. 1. 2 Relevant computer skills for the above tasks (MS Word, Excel, Power Point, La. Tex, Web pages, XML)▪ • 1. 1. 3 Basic computer skills (use of computer for beginners) (chats, email, Intenet)▪ • 1. 2 e-Learning, e-Marketing▪ • 1. 3 The I*Teach document (Leonardo project, http: //i-teach. fmi. uni-sofia. bg/) • 1. 4 Impact of use of computers in society • 1. 5 Studies about use of computers in schools / high schools▪ • 1. 6 Impact of e-Learning on education▪ – 2. Calimera documents (parallel corpus developped in the Calimera FP 5 project, http: //www. calimera. org/ )
Collection of learning materials and linguistic tools • normalization of the learning material▪ commonly agreed DTD and convertors from html/txt to basic XML format▪ Inventarization and classification of existing tools (http: //consilr. info. uaic. ro/uploads_lt 4 el/tools/all. php? ) relevant to: – ▪ the integration of language technology resources in e. Learning (WP 2) – the integration of semantic knowledge (WP 3)▪ • Inventarization and classification of existing language resources▪ corpora and frequencies listsハ: http: //consilr. info. uaic. ro/uploads_lt 4 el/menu/all. php • ▪ lexica: http: //www. let. uu. nl/lt 4 el/wiki/index. php/Lexica_Joint_Table
User Profile LMS Documents SCORM Pseudo-Struct. Basic XML LING. PROCESSOR Lemmatizer, POS, Partial Parser CROSSLINGUAL RETRIEVAL CONVERTOR 2 Documents SCORM Pseudo-Struct Documents HTML Ontology Glossary CONVERTOR 1 EN CZ MT PL GE DT RO Documents User (PDF, DOC, HTML, SCORM, XML) EN Lexikon EN CZ Lexicon Lexikon PT Lexikon RO Lexikon PL Lexikon GE Lexicon Metadata (Keywords) Ling. Annot XML BG PT MT BG Lexikon DT REPOSITORY
WP 2: Integration of language resources in e. Learning Aims of the Workpackage • supporting authors in the generation of metadata for Los • improving keyword-driven search for LOs • supporting the development of glossaries for learning material
Metadata • metadata are essential to make LOs visible for larger groups of users • authors are reluctant or not experienced enough to supply them • NLP tools are supposed to help them in that task • the project uses the LOM metadata schema as a blueprint
Task 1: Identification of keywords • Good keywords have a typical, non random distribution in and across Los • Keywords tend to appear mor often at certain places in texts (headings etc. ) • Keywords are often highlighted / emphasised by authors
Modelling Keywordiness • Residual Inverse document frequency used to model inter text distribution of KW • Term burstiness used to model intra text distribution of KW • Knowledge of text structure used to identify salient regions (e, g, headings) • Layout features of texts used to identify emphasised words and weight them higher
Challenges • Treating multi word keywords (suffix arrays will be used to identify n-gramsof arbitrary length) • Assigning a combined weight which takes into account all the aforementioned factors • Evaluation – manually assigned keywords will be used to measure precision and recall of key word extractor against – inter annotator agreement will be tested to get a upper bound for keyword assignment task
Task 2: Identification of definitory contexts • • This task makes use of the linguistic annotation of Los The approach is empirical Identification of definitory contexts is language specific Workflow – Definitory contexts will be searched and marked in LOs (manually) – Local grammars will be drafted on the basis of these examples – The linguistic annotation will be used for these grammars – The grammars will be applied to new Los Integration – The tools will be integrated as additional functions to the ILIAS LMS – The tools will also be available for integration in other LMS – We consider making the tools available as web services
User Profile LMS Documents SCORM Pseudo-Struct. Basic XML LING. PROCESSOR Lemmatizer, POS, Partial Parser CROSSLINGUAL RETRIEVAL CONVERTOR 2 Documents SCORM Pseudo-Struct Documents HTML Ontology Glossary CONVERTOR 1 EN CZ MT PL GE DT RO Documents User (PDF, DOC, HTML, SCORM, XML) EN Lexikon EN CZ Lexicon Lexikon PT Lexikon RO Lexikon PL Lexikon GE Lexicon Metadata (Keywords) Ling. Annot XML BG PT MT BG Lexikon DT REPOSITORY
WP 3: ontology based crosslingual retrieval • Generic approach • For each domain : – Using computers for beginners – Impact of e. Learning in Society • we built a domain ontology • For consistency reasons we consider also an upper ontology (DOLCE) • Lexical material in all 9 languages is mapped on the ontology and on the upper ontology • According to : • types of relations in the ontology and • Uses cases • Similarity (predefined ontological chunks) we define some search patterns for the user interface
Domain Ontology • First built starting with English documents • Concepts are based on : – Extracted keywords in WP 2 and – Glossaries for the given domains • Concepts have generic names with parts in English (for readability reasons) e. g: C 11_editors • For each concept we provide labels with explanation of the concept in english and ideally in all other languages • Types of relations: – Is_a – Part_of – Here we need some informations about what people are searching • The ontology will be encoded in OWL- DL
Mapping multilingual resources on the domain ontology -1 • Trivial for words having exact a correspondent in the ontology • Problems appear when: 1. One word in a language sub-sums two or more concepts in the ontology 2. One word in a language sub-sums two or more concepts in an ontology but only in relations with some other concepts 3. One word has a much restrictive meaning not present in the ontology
Mapping multilingual resources on the domain ontology -2 • Solution to 1: – Express the lexical items in OWL-DL expressions: disjunction, conjunctions of classes (give example) • Solution to 2: – Express the lexical items in OWL-DL using together with operations on classes also relations between the involved concepts • Solution to 3: – Insert new concept in the ontology
