Apertium has moved from SourceForge to GitHub.
If you have any questions, please come and talk to us on #apertium on irc.freenode.net or contact the GitHub migration team.

Google Summer of Code/Midterm report 2011

From Apertium
Jump to: navigation, search

Contents

[edit] Language pairs

For language pairs, we have two tasks, for some pairs the task was to translate a news article without any diagnostics and evaluate the output. For the other pairs, the task was to create morphological analysers with 80% coverage.

[edit] Turkish → Azerbaijani

See also: apertium-tr-az/dev/midterm
Statistics about input files
-------------------------------------------------------
Number of words in reference: 356
Number of words in test: 364
Number of unknown words (marked with a star) in test: 4
Percentage of unknown words: 1.10 %

Results when removing unknown-word marks (stars)
-------------------------------------------------------
Edit distance: 52
Word error rate (WER): 14.29 %
Number of position-independent word errors: 50
Position-independent word error rate (PER): 13.74 %

Statistics about the translation of unknown words
-------------------------------------------------------
Number of unknown words which were free rides: 1
Percentage of unknown words that were free rides: 25.00 %

[edit] Turkish → Kyrgyz

See also: apertium-tr-ky/dev/midterm
Statistics about input files
-------------------------------------------------------
Number of words in reference: 380
Number of words in test: 371
Number of unknown words (marked with a star) in test: 5
Percentage of unknown words: 1.35 %

Results when removing unknown-word marks (stars)
-------------------------------------------------------
Edit distance: 209
Word error rate (WER): 56.33 %
Number of position-independent word errors: 197
Position-independent word error rate (PER): 53.10 %

Statistics about the translation of unknown words
-------------------------------------------------------
Number of unknown words which were free rides: 0
Percentage of unknown words that were free rides: 0.00 %

[edit] Serbo-Croatian → Macedonian

Statistics about input files
-------------------------------------------------------
Number of words in reference: 456
Number of words in test: 453
Number of unknown words (marked with a star) in test: 2
Percentage of unknown words: 0.44 %

Results when removing unknown-word marks (stars)
-------------------------------------------------------
Edit distance: 275
Word error rate (WER): 60.71 %
Number of position-independent word errors: 207
Position-independent word error rate (PER): 45.70 %

Statistics about the translation of unknown words
-------------------------------------------------------
Number of unknown words which were free rides: 0
Percentage of unknown words that were free rides: 0.00 %

[edit] Slovenian → Spanish

Statistics about input files
-------------------------------------------------------
Number of words in reference: 487
Number of words in test: 457
Number of unknown words (marked with a star) in test: 33
Percentage of unknown words: 7.22 %

Results when removing unknown-word marks (stars)
-------------------------------------------------------
Edit distance: 420
Word error rate (WER): 91.90 %
Number of position-independent word errors: 372
Position-independent word error rate (PER): 81.40 %

Statistics about the translation of unknown words
-------------------------------------------------------
Number of unknown words which were free rides: 11
Percentage of unknown words that were free rides: 33.33 %

[edit] Maltese → Hebrew

[edit] Bengali → English

Personal tools