Voice recognition for English and Mandarin typing

http://languagelog.ldc.upenn.edu/nll/?p=27675

In all tech considered (8/24/16), Arrti Shahani has an article titled "Voice Recognition Software Finally Beats Humans At Typing, Study Finds".

Turns out voice recognition software has improved to the point where it is significantly faster and more accurate at producing text on a mobile device than we are at typing on its keyboard. That's according to a new study by Stanford University, the University of Washington and Baidu, the Chinese Internet giant. The study ran tests in English and Mandarin Chinese.

Baidu chief scientist Andrew Ng says this should not feel like defeat. "Humanity was never designed to communicate by using our fingers to poke at a tiny little keyboard on a mobile phone. Speech has always been a much more natural way for humans to communicate with each other," he says.

The study found that speaking short phrases into an iPhone was three times faster than typing them on an iPhone.

The Stanford University-University of Washington-Baidu team didn't test query skills. They zoomed in on voice recognition software's ability to type the spoken words. In English, they found the software's error rate was 20.4 percent lower than humans typing on a keyboard; and in Mandarin Chinese, it was 63.4 percent lower.

Judging from these figures, it would seem that typists of English must typically make a lot fewer errors than typists of Mandarin.  Considering the complexity of the Chinese script and the complicated nature of human-machine interfaces for entering Chinese text, this is not surprising.  Since the software used in this experiment was a Baidu program called Deep Speech 2, there should not have been a bias in favor of English.

What does this presage?  Greater reliance on speech recognition for typing and increased levels of character amnesia for handwriting.

Who knew?

http://languagelog.ldc.upenn.edu/nll/?p=27707

… that there is an apparently serious and respectable institution called the Center for Advanced Hindsight ("With our ‘Advanced Hindsight’ superpower we develop, apply and share behavioral insights").

This suggests a large space of available institutional names: there could be Institutes (or Centers or Laboratories) for (the Advanced Study of) many interesting things: Higher-Order Cognitive Bias; Unprecedented Errors; Failing Presuppositions; Novel Fallacies; …

 

Annals of parsing

http://languagelog.ldc.upenn.edu/nll/?p=27691

Two of the hardest problems in English-language parsing are prepositional phrase attachment and scope of conjunction. For PP attachment, the problem is to figure out how a phrase-final prepositional phrase relates to the rest of the sentence — the classic example is "I saw a man in the park with a telescope". For conjunction scope, the problem is to figure out just what phrases an instance of and is being used to combine.

The title of a recent article offers some lovely examples of the problems that these ambiguities can cause: Suresh Naidu and Noam Yuchtman, "Back to the future? Lessons on inequality, labour markets, and conflict from the Gilded Age, for the present", VOX 8/23/2016.  The second phrase includes three ambiguous prepositions (on, from, and for) and one conjunction (and), and has more syntactically-valid interpretations than you're likely to be able to imagine unless you're familiar with the problems of automatic parsing.

On the basis of general linguistic and real-world knowledge, I interpret this title as promising lessons for the present time, asserting that these lessons are derived from a comparison with various phenomena in the period known as the Gilded Age, and specifying that those phenomena are inequality, labor markets, and conflict. Reading the article confirms this interpretation.

The relevant linguistic knowledge includes the fact that "lessons for X" and "lessons from Y" are both common idioms. It's reasonable to expect a modern stochastic parser to know this — but of course  for and from are promiscuous modifiers, willing to hook up with almost any noun, verb, or adjective at all.

The relevant real-world knowledge includes the fact that socio-economic inequality is a frequently discussed feature of the Gilded Age and also of the present time, that such inequality is related to labour markets, and that it may lead to conflict.  Modern parsers don't think about things at this level — current directions of progress, like distributional semantics, don't really help much in cases like this one.

