Shanghainese and Mandarin Chinese

Resources for machine translation between Shanghainese and Mandarin Chinese

• Resources developed for Shanghainese:
  • Link to resources developed for Shanghainese: [1]
• Resources developed for Mandarin Chinese:

External Resources

• Link to repository language pair:[2]
• Link to Shanghainese repository transducer:[3]
• Link to Mandarin Chinese repository transducer:[4]
• Link to Shanghainese to Mandarin Tatoeba Text: [5]

Initial Evaluation==

Sentences

• (1) wuu: 我爱侬。
  • zho: 我爱你。
  • eng: I love you.
  • lexical transfer output: ^我<prn><pers><p1><sg>/我<prn><pers><p1><sg>$ ^爱<vblex>/爱<vblex>$ ^侬<prn><pers><p2><sg>/你<prn><pers><p2><sg>$ 。 ^.<sent>/.<sent>$
  • full translation output: 我 爱 你 。

• (2) wuu: 汏手间勒勒阿里耷？
  • zho: 洗手间在哪里？
  • eng: Where is the restroom?
  • lexical transfer output: 洗手间 在 哪里 ？
  • full translation output: ^汏手间<n>/洗手间<n>$ ^勒勒<pr>/在<pr>$ ^阿里耷<adv>/哪里<adv>$？^.<sent>/.<sent>

• (3) wuu: 我勿晓得。
  • zho: 我不知道。
  • eng: I don't know.
  • lexical transfer output: ^我<prn><pers><p1><sg>/我<prn><pers><p1><sg>$ ^勿<adv>/不<adv>$ ^晓得<vblex>/知道<vblex>$。^.<sent>/.<sent>
  • full translation output: 我 不 知道 。

• (4) wuu: 伊啥人啊？
  • zho: 他是谁？
  • eng: Who are they? (sing.)
  • lexical transfer output: ^伊<prn><pers><p3><sg>/他<prn><pers><p3><sg><m>/她<prn><pers><p3><sg><f>/它<prn><pers><p3><sg><nt>$ ^啥人<prn><itg>/谁<prn>$ ^啊<part>/啊<qst>$？^.<sent>/.<sent>$
  • full translation output: 他 #谁 啊 ？

• (5) wuu: 侬要吃茶还是吃咖啡啊？
  • zho: 你想喝茶还是喝咖啡？
  • eng: Do you want to drink tea or coffee?
  • lexical transfer output: ^侬<prn><pers><p2><sg>/你<prn><pers><p2><sg>$ ^要<vaux>/要<vblex>$ ^吃<vblex>/喝<vblex>/吃<vblex>$ ^茶<n>/茶<n>$ ^还是<cnjcoo>/还是<cnjcoo>$ ^吃<vblex>/喝<vblex>/吃<vblex>$ ^咖啡<n>/咖啡<n>$ ^啊<part>/啊<qst>$？^.<sent>/.<sent>
  • full translation output: 你 要 喝 茶 #还是 喝 咖啡 啊 ？

• (6) wuu: 侬要走啦？
  • zho: 你要走了吗？
  • eng: Are you leaving?
  • lexical transfer output: ^侬<prn><pers><p2><sg>/你<prn><pers><p2><sg>$ ^要<vaux>/要<vblex>$ ^走<vblex>/走<vblex>$ ^啦<part>/啦<qst>$ ^.<sent>/.<sent>
  • full translation output: 你 要 走 啦 ？

• (7) wuu: 勿要紧个！
  • zho: 没关系。
  • eng: Don't worry.
  • lexical transfer output: ^勿要紧<ij>/没关系<ij>$ ^个<part>/的<pr>$！^.<sent>/.<sent>
  • full translation output: #没关系 的 ！

• (8) wuu: 我一只也勿要。
  • zho: 我一个也不要。
  • eng: I don't want even one.
  • lexical transfer output: ^我<prn><pers><p1><sg>/我<prn><pers><p1><sg>$ ^一<num>/一<num>$ ^只<cls>/个<num>/只<num>$ ^也<adv>/也<adv>$ ^勿<adv>/不<adv>$ ^要<vaux>/要<vblex>$。^.<sent>/.<sent>
  • full translation output: 我 一 个 也 不 要 。

• (9) wuu: 今朝天好。
  • zho: 今天好天气。
  • eng: The weather's nice today.
  • lexical transfer output: ^今朝<n>/今天<n>$ ^天<n>/天<n>$ ^好<adj>/好<adj>$。^.<sent>/.<sent>
  • full translation output: 今天 天 好 。

• (10) wuu: 一、二、三、四、五、六、七、八、九、十。
  • zho: 一、二、三、四、五、六、七、八、九、十。
  • eng: One, two, three, four, five, six, seven, right, nine, ten.
  • lexical transfer output: ^一<num>/一<num>$、^二<num>/二<num>$、^三<num>/三<num>$、^四<num>/四<num>$、^五<num>/五<num>$、^六<num>/六<num>$、^七<num>/七<num>$、^八<num>/八<num>$、^九<num>/九<num>$、^十<num>/十<num>$。^.<sent>/.<sent>
  • full translation output: 一、二、三、四、五、六、七、八、九、十。

Additions

• 2 new lexical selection rules
• 3 new disambiguation rules
• 2 new transfer rules

Final Evaluation

• Shanghainese Transducer
  • Precision: Coverage hfst is still broken
  • Recall: Coverage hfst is still broken
  • Number of words in Large Corpus: Coverage hfst is still broken
  • Number of stems in Transducer: Coverage hfst is still broken

• wuu → zho
  • WER: apertium-eval-translator-line does not work on space-less orthography
  • PER: apertium-eval-translator-line does not work on space-less orthography
  • Proportion of stems translated correctly in Longer Corpus: apertium-eval-translator-line does not work on space-less orthography
  • Trimmed coverage over longer and large corpora: apertium-eval-translator-line does not work on space-less orthography
  • The number of tokens in longer and large corpora: apertium-eval-translator-line does not work on space-less orthography