NLP in Thailand by Asanee Kawtrakul Kasetsart University

NLP in Thailand by Asanee Kawtrakul Kasetsart University 1

Outline u Thailand Language Features u What we do and the problems u The main actors u Research Model and Infrastructure u What do we need more? 2

Thai Language Characteristics u Dialects and Tone u Isolating language u Uninflected u Monosyllabic u No word delimiters u Same form but several functions u Same form but several meanings 3

Thai Language Characteristics u Grammar u Word coverage formation/Recognition u Compound u Proper words vs Sentences name vs Common noun u Loan word (transliterated foreign words) without special orthography 4

Dialects North 18. 8% Central 33. 7% North-East 34. 2% South 13. 3% 5

What do we do? u TEXT PROCESSING u SPEECH PROCESSING u IMAGE PROCESSING 6

Text Processing 7

Text Processing and Problems u Statistical Based Approach u Knowledge Based Approaches u Lack of Legal Corpus u Small Corpus u Lack of Standard(Pos, Semantic Concepts) u Redundancy work 8

Speech Processing u. Speech recognition u. Speech generation 9

Speech Processing and Problems u Recognition u Generation u Not Only Dialect but Tone u Isolated word not Continuous speech u Word Boundary detection 10

Image Processing u Thai optical character recognition u Hand written recognition 11

OCR and Problems u. Isolated Characters 12

The Main Actors u Universities u NECTEC (National Electronic and Computer Technology Center) , Ministry of Science and Technology Environment u SIGNLP 13

The Main Actors u More than 50 experienced researchers (minimum 5 years research( u More than 100 young researchers 14

Financial Supporter u National Electronics and Computer Technology Center (NECTEC) u National research council of Thailand (NRCT) u Kasetsart University Research and Development Institute (KURDI) u Thai Research Foundation (TRF) u etc. 15

Research Model and Infrastructure u Short Term u Long Term u Simple But Work u Collaboration between end users, universities and Funding Agency (including Private sectors) u Robust and very large scale u Enlarge the number of researchers 16

What do we need more? u Share resources (Corpus, Dictionary, Tools, etc(. u Share Experiences and Knowledge u Set Big Umbrella and distribute workload u Establish research network u Partnership 17

Conclusion u Most Thai uses Thai Language u Thai Language Processing has good future in the market IF. …… 18

u. We have more Collaborative work u. NLP Market for 1/2 of 60 millions 19