Una Zhang
Unaz.siteFor this project, I compiled dumpling names from two sources: historic cookbooks from the Internet Archive and contemporary recipe names scraped from recipe websites, online cookbooks, and food blogs.
1st pass
The first dataset consisted almost entirely of historic cookbook data. The model generated "Apple" repeatedly because Apple Dumpling appeared 489 times out of 1,709 entries (29% of the dataset). This reflects the dominance of apple dumplings in English cookbooks before the 20th century. More broadly, the word dumpling in these sources referred mainly to British dumplings rather than the East Asian dumplings the term is more commonly associated with today.
- sample 1: Apple
- sample 2: Apple
- sample 3: Baked Apple
- sample 4: Apple
- sample 5: Apple
- sample 6: Lemmon
- sample 7: Suet
- sample 8: Apple
- sample 9: Baked Apple
- sample 10: Apple
- sample 11: Baked Apple
- sample 12: Apple
- sample 13: Peach
- sample 14: Apple
- sample 15: Apple
- sample 16: Apple
- sample 17: Apple
- sample 18: Apple
- sample 19: Apple
- sample 20: Apple
2nd pass
I added around 100 contemporary dumpling names and capped each historic recipe at ten repetitions. While this reduced the dominance of Apple Dumpling, the outputs were still fairly generic because the historic names remained overrepresented.
- sample 1: Chicken
- sample 2: Steam
- sample 3: Suet
- sample 4: Gammon
- sample 5: Pigeon
- sample 6: Damson
- sample 7: Cheese
- sample 8: Damson
- sample 9: Marrow
- sample 10: Potato
- sample 11: Currant
- sample 12: Baked Apple
- sample 13: Steam
- sample 14: Steamed
- sample 15: Damassa
- sample 16: Pigeon
- sample 17: Lemon
- sample 18: Hard
- sample 19: Drop
- sample 20: Peach
- sample 21: Plain Suet
- sample 22: Ball
- sample 23: Suet
3rd pass
The final dataset contained 542 unique dumpling names: 162 historic names from the Internet Archive and 380 contemporary names from eight scraped recipe sources. Compared to the historic dataset, the contemporary names contain more descriptors: introduced greater variety in ingredients, cooking methods, transliterated names such as momo and har gow, and expressive spelling online. As a result, the model generated outputs like "Fried Beeeeeef Dum..." and "Steamed Pork..." while also producing hybrids such as "Homo," where it recombined unfamiliar character sequences and naming conventions.
- sample 1: Boiled Chile Dum
- sample 2: Steamed Pork Dum
- sample 3: Siocot Dumpling
- sample 4: Dumpling Stean R
- sample 5: Steamed Pork ark
- sample 6: Steamed Beeefor
- sample 7: Perry Dumpling
- sample 8: Chicken and Stea
- sample 9: Pork and Chicken
- sample 10: Fried Beeeeeef D
- sample 11: Lemon Dumpling
- sample 12: Beef and Cantear
- sample 13: Coubang Dumpling
- sample 14: Pert Dumpling
- sample 15: Rashe Pet Dumpli
- sample 16: Chinese Dumpling
- sample 17: Steamed Dumpling
- sample 18: Momamed Dumpling
- sample 19: Homemade Pork Du
- sample 20: Chicken Dumpling
- sample 21: Pork and Chives
- sample 22: Fried Pota Dumpl
- sample 23: Chicken Dumpling
- sample 24: Pejater Pork and
- sample 25: Steamed Dumpling
- sample 26: Steamed Poth Dum
- sample 27: Banton Dumpling
- sample 28: Boiled Basary Du
- sample 29: Steamed Purrkl D
- sample 30: Homo
- sample 31: Boiled Great Dum