To illustrate some of the ways that parsing can go wrong,  I'll give the (variously wrong) results returned by three available on-line parsers. And let me note in passing that all of these parsers report their results in the general framework adopted by the Penn Treebank project, which flourished due to a sort of treaty that emerged from a summit meeting of computational linguists held 25 years ago (Ezra Black et al., "A Procedure for Quantitatively Comparing the Syntactic Coverage of English Grammars", HLT 1991). This approach has been so widely used in "treebank" projects, across many languages and types of text, that it makes sense in my opinion to teach it more widely, at least to linguistics students if not to a broader audience.

The (rather old-fashioned) Link Grammar parser messes up the worst:

(S (NP (NP Lessons)
       (PP on
           (NP inequality)))
   ,
   (VP labour
       (NP (NP markets)
           , and
           (NP conflict))
       (PP from
           (NP the Gilded Age ,
               (PP for
                   (NP the present)
                   .)))))

You can visualize the induced structure more clearly in a tree diagram:

In other words, the article is about "Lessons on X, labour Y"; X is "inequality"; and Y is "markets and conflict from the Gilded Age", more specifically the "Gilded Age for the present".

The Berkeley parser is a bit better:

(ROOT
  (NP
    (NP (NNS Lessons))
    (PP (IN on)
      (NP
        (NP (NN inequality))
        (, ,)
        (NP (JJ labour) (NNS markets))
        (, ,)
        (CC and)
        (NP
          (NP (NN conflict))
          (PP (IN from)
            (NP (DT the) (NNP Gilded) (NNP Age)))
          (, ,)
          (PP (IN for)
            (NP (DT the) (NN present))))))
    (. .)))

Note that this version adds the "part of speech" tags immediately above the individual words:

Now we're talking about  lessons on X, where X="inequality, labour markets, and Y", and Y="conflict from the Gilded Age for the present". This is coherent but certainly not correct.

The Stanford parser almost gets it right:

(ROOT
  (NP
    (NP (NNS Lessons))
    (PP (IN on)
      (NP
        (NP (NN inequality))
        (, ,)
        (NP (NN labour) (NNS markets))
        (, ,)
        (CC and)
        (NP
          (NP (NN conflict))
          (PP (IN from)
            (NP (DT the) (NNP Gilded) (NNP Age))))
        (, ,)))
    (PP (IN for)
      (NP (DT the) (NN present)))
    (. .)))

Here "from the Gilded Age" modifies only "conflict", and has no direct relationship to "lessons".

I think the correct analysis has the three PPs as all parallel dependents of "lessons"

  • "on inequality, labour markets, and conflict"
  • "from the Gilded Age"
  • "for the present"

with the conjunction joining the three NPs

  • "inequality"
  • "labour markets"
  • "conflict"
(ROOT
  (NP
    (NP (NNS Lessons))
    (PP (IN on)
      (NP
          (NP (NN inequality))
          (, ,)
          (NP (NN labour) (NNS markets))
          (, ,)
          (CC and)
          (NP (NN conflict))))
     (PP (IN from)
            (NP (DT the) (NNP Gilded) (NNP Age)))
    (, ,)
    (PP (IN for)
      (NP (DT the) (NN present))))
  (. .))

Some people might prefer to add additional structure, say by binding "on inequality, labour markets, and conflict" more closely to "lessons", or even stacking up the PP modifiers recursively:

I'll leave this question to the syntacticians.

But whatever exactly the right answer is, it's not what the three online parsers came up with. There may be some systems out there than can do better on this particular example — but PP attachment and conjunction scope in English remain hard problems for computational linguistics.

Unattended luggage

http://languagelog.ldc.upenn.edu/nll/?p=27654

On her way back from Cornwall in April, Janet (Geok Hoon) Williams saw this sign, put up by Great Western Railway, at the train station:

The Chinese reads:

zài ānquánxìng fāngmiàn de lìyì qǐng bùyào líkāi nǐ de xínglǐ wú rén kānguǎn
在安全性方面的利益 請不要離開你的行李無人看管
"For the benefit of the aspect of security, don't depart from your luggage with no one to look after it."

