Small Language Models Also Work With Small Vocabularies
Collection
Models and evaluation data for our 2025 COLING paper (https://aclanthology.org/2025.coling-main.404/).
•
10 items
•
Updated
text
stringlengths 0
8.68k
|
|---|
jæ
|
kɑmpaʊnd'ɛs θɪŋz laɪk sɪks naɪn sɪks naɪn
|
naɪn twɛlv
|
jæ
|
ðæt sɔɹt ʌv stʌf
|
æt ðʌ tɑp
|
ðæt'ɛs ðʌ tɑp nʌmbɜ˞
|
ʌnd kɑmpaʊnd ɪz θɹi fɔɹ
|
ðæts kɑmp ðæts kɑmpaʊnd
|
naʊ kɑmpaʊnd ɪz ɔlweɪz sɪks
|
oʊ
|
sɪmpʌl ɪz dʒʌst ɔɹdʌnɛɹi θɹi bits ʌ bɑɹ
|
jæ
|
fɔɹ bits ɪn ʌ bɑɹ tu bits ɪn ʌ bɑɹ
|
ɪt'ɛs ðʌ θɪŋk ʌv kɑmpaʊnd æz biɪŋ ðʌ wʌn ðæts mɔɹ dɪfʌkʌlt tu θɪŋk ʌbaʊt
|
sɪks eɪt taɪmz naɪn eɪt taɪmz
|
jæ
|
ɹaɪt
|
stɪl livz θɹi tu θɹi ɔɹ fɔɹ bits ɪn ʌ bɑɹ bʌt ðeɪ'ɹeɪ kɑmpaʊnd taɪmz
|
ðeɪ'ɹeɪ dɑtʌd bits
|
ðæt'ɛs ɹɪli ɔl ðæt kɑmpaʊnd minz
|
ɪt minz dɑtʌd bits
|
wʌt wʊd twɛlv twɛlv fɔɹ
|
θɪŋk ʌbaʊt twɛlv fɔɹ kɔs ðæts ʌ kɑmpaʊnd taɪm
|
ɪt'ɛs ɡɑt twɛlv æt ðʌ tɑp
|
ɪt meɪ hɛlp tu lʊk æt ɪt
|
kɔs ju kænt θɪŋk ʌbaʊt ɪt stɹeɪt ʌweɪ
|
θɪŋk twɛlv ʌnd ɪtəl bi twɛlv wʌt
|
dɑtʌd
|
noʊ
|
wʌt dʌz fɔɹ stænd fɔɹ ɪn æn ɔɹdʌnɛɹi taɪm sɪɡnʌtʃɜ˞
|
oʊ
|
ɹaɪt soʊ
|
soʊ
|
ɪf ðɛɹ ɑɹ twɛlv ʌv ðoʊz
|
twɛlv
|
wʌn tu θɹi fɔɹ faɪv sɪks sɛvʌn eɪt naɪn tɛn ɪlɛvʌn twɛlv
|
θɪŋk ʌv ðɛm ɪn ɡɹups ʌv θɹiz ʌnd ðæt ɡɪvz ju fɔɹ dɑtʌd
|
kweɪvɜ˞z
|
noʊ
|
ju'ɹeɪ jʊr ɡoʊɪŋ smɔl
|
jɛs
|
ʌnd ju'vi ɡɑt tu ɡoʊ ðʌ ʌðɜ˞ weɪ
|
oʊ jæ
|
ɪt'ɛs
|
ɪf ðɛɹ ɑɹ ɪf ðɛɹ ɑɹ twɛlv ʌnd itʃ ʌv ðoʊz ɑɹ ɪn ɡɹups ʌv θɹi
|
dɑtʌd
|
soʊ ju'vi ɡɑt fɔɹ dɑtʌd
|
soʊ twɛlv aɪ min ɪf ju kæn ɹɪmɛmbɜ˞ ðɛm twɛlv stændz fɔɹ fɔɹ bits ɪn ʌ bɑɹ ɛniweɪ
|
jæ
|
ʌnd ɪf ju dʒʌst θɪŋk wʌt twɛlv ɑɹ ɡoʊɪŋ tu bi dɪvaɪdɪd ʌp ɪntu
|
m
|
ɪts ɡoʊɪŋ tu dɑtʌd
|
kɔs ju noʊ ɪts ɡoʊɪŋ tu bi dɑtʌd sʌmθɪŋ ɔɹ ʌðɜ˞
