NLP in Thailand

1
NLP in Thailand

• by
• Asanee Kawtrakul
• Kasetsart University

2
Outline

• Thailand Language Features
• What we do and the problems
• The main actors
• Research Model and Infrastructure
• What do we need more?

3
Thai Language Characteristics

• Dialects and Tone
• Isolating language
• Uninflected
• Monosyllabic
• No word delimiters
• Same form but several functions
• Same form but several meanings

4
Thai Language Characteristics

• Grammar coverage
• Word formation/Recognition
• Compound words vs Sentences
• Proper name vs Common noun
• Loan word (transliterated foreign words) without
  special orthography

5
Dialects
North 18.8
North-East 34.2
Central 33.7
South 13.3
6
What do we do?

• TEXT PROCESSING
• SPEECH PROCESSING
• IMAGE PROCESSING

7
Text Processing
8
Text Processing and Problems

• Lack of Legal Corpus
• Small Corpus
• Lack of Standard(Pos, Semantic Concepts)
• Redundancy work

• Statistical Based Approach
• Knowledge Based Approaches

9
Speech Processing

• Speech recognition
• Speech generation

10
Speech Processing and Problems

• Recognition
• Generation

• Not Only Dialect but Tone
• Isolated word not Continuous speech
• Word Boundary detection

11
Image Processing

• Thai optical character recognition
• Hand written recognition

12
OCR and Problems

• Isolated Characters

13
The Main Actors

• Universities
• NECTEC (National Electronic and Computer
  Technology Center) , Ministry of Science and
  Technology Environment
• SIGNLP

14
The Main Actors

• More than 50 experienced researchers (minimum 5
  years research)
• More than 100 young researchers

15
Financial Supporter

• National Electronics and Computer Technology
  Center (NECTEC)
• National research council of Thailand (NRCT)
• Kasetsart University Research and Development
  Institute (KURDI)
• Thai Research Foundation (TRF)
• etc.

16
Research Model and Infrastructure

• Short Term
• Long Term

• Simple But Work
• Collaboration between end users, universities and
  Funding Agency (including Private sectors)
• Robust and very large scale
• Enlarge the number of researchers

17
What do we need more?

• Share resources (Corpus, Dictionary, Tools, etc.)
• Share Experiences and Knowledge
• Set Big Umbrella and distribute workload
• Establish research network
• Partnership

18
Conclusion

• Most Thai uses Thai Language
• Thai Language Processing has good future in the
  market IF.

19

• We have more Collaborative work
• NLP Market for 1/2 of 60 millions