Ontology enrichment • If one word cannot be mapped directly on the ontology look if a similar meaning can be retrieved in some other languages. • If this seems to be not an isolated case insert the new concept in the ontology. • In any case assign to each concept a label indicated the languages in which this concept is lexicalised • The insertion of a new concept will be done with FACT or RACER
Linking lexicon, domain ontology and upper ontology • Domain ontology concepts will be mapped on the upper ontology. • This will ensure that all important properties of main classes are considered. • Not relevant senses of some lexical items could be also mapped directly on the upper ontology
User Profile LMS Documents SCORM Pseudo-Struct. Basic XML LING. PROCESSOR Lemmatizer, POS, Partial Parser CROSSLINGUAL RETRIEVAL CONVERTOR 2 Documents SCORM Pseudo-Struct Documents HTML Ontology Glossary CONVERTOR 1 EN CZ MT PL GE DT RO Documents User (PDF, DOC, HTML, SCORM, XML) EN Lexikon EN CZ Lexicon Lexikon PT Lexikon RO Lexikon PL Lexikon GE Lexicon Metadata (Keywords) Ling. Annot XML BG PT MT BG Lexikon DT REPOSITORY
WP 4: Tasks • Integration of LT 4 e. L Tools for semi-automated metadata generation, definitory context extraction and ontology supported extended data retrieval into a learning management system (prototype based on ILIAS LMS) • Developing and providing documentation for a standard-technology-based interface between the language technology tools and learning management systems
WP 4: Objective - Fostering Re-Use of LT-Tools LMS 1 (ILIAS) LMS 2 (e. g. Moodle) LMS 3 (e. g. ATutor) LT-Interface XML-RPC / Web Service Language Technology Tools • Simple-as-possible, well-documented and standards-based interface
WP 4: Using LT-Tools in Learning Managements Systems • Possible Use Case Scenarios – – Author annotates learning object with keywords Author generates glossary for learning object Tutor searches for learning objects Learner searches for learning material in multiple languages – Learner browses through learning material with ontology based information
WP 4: Example ILIAS-LT-Tools Use Case Scenario: Keyword Generation 1. Author adds new learning object to the LMS (e. g. HTML file) 2. ILIAS displays a form including input fields for title, language and filename 3. Author enters title and language, selects a local. pdf file and hits Upload File 4. ILIAS+LTTools display the (LOM) metadata input form, including a list of auto-generated, suggested keywords 5. Author selects some of the suggested keywords, enters some new keywords and hits Save 6. ILIAS saves the metadata
User Profile LMS Documents SCORM Pseudo-Struct. Basic XML LING. PROCESSOR Lemmatizer, POS, Partial Parser CROSSLINGUAL RETRIEVAL CONVERTOR 2 Documents SCORM Pseudo-Struct Documents HTML Ontology Glossary CONVERTOR 1 EN CZ MT PL GE DT RO Documents User (PDF, DOC, HTML, SCORM, XML) EN Lexikon EN CZ Lexicon Lexikon PT Lexikon RO Lexikon PL Lexikon GE Lexicon Metadata (Keywords) Ling. Annot XML BG PT MT BG Lexikon DT REPOSITORY
WP 5: Validation of enhanced LMS. • Challenge is to answer these questions: – How does this compare with what can already be done with existing systems? – What added value is there? – What is the educational / pedagogic value of these functionalities? • Problem is to evaluate the functionality and separate from issues of usability or unfamiliarity with the LMS platform. How can we expect users to identify any benefit?
How can we expect users to identify any benefit? • Present them with tasks to complete using LMS • With no project functionality • With project functionality – – Partial Full • Identify potential users – Course Creators – Content Authors or Providers – Teachers – Sudents • studying in their own language • studying in a second language
Create outline User Scenarios • We define scenarios, in this context, as – a story focused on a user or group of users which provides information on • the nature of the users, • the goals they wish to achieve and • the context in which the activities will take place. – They are written in ordinary language, and are therefore understandable to various stakeholders, including users. – They may also contain different degrees of detail.
Example Outline Scenario for a student • A student has just completed studying in English a topic on 'The use of computers in Schools'. • They are interested in finding more information on the use of this topic within their subject domain. • Their first language is German • Suggested search approaches might be: – standard search as available within the LMS not using any of LT 4 e. L functionality. – add in the lexicon – add in the multi-linguality – add in the ontology • Users will be given guidance / familiarisation activities in using each of the tools beforehand. ▪ • User Scenarios are under development for all the identified users. • Each scenario will focus on one or more of the new functionalities dependent on the roles of a particular user.
Possible Teachers /Course creators tasks • • • Add new content to new course structure Search for existing content and add to course structure Add new content to existing course Add supplementary content (could be another language) Modify existing content Create new content and make available to the system.
Feedback from Users • Sessions will be used to gather some initial feedback using – individual interviews – group plenary – questionnaires
Project plan • Preparatory work in place (May 06). • Development functionalities complete (November 2006). • Integration functionalities in LMS complete (May 2007) • First cycle integration functionalities in LMS and their validationcomplete (November 2007) • Second cycle integration functionalities in LMS and their validationcomplete (May 2008)