R 2017

Corpus Annotation Tool

RYCHLÝ, Pavel

Basic information

Original name

Corpus Annotation Tool

Authors

RYCHLÝ, Pavel (203 Czech Republic, guarantor, belonging to the institution)

Edition

2017

Other information

Language

English

Type of outcome

Software

Field of Study

60200 6.2 Languages and Literature

Country of publisher

Czech Republic

Confidentiality degree

není předmětem státního či obchodního tajemství

References:

RIV identification code

RIV/00216224:14330/17:00096859

Organization unit

Faculty of Informatics

Keywords in English

text corpora; corpus annotation; part of speech tagging

Technical parameters

The tool was used to annotate texts in 6 languages by 16 annotators in total. Czech and Norwegian corpora were annotated mainly for evaluation reasons, there are several PoS annotated (including UD tag set) corpora for both languages. Annotation of 4 Ethiopian languages (Amharic, Oromo, Somali, Tigrinya) was used to build respective PoS taggers which are part of the HaBiT system.
Změněno: 1/6/2017 13:53, doc. Mgr. Pavel Rychlý, Ph.D.

Abstract

V originále

The main goal of this work is an annotation tool for easy and fast production of small annotated corpora for languages without linguistics resources. The tool is optimised for the following priorities: Simple tool for instant usage: The client part of the tool is a web application which works in any browser, it requires small amount of all resources (memory, internet connection, screen size), it could be used even from mobile devices (tablets, phones). Using small standard tag set: If possible, we have used Universal Dependencies (UD) tag set (version 2). It is well documented and used for many different languages. The main description of the tags is directly included in the tool with links to the UD site. Fast navigation: Users can use mouse or keyboard for navigation, only one or two mouse clicks or key strokes are needed for annotation of one token. All sentences are pre-annotated via an builtin adaptive tagger and only incorrect annotation have to be corrected by users. Clean texts: Users can (and are instructed to) reject to annotate a sentence if the sentence is not clear.

Links

7F14047, research and development project
Name: Harvesting big text data for under-resourced languages (Acronym: HaBiT)
Investor: Ministry of Education, Youth and Sports of the CR