Although it is fairly easy to get the basic gist of the Chinese sentence, it sounds awkward.

It's interesting that all six of the foreign languages on the sign seem to be working from this English sentence which is not on the sign:  "In the interest of security, please do not leave your luggage unattended".

I don't know about the other languages on the sign, but it seems difficult to me to express "unattended" in Chinese.  One could say, in a rather forced, unidiomatic manner, wèi jīng zhàokàn 未經照看 ("not having been taken care of"), but I think that Chinese would prefer to state the matter positively as zhàokàn hǎo 照看好 ("well taken care of"), or some such.

In English, "unattended" is an adjective, but the Chinese phrase "wú rén kānguǎn 無人看管" ("no one / nobody looking after [it]") has the subject expressed.  If you want to turn this phrase into an adjective, you have to add the magical marker "de 的" at the end — "which has no one looking after it" — and then you can modify "luggage" with it, but that would still sound unnatural.

Here are three alternative Chinese translations of the standard, but unwritten, English announcement:

chū yú ānquán kǎolǜ, qǐng bùyào ràng nǐ de xínglǐ wú rén kānguǎn
出於安全考慮,請不要讓你的行李無人看管
("…please don't let your luggage [be in a condition of having] no one to look after it")

chū yú ānquán kǎolǜ, qǐng bùyào líkāi nín de xínglǐ
出於安全考慮,請不要離開您的行李
("…please don't go away from your luggage")

chū yú ānquán kǎolǜ, qǐng kānguǎn hǎo nín de xínglǐ
出於安全考慮,請看管好您的行李
("…please watch over your luggage well")

[Thanks to Fanyi Cheng and Yixue Yang]

Gestures of death

http://languagelog.ldc.upenn.edu/nll/?p=27678

Shaun King, "North Carolina police kill unarmed deaf man using sign language", New York Daily News 8/22/2016:

This is as bad as it gets.

A North Carolina state trooper shot and killed 29-year-old Daniel Harris — who was not only unarmed, but deaf — just feet from his home, over a speeding violation. According to early reports from neighbors who witnessed the shooting this past Thursday night, Harris was shot and killed "almost immediately" after exiting his vehicle.

He appeared to be trying to communicate with the officer via sign language.

Screenshot:

Local news coverage is here (warning: autoplay).

Update — after looking at more than 200 examples of the form VERB NOUNPHRASE1 using NOUNPHRASE2, I could find only a handful where the using-phrase is a modifier of NOUNPHRASE1, or part of an untensed clause [NOUNPHRASE1 using NOUNPHRASE2], rather than an instrumental adjunct modifying the VERB:

… turning his back to avoid a septuagenarian using an aluminum walker …
… you need a tablet using an x86-based processor …
… he overheard his father using those words once …

So the crash blossom misparsing is overwhelmingly the stochastic-parsing choice, never mind what makes sense.

Update #2 — The described event is horrific, and underlines what seems to be a serious problem with the training, attitudes and behavior of police forces in the U.S. I apologize for seemed to trivialize the incident by focusing on the parsing of a misleading headline.

 

The Festival are clear

http://languagelog.ldc.upenn.edu/nll/?p=27676

One of the rare syntactic dialect differences between British and American English (there really aren't many) concerns verb agreement in present-tense clauses: British English strongly favors plural agreement with any singular subject noun phrase that denotes a collectivity of individuals rather than a unitary individual. And the extent to which it favors that plural agreement is likely to raise eyebrows with speakers of American English. This example, for example, from an email about a lecture at the Edinburgh International Book Festival:

The Festival are very clear that if you arrive after the start of the lecture you will not be admitted.

