Experiment 1: small data

English-Mandarin codeswitch

0.5h

e.g. 1: 你的Android为什么那么快,我的iphone好慢呀。

CN_MS1
EN_MS1

e.g. 2: 明天的presentation你prepare一下吧,一定要弄得很amazing。

CN_MS1
EN_MS1

2h

e.g. 1: 你的Android为什么那么快,我的iphone好慢呀。

CN_MS1
EN_MS1

e.g. 2: 明天的presentation你prepare一下吧,一定要弄得很amazing。

CN_MS1
EN_MS1

8h

e.g. 1: 你的Android为什么那么快,我的iphone好慢呀。

CN_MS1
EN_MS1

e.g. 2: 明天的presentation你prepare一下吧,一定要弄得很amazing。

CN_MS1
EN_MS1

English

0.5h

e.g. 3: My mother is making the bed.

CN_MS1
EN_MS1

e.g. 4: Houses, often three storeys high, were built of stone.

CN_MS1
EN_MS1

2h

e.g. 3: My mother is making the bed.

CN_MS1
EN_MS1

e.g. 4: Houses, often three storeys high, were built of stone.

CN_MS1
EN_MS1

8h

e.g. 3: My mother is making the bed.

CN_MS1
EN_MS1

e.g. 4: Houses, often three storeys high, were built of stone.

CN_MS1
EN_MS1

Mandarin

0.5h

e.g. 5: 明天我们同一时间相约直播间,再见。

CN_MS1
EN_MS1

e.g. 6: 买不了房子,买不了车,旅游也到不了莫斯科。

CN_MS1
EN_MS1

2h

e.g. 5: 明天我们同一时间相约直播间,再见。

CN_MS1
EN_MS1

e.g. 6: 买不了房子,买不了车,旅游也到不了莫斯科。

CN_MS1
EN_MS1

8h

e.g. 5: 明天我们同一时间相约直播间,再见。

CN_MS1
EN_MS1

e.g. 6: 买不了房子,买不了车,旅游也到不了莫斯科。

CN_MS1
EN_MS1

Experiment 2: multi-speaker cross-lingual data

EN_MS1

e.g.1

你的Android为什么那么快,我的iphone好慢呀。

e.g.2

明天的presentation你prepare一下吧,一定要弄得很amazing。

e.g.3

My mother is making the bed.

e.g.4

Houses, often three storeys high, were built of stone.

e.g.5

明天我们同一时间相约直播间,再见。

e.g.6

买不了房子,买不了车,旅游也到不了莫斯科。

CN_MS2

e.g.1

你的Android为什么那么快,我的iphone好慢呀。

e.g.2

明天的presentation你prepare一下吧,一定要弄得很amazing。

e.g.3

My mother is making the bed.

e.g.4

Houses, often three storeys high, were built of stone.

e.g.5

明天我们同一时间相约直播间,再见。

e.g.6

买不了房子,买不了车,旅游也到不了莫斯科。

CN_FS1

e.g.1

你的Android为什么那么快,我的iphone好慢呀。

e.g.2

明天的presentation你prepare一下吧,一定要弄得很amazing。

e.g.3

My mother is making the bed.

e.g.4

Houses, often three storeys high, were built of stone.

e.g.5

明天我们同一时间相约直播间,再见。

e.g.6

买不了房子,买不了车,旅游也到不了莫斯科。