|
ʌnd ɪf ju wɜ˞ tu si ɜ˞ æn ɛkstɹækt ʌv mjuzɪk fɔɹ ɪɡzæmpʌl hiɹ'ɛs ʌ twɛlv eɪt wʌn
|
ðɪs wʌn æktʃuʌli tɛlz ju ɪts ɪn twɛlv eɪt taɪm
|
ʌnd ɪts æskɪŋ ju tu pʊt ɪn ðʌ bɑɹ laɪnz
|
wɛɹ wʊd ju pʊt ðoʊz bɑɹ laɪnz du ju θɪŋk
|
ɪt wʊd æl aɪ θɪŋk ɪt pɹɑbʌbli ɔlsoʊ æsks ju tu tu ɡɹup ðɛm pɹɑpɜ˞li
|
bʌt dʒʌst fɔɹ ðʌ moʊmʌnt wi wʌn'ti wɜ˞i ʌbaʊt ðʌ ɡɹupɪŋ æspɛkt
|
aɪ θɪŋk ʌsum ɛvɹiθɪŋ stɑɹts ɑn ðʌ fɜ˞st bit ʌv ðʌ bɑɹ æz wɛl
|
oʊkeɪ
|
ɪt' tɛl ju ɪf ɪt dʌzənt
|
eɪt stændz fɔɹ
|
kweɪvɜ˞z
|
ɹaɪt
|
wʌn tu θɹi fɔɹ faɪv sɪks sɛvʌn eɪt naɪn tɛn ɪlɛvʌn twɛlv
|
soʊ ɑbviʌsli ðʌ fɜ˞st wʌn wʊd bi ðɛɹ
|
ʌnd pɹɑbʌbli ɪf ɪts ɡoʊɪŋ tu bi ʌ hoʊl wʌn
|
wʌn tu θɹi fɔɹ faɪv sɪks sɛvʌn eɪt naɪn tɛn ɪlɛvʌn twɛlv
|
soʊ ɪn fækt ðɛɹ ɑɹ dʒʌst tu bɑɹz
|
naʊ haʊ wʊd ju æktʃuʌli ɡɹup ðoʊz noʊts
|
ɪtəd bi ɡɹups ʌv θɹi ʌɡɛn wʊdənt ɪt
|
ɪn ɡɹups ʌv θɹi
|
ðoʊz tu'vi ɡɑt tu bi ɡɹupt tʌɡɛðɜ˞
|
wʌn tu θɹi fɔɹ
|
ɜ˞ ɪtəd bi sɔɹt ʌv hæf ʌv ðæt
|
hæf ʌv ðʌ
|
aɪ doʊnt noʊ haʊ ju du ðæt
|
kæn wi ɡoʊ bækwɜ˞dz
|
wʌn tu θɹi fɔɹ faɪv
|
ju kæn oʊnli du ɪt ɪf ju mɪs aʊt ðæt
|
wʌn tu θɹi
|
wʌn tu θɹi
|
ʌnd ðæt livz ju wɪð ʌ ʌ lɪtʌl bɪt
|
ɹaɪt
|
jɛs
|
ɪt æktʃuʌli ɪkspleɪnz ɪt
|
ɪɡzæktli wʌt ðeɪ wɑnt ju tu du
|
ɜ˞ wɛɹ ju'vi ɡɑt ʌ noʊt ðæt ɑbviʌsli ɡoʊz oʊvɜ˞ ʌ bit æz ɪn ðæt wʌn ðɛɹ
|
jæ
|
jʊr jʊr ɡoʊɪŋ tu hæv tu ɹaɪt tu kweɪvɜ˞z ʌnd taɪ ðɛm
|
ɪz ðæt wʌt ju sɛd
|
jæ
|
ɪt maɪt hæv bɪn wʌt ju sɛd
|
soʊ ðoʊz tu oʊkeɪ
|
ðæt wʌn wɪl bi dʒɔɪnd tu ʌ kweɪvɜ˞ bit soʊ ðæt jul hæv ʌ ɡɹup ʌv θɹi
|
End of preview. Expand
in Data Studio
No dataset card yet
- Downloads last month
- 6
Size of downloaded dataset files:
171 MB
Size of the auto-converted Parquet files:
97.8 MB
Number of rows:
3,915,418