Generative Speech Recognition: A competence model of ASR—Stanislaus Gorky SpecGram Vol CLI, No 2 Contents U.S. Government Linguists in Action—A. Nonymous, B. Nonymous and C. Nonymous

Re-Rating the World’s Languages

Waxaklahun Ubah K’awil
and
José Felipe Hernandez y Fernandez
University of Sprouts, Brussels

#include 

int main()
{
    std::cout << "Hello, world!\n";
}
— C++

In 1991, Dikembe Mutombo and John Thompson rocked the world of linguistics with their iconoclastic article “Rating the World’s Languages”. With the bold statement that “the canard of linguistic equality has to be abandoned by anyone wanting to be a realistic student of language,” Mutombo and Thompson opened the door to what should have been a new era in applied linguistics.

But that didn’t happen. Though
PLEASE DO ,1 <- #13
DO ,1 SUB #1 <- #238
DO ,1 SUB #2 <- #112
DO ,1 SUB #3 <- #112
DO ,1 SUB #4 <- #0
DO ,1 SUB #5 <- #64
DO ,1 SUB #6 <- #238
DO ,1 SUB #7 <- #26

DO ,1 SUB #8 <- #248
DO ,1 SUB #9 <- #168
DO ,1 SUB #10 <- #24
DO ,1 SUB #11 <- #16
DO ,1 SUB #12 <- #158
DO ,1 SUB #13 <- #52
PLEASE READ OUT ,1
PLEASE GIVE UP
— INTERCAL
it would indubitably make an interesting thesis topic in the field of History of Science or Psychology of Academia, the reasons for this failure to break through are in broad scope obvious, and in detail irrelevant. The two present authors have long harbored a clandestine fascination with this under-explored topic. Having each recently achieved tenure, we can now continue Mutombo and Thompson’s important work with significantly less fear of the kind of retaliation that seems to have destroyed their careers.

As the original authors discussed, the empirical scoring is ironclad, though the choice of which parameters to score is perhaps more subjective. To the original six parameters of simplicity of expression, clarity of expression, range of expressible content, ease of acquisition, writing system quality, and euphoniousness, we have added nine additional important parameters: poeticality (as suggested by the original authors), practical utility, conservativeness, linguistic imperialism, sexiness, coolness, cussing/insulting capability, obstreperousness, and idiosyncrasy.

Unlike the original six parameters, not all of the new parameters’ scores are proportional to the degree to which the trait described is present in the language. Several (conservativeness, linguistic imperialism, and idiosyncrasy) have more-or-less U-shaped scoring curves. Obstreperousness is inversely proportional to the score assigned. These parameters have been normalized to the linear scoring scale. As an example, under linguistic imperialism, both Dyirbal and English score very poorly: Dyirbal for being insufficiently imperialistic, English for being overly so.

While there may have been some gains in the general level of mathematical sophistication in the field, our very objective and highly scientific methodology is probably still too mathematical for the average linguist to understand. Our results are presented in the table below. Extrema for each language are in red. Extrema for each parameter are in bold.