That's not a syntactic slip. The Festival  has been treated as denoting not the unitary institution but the collection of officials who constitute it or run it. Hence it gets plural agreement, as in Manchester United are playing well or Apple are releasing a new phone model. American usage strongly favors singular agreement with subjects that look singular, unless something really clashes with that (I take it that even for Americans The Festival are really beating each other up over the funding snafu is much better than ?The Festival is really beating each other up over the funding snafu). But there are many other subtleties. Never think that verb agreement with subjects is a simple example of a syntactic rule, let alone one that universally works in the same way.

Modal logic of traffic signs

http://languagelog.ldc.upenn.edu/nll/?p=27668

Sent in by Michael Robinson:

I saw this traffic sign in Toledo, Ohio. Luckily I wasn't driving a truck, or I would have had no idea what I was allowed to do. Since we were in a car, we figured U-turns must be OK. Because we were heading to a place that sold coffee, and nothing must stand between us and our morning latte.


I'm not sure that Michael was in the clear. Working out his concern in more detail, the full deontic-logic table for this sign is

Other Vehicles Trucks
Left Turn  ?  ?
U Turn  ?  NO
No Turn  ?  ?

The sign could mean that U-turns by trucks are forbidden — and by Gricean implication, the other five conditions are allowed. Or it could mean that trucks are not allowed to make any turns, and no vehicles are allowed to make U-turns.

Or …

Anyhow, according to Ohio Laws and Rules 4511.36 Rules for turns at intersections:

(C) The department of transportation and local authorities in their respective jurisdictions may cause markers, buttons, or signs to be placed within or adjacent to intersections and thereby require and direct that a different course from that specified in this section be traveled by vehicles, streetcars, or trackless trolleys, turning at an intersection, and when markers, buttons, or signs are so placed, no operator of a vehicle, streetcar, or trackless trolley shall turn such vehicle, streetcar, or trackless trolley at an intersection other than as directed and required by such markers, buttons, or signs.

So there's that.

And Internet search turns up this troubling item on amazon.com:

Perhaps a broader version of this traffic-sign semantics puzzle will play a role in Dan Brown's next semiotic thriller.

 

 

Bilingual Spanish-Chinese street signs

http://languagelog.ldc.upenn.edu/nll/?p=27579

Germán Renedo recently noticed that the government has installed bilingual street signs in the Belgrano neighborhood of Buenos Aires, where Chinatown is located. The signs transcribe the sounds of the Spanish words rather than translate their meanings.

For instance, Arribeños ("those who came from the highlands and live on the coast"):

Ālǐbèiniǔsī jiē 阿里贝紐斯街 (jiē 街 means "street")

And here is Juramento ("oath; sacrament; vow"):

Hūlāménduō jiē 呼拉门多街

I leave it to readers who are fluent in Spanish to determine whether the transcriptions are faithful renderings of the Spanish sounds.

These are the previous (Spanish-only) versions:

Aside from the accuracy of the transcriptions, I'd also like to ask whether readers think that transcriptions or translations are more useful in such circumstances (I have my own view on this).

Ask Language Log: why is "inch" a family relationship in Korean?

http://languagelog.ldc.upenn.edu/nll/?p=27637

Katie Odhner asks:

I have lately been teaching myself Korean and have become quite interested in Sino-Korean vocabulary. Recently two words in particular caught my attention: samchon 삼촌 ("paternal uncle"), from Chinese s ān cùn 三寸 ("three inches"), and sachon 사촌 ("cousin"), from Chinese sì cùn 四寸 ("four inches"). I wondered how "three inches" and "four inches" could turn into family members. According to one website I found, chon 寸 can refer to "degree (of kinship)", which makes some sense. But when I looked on ctext.org (Chinese Text Project), I couldn't find classical Chinese examples of this usage, so I'm thinking maybe it's a Korean invention.

Have you ever encountered cùn 寸 ("inch") in Classical Chinese to refer to degree of kinship? Do you think it's a Korean invention? And does "third degree of kinship" for uncle and "fourth degree of kinship" for cousin have any roots that you can think of in the Confucian tradition, or is that also a native Korean concept?

