Louis Lecailliez

Graduate student in Computer Science (B.S.), Natural Language Processing (M.S.) and Japanese Studies (B.A. & M.A.)

Interest for programming, Japanese, Mandarin, writing systems and mobile applications.

EmailLinkedInBlog

Publications

Construction of a Knowledge Map-based System for Personalized Second Language Learning (2019)
Abstract: A great number of Computer-Assisted Language Learning (CALL) systems have been developed, focusing on one particular language skill and a limited range of user expertise. In this paper, we propose the use of a knowledge graph that contains both lexicon and grammatical data that serve as the basis of learner’s model. By being populated with event log data from external sources, the graph could serve as a user model for client systems in a lifelong learning perspective. The knowledge map is exposed as a service that can be interfaced with existing CALL and learning analytics tools. Beyond the direct display of the knowledge map to the learner that most existing systems provide, we investigate text-based use cases, such as: recommending grammatical concepts with examples tailored to the user's current knowledge, and quiz generation that feeds the system's feedback loop.
A Hypergraph Data Model for Building Multilingual Dictionary Applications (2018)
Abstract.
A non-negligible part of learners of an East Asian language have an interest for another tongue from East Asia that may share some common areal features such as the use of Chinese characters as well as limited word morphology. It makes sense to build for this niche a dictionary application that provides multiple languages in one bundle and allows easy navigation between them and in the lexicon. This paper describes a data model and a generic dictionary application architecture that addresses this use case.

The task of merging lexical resources with vastly differing micro-structures is complex. Even more so when updating it to include new data types or languages after release. In this regard, lexical networks are appealing: they solve the problem by exploding the micro-structure into data nodes and explicitly linking them with edges that can be discovered and traversed automatically. One of these approaches, The Linked Data, is gaining traction in lexicography. It is however plagued with issues within the Resource Description Framework (RDF) that backs it up. Most notably, the lack of three-valent relationships makes it harder than it should to handle the Chinese writing system.

We therefore came up with a simple and consistent hypergraph data model whose main features are: hyperlinks (links of arity greater than two), a flat type system (non-ontological lexical network) and annotations for both node and link instances. We propose a generic software architecture based on this model and illustrate it with a working mobile application. The user interface is constructed from independent components, allowing the display of complex data while increasing further its updatability and maintainability. We use data from the Revised Mandarin Chinese Dictionary of the Ministry of Education of Taiwan and augmented it with open-data Japanese readings to feed the prototype.
Preliminary Thoughts on Issues of Modeling Japanese Dictionaries Using the OntoLex Model (2017)
Abstract.
Recent works aiming at making Linked Data dictionaries make use of the Lemon or OntoLex models. Application to existing dictionaries revealed the need for extensions to the model to properly deal with lexicographic data without loss of information. These works however focus on languages found in Europe, and thus let the issue of Est-Asian lexicography for future exploration. This paper provides a small typology of existing dictionaries in Japan and exposes issues in existing related works that could form the ground of new modules for OntoLex.

Master Dissertations

Towards a Graph-based Model for Japanese Dictionaries (2016)
Creating and analyzing a type system for building and displaying graph-based dictionaries. Including two cases studies using this model.
Approaches for Good Quality Optical Character Recognition (OCR) of a Vietnamese-French Dictionary Containing Nôm (2015)
Keywords: OCR, Vietnamese, chữ nôm, finite state machine.
Automated Method for Complex Chinese Character Search in Electronic Dictionary Based on their Structural Decomposition (2015)
Keywords: Chinese Characters, Graph, IDS, Smartwatch.

Softwares

Japanese-French Dictionary (2012-2014)
Windows Phone 7 & 8 application build with C#. Uses data from dictionnaire-japonais.com.

Academic Degrees

Master of Japanese Studies (2016)
Master of Natural Language Processing (2015)
Bachelor of Japanese Studies (2013)
Bachelor of Computer Science (2012)

Language Certificates

Japanese Language Proficiency Test (JLPT) N2 (2017)
CEFRL equivalent: B2/C1
TOEIC: 950 (2015)
CEFRL equivalent: C1
Hanyu Shuiping Kaoshi (HSK) level 2 (2013)
CEFRL equivalent: A1.1

Awards

Recipient of the scholarship of the Ministry of Education (MEXT) of Japan 🇯🇵 (2018~)
Recipient of the scholarship of the Ministry of Education (MEYS) of the Czech Republic 🇨🇿 (2017-2018)
Chercheur à l'université Masaryk dans le centre de traitemement automatique des langues pendant 10 mois.