Language SofE CofE RofE EofA WSQ Euph. Poet. Prac. Cons. Imp. Sexy Cool Cuss Obst. Idio. Avg.
Anglo-Saxon 5.99 6.15 5.31 4.35 5.26 7.43 9.77 1.23 0.73 0.04 6.25 7.83 7.66 5.27 7.29 5.37
Arabic 5.57 7.29 8.42 4.28 7.33 2.59 8.26 8.74 5.67 4.56 7.93 6.32 8.62 4.32 5.32 6.35
ASL 8.23 6.94 7.43 5.34 0.79 8.73 9.02 5.32 7.43 3.20 4.86 8.71 9.54 7.65 3.56 6.45
Bella Coola 5.26 3.37 4.41 5.11 6.82 0.48 3.24 0.32 0.44 0.46 1.12 1.89 0.04 2.34 3.45 2.58
Dyirbal 4.32 3.12 4.40 5.77 5.48 4.89 6.93 0.01 0.54 0.03 3.25 6.25 6.32 2.04 6.43 3.99
English 4.01 5.79 9.13 4.33 2.73 5.01 4.84 9.87 1.23 0.01 4.53 5.01 8.81 4.22 5.43 5.00
Esperanto 7.54 6.57 6.54 8.87 4.34 4.33 5.33 2.77 8.54 8.87 2.12 3.23 2.54 7.65 3.93 5.54
French 6.59 6.34 7.89 6.02 6.93 6.87 9.82 8.93 1.32 6.76 8.43 5.64 8.93 3.21 7.89 6.77
German 3.22 5.93 8.23 5.43 5.32 2.34 2.85 8.86 7.43 1.23 0.79 4.25 5.43 8.32 5.21 4.99
Gothic 4.32 6.23 5.62 4.76 4.35 3.42 6.82 1.21 0.17 0.36 6.87 8.99 9.96 5.44 6.44 5.00
Guaraní 4.86 4.23 7.03 6.72 6.03 5.82 5.83 4.11 8.93 2.23 4.77 6.88 8.32 6.23 8.43 6.03
Hindi 6.42 7.42 8.43 6.77 8.43 5.42 8.33 9.30 8.64 3.40 7.44 7.29 6.43 7.54 5.63 7.13
Hittite 3.98 2.12 5.71 3.35 1.52 4.10 1.12 0.45 0.02 0.11 0.37 2.45 1.59 8.43 8.54 2.92
Inuktitut 2.38 3.46 4.34 2.43 1.23 3.29 4.35 2.13 1.43 0.16 2.35 7.33 2.34 2.84 3.64 2.91
Italian 7.11 6.21 7.54 7.95 7.31 6.48 9.12 8.64 8.63 8.33 8.79 6.78 9.33 8.65 8.64 7.97
Japanese 2.11 2.67 8.62 2.23 3.86 6.77 9.83 8.65 6.82 5.43 5.34 7.62 2.31 3.62 5.72 5.44
Klingon 4.23 4.35 3.54 1.76 2.22 0.19 5.64 2.54 2.34 0.02 2.58 5.43 9.32 1.87 7.60 3.58
Latin 6.12 7.98 8.72 6.54 6.09 5.67 8.43 1.65 0.34 1.33 1.45 4.82 4.33 8.25 7.52 5.28
Lojban 0.22 9.99 8.23 0.12 7.01 3.45 0.01 0.10 3.20 2.56 1.34 5.73 2.24 3.23 2.33 3.32
Mandarin 8.32 6.23 7.87 3.84 3.44 4.82 5.28 9.52 7.43 1.89 1.43 5.99 6.75 2.10 3.25 5.21
Spanish 7.31 6.50 7.77 8.10 7.43 6.11 9.23 9.37 8.76 7.65 7.98 6.54 9.42 9.87 8.54 8.04
Swahili 5.03 7.03 6.99 7.78 6.11 4.12 5.99 4.02 7.35 5.43 5.42 6.87 5.43 4.73 5.66 5.86
Tagalog 4.25 4.99 5.82 5.83 5.97 4.03 6.48 5.23 5.63 7.45 4.52 5.22 8.45 8.32 7.42 5.97
Volapük 8.01 6.45 2.12 6.88 6.57 0.97 0.13 0.02 0.01 0.05 0.03 0.53 1.34 7.23 3.88 2.94
Zulu 3.98 6.43 5.49 3.82 7.27 4.77 6.56 4.35 2.43 6.54 3.42 8.42 6.28 2.34 6.57 5.24
σ = 2.64     μ = 5.20     Σ = 2078.25     ε << ι     α < Ω     ∀x∃y s.t. x ⊥ y

The collection of languages represented in this sample is somewhat irregular. We have included the
puts "Hello, world!"
— Ruby
original authors’ list of ten, along with a number of languages commonly taught in high-school and college (to assist students in choosing an appropriate language to study), and a small number of additional constructed languages at the request of our friends in the ConLang community.

We provide some interpretation of the data above as a guide to others seeking to properly use this resource. While Spanish has the highest overall score, English, with the highest practical utility score, is still the best general-purpose second-language choice throughout the world. Students who are native speakers of English should look to Spanish to fulfill their language requirementsnot only is it a wonderful, useful language, it is the least
Test: proc options(main) reorder;
  put skip edit('Hello, world!') (a);
end Test;
— PL/I
obstreperous among commonly-taught languages: no uvular nasals, no contrastive aspiration, tractable phonemic inventory, and a generally non-ridiculous grammar (though there were deductions for that whole “a mi me gusta ... debacle).

Those thinking they should run out and learn Italianthe sexiest language on the listare advised to keep in mind that most of the people who speak Italian are used to being wooed by native speakers of Italian, and will not likely be impressed by anything less. However, second year Italian is very good for wooing first year Italian students.

To the Klingons out there who are dissatisfied with the results of this very objective and highly scientific methodology, we would apologize, though it would do no good. Rather we refer them to the scores for Gothic and say instead 𐌺𐌿𐌽𐌹 𐌽𐌰𐌳𐍂𐌴!”

Generative Speech Recognition: A competence model of ASR—Stanislaus Gorky
U.S. Government Linguists in Action—A. Nonymous, B. Nonymous and C. Nonymous
SpecGram Vol CLI, No 2 Contents