Kperalt1/Final Project - Revision history

Kperalt1 at 16:33, 24 July 2025

2025-07-24T16:33:17Z

Kperalt1: /* Statistics about input files */

2025-05-16T02:48:18Z

‎Statistics about input files

Kperalt1: /* Statistics about input files */

2025-05-16T02:48:06Z

‎Statistics about input files

Kperalt1: /* External Resources */

2025-05-16T02:47:32Z

‎External Resources

Kperalt1: /* External Resources */

2025-05-16T02:45:39Z

‎External Resources

Kperalt1: /* Statistics about input files */

2025-05-16T02:36:50Z

‎Statistics about input files

Kperalt1: /* Statistics about input files */

2025-05-16T02:33:54Z

‎Statistics about input files

Kperalt1: /* Statistics about input files */

2025-05-16T02:33:39Z

‎Statistics about input files

Kperalt1: /* Statistics about input files */

2025-05-16T02:33:12Z

‎Statistics about input files

Kperalt1: /* Final Evaluation */

2025-05-16T01:24:59Z

‎Final Evaluation

← Older revision		Revision as of 16:33, 24 July 2025
Line 272:		Line 272:
	Based on the examples shown on this wiki, the system performs reasonably well on simple sentence structures and produces intelligible translations. With continued refinement, especially in grammatical handling and structural rules, the system will become more defined. Most importantly, this project demonstrates that it is possible to build translation systems even for languages with different grammatical structures.		Based on the examples shown on this wiki, the system performs reasonably well on simple sentence structures and produces intelligible translations. With continued refinement, especially in grammatical handling and structural rules, the system will become more defined. Most importantly, this project demonstrates that it is possible to build translation systems even for languages with different grammatical structures.

−	[[Category:Shanghainese]]	+	[[Category:Sp25_FinalProjects]] [[Category:Shanghainese]]

@@ Line 239: / Line 239: @@
 ==Statistics about input files==
-*Evaluation Corpus:
+Evaluation Corpus:
 **Number of words in reference: 8105
 **Number of words in test: 9263

@@ Line 252: / Line 252: @@
 **Position-independent word error rate (PER): 112.36 %
-*Development Corpus:
+Development Corpus:
 **Number of words in reference: 80
 **Number of words in test: 74

@@ Line 5: / Line 5: @@
 ===External Resources===
 *Link to repository language pair:[https://github.swarthmore.edu/Ling073-sp25/ling073-wuu-eng.git]
 *Link to Shanghainese repository transducer:[https://github.swarthmore.edu/Ling073-sp25/ling073-wuu.git]
 *Link to English repository transducer:[https://github.com/apertium/apertium-eng.git]
 *Link to Shanghainese to English Text: [https://tatoeba.org/en/sentences/search?from=wuu&query=&to=eng]
 ==Shanghainese <code>wuu</code> → English <code>eng</code> Sentence Evaluation==

@@ Line 9: / Line 9: @@
 *Link to English repository transducer:[https://github.com/apertium/apertium-eng.git]
 *Link to Shanghainese to English Text: [https://tatoeba.org/en/sentences/search?from=wuu&query=&to=eng]
 ==Shanghainese <code>wuu</code> → English <code>eng</code> Sentence Evaluation==

@@ Line 250: / Line 250: @@
 **Number of position-independent correct words: 156
 **Position-independent word error rate (PER): 112.36 %
 ==Final Evaluation==

@@ Line 244: / Line 244: @@
 **Percentage of unknown words: 50.70 %
-**Results when removing unknown-word marks (stars)
+Results when removing unknown-word marks (stars)
 -------------------------------------------------------
 **Edit distance: 9777

← Older revision		Revision as of 01:24, 16 May 2025
Line 251:		Line 251:

	==Final Evaluation==		==Final Evaluation==
		+	The goal of this project was to develop a simple yet intelligible translation system from Shanghainese to English. While there is still a long way to go, the current progress demonstrates that this goal is achievable. Currently, the system produces literal or direct translations. As a result, the English output may lack grammatical function words such as "the" or "a," which are typically absent in Shanghainese. A key future improvement would be refining the translations to produce more grammatically correct English, making the system more robust. Another issue to address is the handling of verb tags, which currently affect the system’s accuracy.

		+	As for the evaluation, the current WER and PER are relatively high. However, there's been an great improvement, over a 10% increase in performance, after expanding the development corpus from 10 to over 40 sentences. The high error rates can likely be attributed to incomplete verb tagging and perhaps the limited amount of structural transfer rules.
		+
		+	Based on the examples shown on this wiki, the system performs reasonably well on simple sentence structures and produces intelligible translations. With continued refinement, especially in grammatical handling and structural rules, the system will become more defined. Most importantly, this project demonstrates that it is possible to build translation systems even for languages with different grammatical structures.

	[[Category:Shanghainese]]		[[Category:Shanghainese]]