In Chinese, whether modern or premodern, cùn 寸 ("inch") does not refer to a degree of kinship.  It has the following meanings:  an ancient unit of length (one tenth of a chǐ 尺 [roughly a foot; in the Western Han over two thousand years ago, it was 0.231 meter]; an inch; short; small, tiny).  As for metaphorical and allusory usages of cùn 寸 ("inch") in early Chinese texts:

cùnxīn 寸心 (lit., "inch heart") refers to the feelings, inner thoughts

sāncùn 三寸 (lit., "three inches") refers to the tongue

sì cùn 四寸 (lit., "four inches") is an allusion to a very large pearl in early medieval times

From Haewon Cho:

Samchon 三寸 ("paternal uncle") and sachon 四寸 ("cousin") are both Korean Chinese words that are used in Korea only, and Koreans are known to have used this chon 촌 system as early as the Koryo Dynasty ( 918–1392). The Korean dictionary (NAVER) lists two meanings of chon : 1) degree of kinship by blood; 2) a measure of length, which is about 3.03 cm. chon 촌  is a synonym of chi 치 , which is 1/10 of ca 자 (Korean foot). 촌 and 자 are said to be pure Korean.

Some say chon 촌 derived from mati 마디 (節), which denotes "a bamboo joint; node" or a "finger segment (between two knuckles)." According to Ugyo.net (Ugyo 유교 [儒敎] is Confucianism in Korean), a family tree is like a bamboo growing one joint by one joint, and people figure out how close/distant a kinship relation is by counting these joints (chon 촌).

We add one 촌 for any vertical relationship (i.e., father-son) and two 촌s for any horizontal relationships (i.e., brother and sister). So basically, even numbered 촌s means people are of the same generation or they have grandchild-grandparent relationships. As 촌 refers to any blood related relationship, husband and wife are 0 촌 (無寸).

From Bob Ramsey:

You'll find the word associated with the character 寸 in a lot of Korean reference works. But I think most knowledgeable people say that it's an example of what the Japanese call ateji (当て字)–in other words, a false association. I've said before that I thought the word–and the concept–was a Korean invention, and I see no reason to change my mind about that.

Thus chon 촌 is both a measure of actual distance and of relational distance. The same word / character with the same pronunciation is used to indicate physical distance and is also used metaphorically for the idea of kinship relationship distance.  Korean dictionaries treat chon 촌 as Sino-Korean, and apparently etymologically it is. But the usage is not Chinese, so if chon 촌 really is Sino-Korean, Koreans long ago repurposed it in a way that is uniquely their own.

The Female Brain movie

http://languagelog.ldc.upenn.edu/nll/?p=27641

Silas Lesnick, "An ensemble cast has come together for Whitney Cummings’ The Female Brain movie", comingsoon.net 8/17/2016:

Black Bicycle Entertainment has today announced the ensemble cast for their upcoming The Female Brain movie, which marks the directorial debut of Whitney Cummings. Cummings herself will also star in the film, which she co-wrote alongside Neal Brennan, adapting the nonfiction book by neuropsychiatrist Louann Brizendine. […]

The Female Brain movie aims to comically detail the inner workings and complex power of brain chemistry among couples at different stages of their relationships. […] The film’s story follows five couples struggling through various stages of their relationships: whether it’s finding the right romantic balance; parenting; overcoming commitment issues; expressing emotion; or simply admitting to being useless around the house.


