Building Cendana: a Treebank for Informal Indonesian

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989592%3A15210%2F19%3A73596472" target="_blank" >RIV/61989592:15210/19:73596472 - isvavai.cz</a>

Výsledek na webu

<a href="https://obd.upol.cz/id_publ/333176358" target="_blank" >https://obd.upol.cz/id_publ/333176358</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Building Cendana: a Treebank for Informal Indonesian

Popis výsledku v původním jazyce

This paper introduces Cendana, a treebank for informal Indonesian. The corpus is from a subset of online chat data between customer service staff and customers at Traveloka (traveloka.com), an online travel agency (OTA) from Indonesia that provides airline ticketing and hotel booking services. Lines of conversation text are parsed using the Indonesian Resource Grammar (INDRA) (Moeljadi et al., 2015), a computational grammar for Indonesian in the Head-Driven Phrase Structure Grammar (HPSG) framework (Pollard and Sag, 1994; Sag et al., 2003) and Minimal Recursion Semantics (MRS) (Copestake et al., 2005). The annotation was done using Full Forest TreeBanker (FFTB) (Packard, 2015). Our purpose is to create a treebank, as well as to develop INDRA for informal Indonesian. Testing on 2,000 lexically dense sentences, the coverage is 64.1% and 715 items or 35.8% was treebanked, with correct syntactic parses and semantics. INDRA has been developed by adding 6,741 new lexical items and 22 new rules, especially the ones for informal Indonesian. The treebank data was employed to build a Feature Forest-based Maximum Entropy Model Trainer. Testing against the annotated data, the precision was around 90%. Moreover, we leveraged the treebank data to develop a POS tagger and present benchmark results evaluating the same.

Název v anglickém jazyce

Building Cendana: a Treebank for Informal Indonesian

Popis výsledku anglicky

This paper introduces Cendana, a treebank for informal Indonesian. The corpus is from a subset of online chat data between customer service staff and customers at Traveloka (traveloka.com), an online travel agency (OTA) from Indonesia that provides airline ticketing and hotel booking services. Lines of conversation text are parsed using the Indonesian Resource Grammar (INDRA) (Moeljadi et al., 2015), a computational grammar for Indonesian in the Head-Driven Phrase Structure Grammar (HPSG) framework (Pollard and Sag, 1994; Sag et al., 2003) and Minimal Recursion Semantics (MRS) (Copestake et al., 2005). The annotation was done using Full Forest TreeBanker (FFTB) (Packard, 2015). Our purpose is to create a treebank, as well as to develop INDRA for informal Indonesian. Testing on 2,000 lexically dense sentences, the coverage is 64.1% and 715 items or 35.8% was treebanked, with correct syntactic parses and semantics. INDRA has been developed by adding 6,741 new lexical items and 22 new rules, especially the ones for informal Indonesian. The treebank data was employed to build a Feature Forest-based Maximum Entropy Model Trainer. Testing against the annotated data, the precision was around 90%. Moreover, we leveraged the treebank data to develop a POS tagger and present benchmark results evaluating the same.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

60202 - Specific languages

Projekt

—

Návaznosti

O - Projekt operacniho programu

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of the 33rd Pacific Asia Conference on Language, Information and Computation

ISBN

—

ISSN

2619-7782

e-ISSN

—

Počet stran výsledku

9

Strana od-do

"156–164"

Název nakladatele

Waseda Institute for the Study of Language and Information

Místo vydání

Tokyo

Místo konání akce

Hokodate

Datum konání akce

13. 9. 2019

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—