Research
-
Hybrid Machine Translation Model
-
Hierarchical PB Translation Model
-
Multiple System Combination
Projects
-
NGL: Next Generation Localisation (07/2008- )
Language barriers constitute a formidable obstacle to the free flow of information, products and services in an increasingly globalised economy and information society.
“Localisation” refers to the process of adapting digital content to culture, locale and linguistic environments at high quality and speed. Localisation is a key enabling, value-adding, multiplier component of the global software and content distribution industry. Localisation seeks to overcome language barriers.
The objective is to produce substantial advances in the basic and applied research underpinning the design, implementation and evaluation of the blueprints for the Next Generation Localisation Factory.
The mission is to revolutionise localisation via breakthroughs in automation, composition and integration, focusing on:
-
Integrated machine translation technology,
-
Speech-based interfaces and more personalised speech output,
-
Multilingual digital content management for personalised multilingual content access and delivery,
-
Localisation workflows and system integration.
For more details about it, see http://www.cngl.ie.
-
-
CASIA: A SMT Platform based on Multiple System Combination Strategy (January 2007 - June 2008)
The multi-engine platform is developed for research on SMT models and algorithms. Meanwhile, it also could be used as a transferring platform for application development.
This Platform includes 4 core modules: 1) Automatic Preprocessing Module; 2) Alignment Post-processing & Models Generation Module; 3) Decoding & MER Training Module; 4) Multiple System Combination & Post-processing Module.
More details could be found in my Ph.D Thesis.
CASIA participated several international and demestic Chinese-to-English machine translation evaluation campaigns:
-
Won 13th place in NIST MT Eval 2008 (about 21 teams)
-
Won 3rd place in SSMT Eval 2007 (about 10 teams)
-
-
Matrix: a Phrase-based SMT system (March 2006 - June 2008)
Matrix is a typical phrase-based SMT system with improved re-ordering model.
-
Intelligent Temperature Control Module for Injection Moulding Machine (December 2004 - August 2005)
This is an electrical engineering project which requires a fast and accurate respond when the machine starts. Moreover, in the process of production, the error of temperature must be controlled between <-0.5,+0.5>. This is the first project at my first Ph.D year.