This won't be the first, or even the millionth, fictional exploration of popular gender stereotypes:

It won't even be the first dramatization of a general theory about gendered communication — for example, see the exchange between Gloria Clemente (Rosie Perez) and Billy Hoyle (Woody Harrelson) in White Men Can't Jump,  which explores the distinction between rapport talk and report talk that Deborah Tannen popularized in  "You just don't understand: women and men in conversation" (1990):

Gloria:  Honey? My mouth is dry. Honey. I'm thirsty.
Billy: Umm… [ Water Runs ] There you go honey.
Gloria: When I said I was thirsty, it doesn't mean I want a glass of water.
Billy: It doesn't?
Gloria: You're missing the whole point of me saying I'm thirsty. If I have a problem, you're not supposed to solve it. Men always make the mistake of thinking they can solve a woman's problem. It makes them feel omnipotent.
Billy: Omnipotent? Did you have a bad dream?
Gloria: It's a way of controlling a woman.
Billy: Bringing them a glass of water?
Gloria: Yes. I read it in a magazine. See, if I'm thirsty, I don't want a glass of water. I want you to sympathize. I want you to say "Gloria, I too know what it feels like to be thirsty. I too have had a dry mouth." I want you to connect with me through sharing and understanding the concept of dry mouthedness.
Billy: This is all in the same magazine?

But it may be the first movie to explore the idea that popular gender stereotypes represent "the inner workings and complex power of brain chemistry".

That idea, in one form or another, has been popular among philosophers, scientists and the general public for hundreds if not thousands of years, often taking the form of speculation about the superiority of female brains in certain respects. Thus William Thomas, Sex and Society (1907):

Indeed, when we take into consideration the superior cunning as well as the superior endurance of women, we may even raise the question whether their capacity for intellectual work is not under equal conditions greater than in men. Cunning is the analogue of constructive thought — an indirect, mediated, and intelligent approach to a problem — and characteristic of the female, in contrast with the more direct and open procedure of the male. Owing to the limited and personal nature of the activities of woman, this trait has expressed itself historically in womankind as intrigue rather than invention, but that it is very deeply based in the instincts is shown by the important role it plays in the life of the female in animal life.

But most of the stereotypes are empirically false — the actual group differences are generally small compared to the within-group variation, and often go in the opposite of the stereotypical direction. And Dr. Brizendine's claims about the neurophysiological foundations of the stereotypes seem to be mostly empty. For a long engagement with the details, you can sample the list of LLOG posts below. For a shorter and more authoritative account, see Rebecca Young and Evan Balaban's review of The Female Brain, "Psychoneuroindoctrinology", Nature 10/12/2006.

Overall, the evidence lines up behind Janet Shibley Hyde's conclusion in her review of meta-analyses "The Gender Similarities Hypothesis", American Psychologist 2005:

The differences model, which argues that males and females are vastly different psychologically, dominates the popular media. Here, the author advances a very different view, the gender similarities hypothesis, which holds that males and females are similar on most, but not all, psychological variables. Results from a review of 46 meta-analyses support the gender similarities hypothesis. Gender differences can vary substantially in magnitude at different ages and depend on the context in which measurement occurs. Overinflated claims of gender differences carry substantial costs in areas such as the workplace and relationships.

But the Female Brain movie might well turn out to be a hit.

"Neuroscience in the service of sexual stereotypes", 8/6/2006
"Sex-linked lexical budgets", 8/6/2006
"Sex and speaking rate", 8/7/2006
"Yet another sex-n-wordcount sighting" , 8/14/2006
"The main job of the girl brain", 9/2/2006
"The superior cunning of women", 9/2/2006
"The laconic rapist in the womb", 9/4/2006
"Open-access sex stereotypes", 9/10/2006
"David Brooks, Neuroendocrinologist", 9/17/2006
"Gabby guys: the effect size", 9/25/2006)
"'Every 52 seconds': wrong by 23,736 percent?", 10/13/2006
"Guys are a bit gabbier in Dutch, too", 10/16/2006
"Two new reviews of Brizendine", 10/30/2006
" Word counts", 11/28/2006
"Sex differences in "communication events" per day?", 12/11/2006
"Male and female college students are equally talkative", 7/5/2007
"The first time?", 7/5/2007
"Female talkativeness: 'Knowledge protected again induction'?", 7/6/2007
"Why men don't listen", 4/4/2010
"An invented statistic returns", 2/22/2013
"Sex and FOXP2: Preservation of endangered stereotypes", 2/28/2013