Tesis doctoral en la redInfoling 9.17 (2014)

Autor/ra:Burgos, Diego A.
Fecha de lectura o defensa:19 de mayo de 2014
Título de la tesis:Towards an image-term co-occurence model for multilingual terminology alignment and cross-language image indexing
Director/a de la tesis:Leo Wanner
Universidad:Universidad Pompeu Fabra
Departamento:Institut Universitari de Lingüística Aplicada
País:España
Descripción de la tesisThis thesis addresses the potential that the relation between terms and images in multilingual specialized documentation has for glossary compilation, terminology alignment, and image indexing. It takes advantage of the recurrent use of these two modes of communication (i.e., text and images) in digital documents to build a bimodal co-occurrence model which aims at dynamically compiling glossaries of a wider coverage. The model relies on the developments of content-based image retrieval (CBIR) and text processing techniques. CBIR is used to make two images from different origin match, and text processing supports term recognition, artifact noun classification, and image-term association. The model aligns one image with its denominating term from collateral text, and then aligns this image with another image of the same artifact from a different document, which also enables the alignment of the two equivalent denominating terms. The ultimate goal of the model is to tackle the limitations and drawbacks of current static terminological repositories by generating bimodal, bilingual glossaries that reflect real usage, even when terms and images may originate from noisy corpora.

Más información: http://www.tdx.cat/handle/10803/145644
Área temática:Lingüística computacional, Lingüística de corpus, Otras especialidades, Terminología
Índice1. Introduction
1.1. The problem
1.1.1. Corpus-based dictionary analysis
1.1.2. Usage-based dictionary analysis
1.2. Thesis proposal: the BC model
1.2.1. The Bimodal Co-occurrence hypothesis
1.2.2. Theoretical Background
1.2.3. Implementing the BC hypothesis
1.3. Assumptions and hypothesis
1.3.1. Assumptions
1.3.2. Hypothesis
1.4. Objectives and delimitation of scope
1.4.1. General Objective
1.4.2. Specific objectives
1.4.3. Delimitation of scope
1.5. Fields of application
1.6. The Structure of the Thesis

2. State-of-the-Art
2.1. Introduction
2.2. Image- and text-based information retrieval systems
2.2.1. Image-term alignment
2.2.2. Content-based image retrieval
2.2.3. Cross-language image retrieval
2.2.4. Domain-specific CLIR
2.3. MWT and artifact noun recognition
2.3.1. Multi-word term recognition
2.3.2. Artifact noun recognition

3. Search Space Characterization
3.1. Outline
3.2. Data macrostructure
3.2.1. General figures
3.2.2. Category delimitation
3.2.3. Category filtering and distribution
3.3. Data microstructure
3.3.1. URL analysis
3.3.2. Analysis of BC components
3.4. Attained goals and drawbacks
3.4.1. Attained goals
3.4.2. Search space drawbacks

4. Nominal, Artifact MWT Recognition
4.1. Introduction
4.2. Nominal MWT recognition
4.2.1. Preliminary considerations
4.2.2. Rule-based MWT recognition
4.2.3. Seed-based bootstrap MWT recognition
4.2.4. Predefined-categories-based MWT recognition
4.2.5. Evaluation of MWT recognition
4.2.6. Results
4.2.7. Discussion
4.3. Artifact MWT recognition
4.3.1. Anchor-based selection
4.3.2. Noun classification

5. Implementation of the BC-Model
5.1. Introduction
5.2. System overview
5.3. System Design
5.3.1. Behavioral Model
5.3.2. Structural/Architectural Model:
5.4. User interface
5.5. System evaluation
5.5.1. Evaluation of the CBIR component
5.5.2. Evaluation results
5.6. Remarks

6. Conclusions
6.1. Hypothesis validation
6.2. Objectives attainment
6.3. Contributions
6.3.1. Characterization of the problem
6.3.2. The BC model
6.3.3. The Search Space
6.3.4. MWT and artifact noun recognition
6.3.5. Prototype
6.4. Limitations
6.4.1. The BC model
6.4.2. MWT and artifact noun recognition
6.4.3. Prototype
6.5. Lines of future research

References
Annexes
Número de págs.:186
Remitente:Diego A. Burgos
Correo-e: <burgosdawfu.edu>
Fecha de publicación en Infoling:11 de septiembre de 2014