Reuse: a symbiosis between developers and researchers

Transcription

1 Contemporary Bioethics and the History of the Unborn in Islam ForGe 2016 Reuse: a symbiosis between developers and researchers Alicia González and Tillmann Feige

2 Outline 1 Setup 2 Requirements 3 Available technology 4 Decision

3 Setup The COBHUNI project aims at diversifying our understanding of how pre-natal life is conceptualized in texts of Islamic normativity.

4 Setup

5 Setup

6 Setup 8th now

7 Setup يع ي ة الط ب تا غ الل ال جة مع RTL

8 Setup يع ي ة الط ب تا غ الل ال جة مع RTL م عالجة اللغات الطبيعي ة

9 Setup يع ي ة الط ب تا غ الل ال جة مع RTL م عالجة اللغات الطبيعي ة معالجة اللغاب الطبيعية

10 جة ل عا م Setup م عال ج ة الل غا ت ال ط بي عي ة RTL Root + Derivational morpheme

11 جة ل عا م Setup م عال ج ة الل غا ت ال ط بي عي ة RTL Root + Derivational morpheme + Inflectional morpheme

12 عا م Setup م عال ج ة الل غا ت ال ط بي عي ة RTL ه لج ت Root + Derivational morpheme + Inflectional morpheme + Pronominal clitic

13 Setup م عال ج ة الل غا ت ال ط بي عي ة RTL ه م عال ج ت Root + Derivational morpheme + Inflectional morpheme + Pronominal clitic

14 Setup Source material No. tokens No. types Quran altafsir.com (Quran exegesis) 2,059, ,024 hadith.al-islam.com 1M - Ocred/types texts 23,387 6,954 TOTAL 3M -

15 Setup Great ideas query insert insert processed information Visualization Tool import Annotation Tool COBHUNI Corpus import export Prof. Thomas Eich Universität Hamburg

16 Requirements 1 Full utf8 support and BIDI visualization 2 NLP tools for Arabic 3 Annotation: 3.1 Customised tagsets 3.2 Multitoken 3.3 Subtoken 3.4 Multilayer 3.5 Overlaps 4 Pre-annotated documents. Flexible import and export.

17 Requirements 1 Full utf8 support and BIDI visualization جنين (fetus) 2 NLP tools for Arabic LTR 3 Annotation: 3.1 Customised tagsets 3.2 Multitoken 3.3 Subtoken 3.4 Multilayer 3.5 Overlaps RTL 4 Pre-annotated documents. Flexible import and export.

18 Requirements 1 Full utf8 support and BIDI visualization جنين (fetus) 2 NLP tools for Arabic LTR 3 Annotation: RTL 3.1 Customised tagsets 3.2 Multitoken 3.3 Subtoken 3.4 Multilayer 3.5 Overlaps POS tagging عل ج الجنين بالجراحة المف توحة من أجل... PREP NOUN NOUN PREP NOUN NOUN 4 Pre-annotated documents. Flexible import and export.

19 Requirements 1 Full utf8 support and BIDI visualization جنين (fetus) 2 NLP tools for Arabic LTR 3 Annotation: RTL 3.1 Customised tagsets 3.2 Multitoken 3.3 Subtoken POS tagging عل ج الجنين بالجراحة المف توحة من أجل... PREP NOUN NOUN PREP NOUN NOUN 3.4 Multilayer 3.5 Overlaps semantic KEY TERM 4 Pre-annotated documents. Flexible import and export.

20 Available technology 1 Available toolkits for Arabic processing 2 Not extended support for RTL languages 3 Most software support all annotation requirements 4 Fairly flexible import and export capabilities

21 Available technology 1 Available toolkits for Arabic processing 2 Not extended support for RTL languages 3 Most software don't support all annotation requirements 4 Not flexible import and export capabilities

22 Decision Arabic NLP Component ه م عال ج ت POS PRON NOUN

23 Decision Arabic NLP Component ه م عال ج ت POS PRON NOUN lemma <معالجة> <هو>

24 Annotation Decision Powerful annotation component Import and export flexibility Faulty visualization of Arabic script RTL support not fully integrated

25 Annotation Decision Powerful annotation component Import and export flexibility Faulty visualization of Arabic script RTL support not fully integrated Developers fixed issues very fast

26 Annotation Decision

27 Annotation Decision

28 Annotation Decision

29 Selection Annotation RTL support Meet our annotation requirements Import and export flexibility Robust framework (DKPro Core, Apache UIMA) Excellent support from developers team

30 Selection Annotation RTL support Meet our annotation requirements Import and export flexibility Robust framework (DKPro Core, Apache UIMA) Excellent support from developers team Visualization Extremely powerful searches Import flexibility Portability

31 Conclusion Reuse: using existing technology Arabic is not the worst language to work with Tools can be adapted Communication amongst researchers, users and developers is the key to success

32 Thank you