Difference between revisions of "Eastern Burushaski/Final Project"

From LING073
Jump to: navigation, search
(→‎Methodology)
 
(3 intermediate revisions by one other user not shown)
Line 10: Line 10:
 
Both orthographies are somewhat inconsistent or are not completely documented. We have therefore done our best to reconstruct the intentions of Dr. Munshi and Dr. Chitrali, but the orthographies we implement may differ slightly from either original. Small orthographic adjustments can easily be made to the completed transducer.  
 
Both orthographies are somewhat inconsistent or are not completely documented. We have therefore done our best to reconstruct the intentions of Dr. Munshi and Dr. Chitrali, but the orthographies we implement may differ slightly from either original. Small orthographic adjustments can easily be made to the completed transducer.  
 
To build our transducer, we used the Apertium framework.[https://wiki.apertium.org] We encoded the correspondence between the Latin and Arabic characters in our .lexc file, and wrote a number of rules to correct for inconsistencies and edge cases.
 
To build our transducer, we used the Apertium framework.[https://wiki.apertium.org] We encoded the correspondence between the Latin and Arabic characters in our .lexc file, and wrote a number of rules to correct for inconsistencies and edge cases.
 +
We tested the functioning of the transducer by making a .yaml file [https://github.swarthmore.edu/cwillia7/ling073-bsk/blob/master/tests/ARAtests.yaml] with some expected arabic orthographic forms.
  
 
===Code===
 
===Code===
 
The github repository for this project overall can be found at https://github.swarthmore.edu/cwillia7/ling073-bsk.
 
The github repository for this project overall can be found at https://github.swarthmore.edu/cwillia7/ling073-bsk.
  
−
[[Category:Sp19_FinalProjects]]
+
===Evaluation===
 +
Despite the existence of an Arabic orthography, no corpus utilizing it could be found. To evaluate the effectiveness of our transducer, we hand-generated a small corpus of Arabic Burushaski and ran the aq-morftest script[] on the corpus to test for effectiveness and accuracy.
 +
 
 +
Test Results:
 +
 
 +
{| class="wikitable"
 +
|-
 +
! Test # !! pass !! fail !! total
 +
|-
 +
!0.
 +
|46
 +
|8
 +
|54
 +
|-
 +
!1.
 +
|46
 +
|16
 +
|62
 +
|-
 +
!2.
 +
| 0
 +
|12
 +
|12
 +
|-
 +
!3.
 +
| 0
 +
|12
 +
|12
 +
|-
 +
!4.
 +
| 0
 +
|24
 +
|24
 +
|-
 +
!5.
 +
| 0
 +
|24
 +
|24
 +
|-
 +
!6.
 +
| 17
 +
|26
 +
|43
 +
|-
 +
!7.
 +
| 17
 +
|25
 +
|42
 +
|-
 +
!8.
 +
| 12
 +
|17
 +
|29
 +
|-
 +
!9.
 +
| 12
 +
|14
 +
|26
 +
|-
 +
!10.
 +
| 5
 +
|0
 +
|5
 +
|-
 +
!11.
 +
| 5
 +
|32
 +
|37
 +
|-
 +
! total
 +
|  160
 +
|210
 +
|370
 +
 
 +
|}
 +
 
 +
Coverage: 0.43243 or 43.24%
 +
 
 +
 
 +
 
 +
[[Category:Sp19_FinalProjects]][[Category:Eastern_Burushaski]]

Latest revision as of 03:29, 15 May 2019

outlines what you did, links to your code, and provides the results of your evaluation.

Arabic-Latin Orthographic Transducer of Hunza Burushaski

The Burushaski language has no consensus orthography, and literacy is low. We aim to create a transducer between two proposed orthographies which are good candidates for uptake by the Burushaski speaker community. The Latin orthography was proposed by Dr. Sadaf Munshi, curator of the North Texas University Burushaski Language Corpus. [1] This orthography is used in all of her transcriptions. The Arabic orthography was proposed by Rehmat Aziz Chitrali, a linguist and researcher at Khowar Academy Chitral in Pakistan. He released a keyboard layout for PC and Mac on GitHub, [2] which we have implemented for Linux. [3]

Methodology

Both orthographies are somewhat inconsistent or are not completely documented. We have therefore done our best to reconstruct the intentions of Dr. Munshi and Dr. Chitrali, but the orthographies we implement may differ slightly from either original. Small orthographic adjustments can easily be made to the completed transducer. To build our transducer, we used the Apertium framework.[4] We encoded the correspondence between the Latin and Arabic characters in our .lexc file, and wrote a number of rules to correct for inconsistencies and edge cases. We tested the functioning of the transducer by making a .yaml file [5] with some expected arabic orthographic forms.

Code

The github repository for this project overall can be found at https://github.swarthmore.edu/cwillia7/ling073-bsk.

Evaluation

Despite the existence of an Arabic orthography, no corpus utilizing it could be found. To evaluate the effectiveness of our transducer, we hand-generated a small corpus of Arabic Burushaski and ran the aq-morftest script[] on the corpus to test for effectiveness and accuracy.

Test Results:

Test # pass fail total
0. 46 8 54
1. 46 16 62
2. 0 12 12
3. 0 12 12
4. 0 24 24
5. 0 24 24
6. 17 26 43
7. 17 25 42
8. 12 17 29
9. 12 14 26
10. 5 0 5
11. 5 32 37
total 160 210 370

Coverage: 0.43243 or 43.24%

Retrieved from "http://wikis.swarthmore.edu/ling073/index.php?title=Eastern_Burushaski/Final_Project&oldid=10682"