January 25, 2008

Bowlingual education, Critical Pet Studies, and human-poultry interaction

Are European researchers copying Japanese innovation in cross-species communication? This question arises in connection with our posts on Csaba Molnár et al. ("Classification of dog barks: a machine learning approach", Animal Cognition, published online 1/15/2008; see "It's a dog's life -- 0.3 bits at a time", 1/19/2008; "Dog language mailbag", 1/21/2008; "Not Prof. Milton, but Prof. Schwartzman", 1/21/2008). The issue was raised by Rob Troyer, who wrote:

Not to beat a dead dog, but a quick Google search of "dog translation" pulled up this Reuter article from 24 March 2003, "Dog translation device coming to U.S.", published at CNN.com/technology.

It discusses the "Bowlingual" a device "cited as one of the coolest inventions of 2002 by Time magazine" and developed with a cost of "hundreds of million of yen" by Tokyo-based Takara Company Ltd.

"The console classifies each woof, yip or whine into six emotional categories -- happiness, sadness, frustration, anger, assertion and desire" much like the computer program in the recent paper by the Hungarian researchers.

According to the article some 300,000 of these devices were sold in Japan in 2002-3, and the company was hoping to meet great success with its English version in the US which "is home to about 67 million dogs, more than six times the number in Japan."

The Wikipedia article  on the Bowlingual links to an inspiring Takara press release about the potential impact of interspecies communiation on world peace("Bowlingual Presented to Russian President Putin by Japanese Prime Minister Koizumi", 5/30/2003):

The devices were prototypes of the U.S. version, currently in the final phase of testing, and included prototype English packaging and English user manual. The Ministry of Foreign Affairs, Russian Division, also assisted in the preparation of a summary of the user manual in Russian. Although the prototypes were available only in Red, they were adjusted to two different frequencies and labeled accordingly so that they could be used together with President Putin's two dogs--a standard feature when using two different-colored off-the-shelf Bowlingual units. The units' Set Up menus were also set to match the breeds of the President's dogs, which were said to be a standard Poodle and a Labrador Retriever.

It with great pride and satisfaction that Takara has been able to help facilitate the bonds of friendship between two leaders, Prime Minister Koizumi and President Putin, as well as promote peace and strong international relations between Japan and Russia.

Alas, Wikipedia's reference to the Bowlingual web page links via the Wayback Machine -- suggesting that this is an ex-product. There's a 2003 SF Chronicle article that suggests why (Sophia Lin, "Gadget's bark is bigger than its hype: Vet puts 'bark translator' to the test. Verdict: nothing more than $120 curiosity", 8/16/2003).

Dr. Lin describes some informal tests, in which the Bowlingual seems to perform non-randomly but also uninformatively, and concludes:

My final ruling? The Bowlingual is fun to play with for a while if you got it for free, but it's not very useful because the translations aren't trustworthy and most don't make sense. The toy is marketed as being backed by strong science carried out by respected researchers but somehow, despite their accolades, they produced a dud.

I haven't been able to figure out who the "respected researchers" were, or where their "strong science" was published. In particular, I was unable to find any references to the Bowlingual in the machine-learning or animal-communication literature. Google Scholar does turn up some published discussion by literary scholars, including B. Lennon, "Misunderstanding Media: The Bomb and Bad Translation", Criticism, 2005; H.J. Nast, "Loving ... Whatever; Alienation, Neoliberalism and Pet-Love in the 21st Century", ACME; H.J. Nast, "Critical Pet Studies", Antipode 38(5):894-906. Unfortunately, none of these offer any information about the device beyond quotations from the popular press.

There is also at least one paper in the computer-science literature that mentions the Bowlingual, and it's a doozy: Shang Ping Lee, Adrian David Cheok, Teh Keng Soon James, Goh Pae Lyn Debra, Chio En Jie, Wang Chuang and Farzam Farbiz, , "A mobile pet wearable computer and mixed reality system for human-poultry interaction through the internet", Personal and Ubiquitous Computing, 10(5) 301-317, 2006. The abstract is fascinating:

Poultry are one of the most badly treated animals in the modern world. It has been shown that they have high levels of both cognition and feelings and as a result there has been a recent trend of promoting poultry welfare. There is also a tradition of keeping poultry as pets in some parts of the world. However, in modern cities and societies, it is often difficult to maintain contact with pets, particularly for office workers. We propose and describe a novel cybernetics system to use mobile and Internet technology to improve human-pet interaction. It can also be used for people who are allergic to touching animals and thus cannot stroke them directly. This interaction encompasses both visualization and tactile sensation of real objects.

However, the paper's only information about the Bowlingual is this:

The growing importance of human-pet communication can also be seen in recent related company products. Recently, an entertainment toy company [12] has produced a Bowlingual dog language translator device. It displays some words on its LCD panel when the dog barks.

The reference is "12. TT Company http://www.takaratoys.co.jp/bowlingual", which is a dead link, suggesting again that this is an ex-product.

So, to sum up, it's hard to tell whether there is any connection between the pioneering but under-documented (and perhaps useless) Bowlingual, and the serious research described in detail by Molnár et al., demonstrated to derive 0.3 bits of information per doggie utterance. The idea of translating dog barks into six categories is similar, but the mapping between the Bowlingual six categories ("happiness, sadness, frustration, anger, assertion and desire") and the Hungarian researchers' six categories ("Stranger, Fight, Walk, Alone, Ball, Play") is not transparent. Neither Bowlingual nor Takara is mentioned in the Molnár et al. paper.

[On a related topic, I wonder how many patents on pet communication devices are out there...]

[Update -- at some point, Vodaphone released a Bowlingual-equipped cell phone, according to the Guardian's technology section (" Pet practice", 2/12/2004):

'Woof!" It might sound like a meaningless bark but, in fact, the dog is saying "Ya ne! Soba ni konai de!" (Hey! Don't come near me!). And while a European might make the mistake of approaching the diffident hound, Japanese dog owners would know to steer clear. Why? Because their phones would translate for them.

Bowlingual, a mobile application available to Vodafone subscribers in Japan, has a repertoire of about 200 dog phrases. It's just one of the many strange but innovative mobile products available in the Far East - and another reminder of how far ahead the Japanese are in non-voice applications.

It's not clear whether the upgrade from 6 to "200 dog phrases" represents a genuine research breakthrough, or a simple hack (e.g. choosing randomly among multiple "translations" for each of 6 categories), or just a random journalistic misunderstanding.]

[Update #2 -- a lead! The 2002 "Ig Nobel Peace Prize "

... went to Keita Sato, President of Takara a major Japanese-based toy company, Dr. Matsumi Suzuki, President of Japan Acoustic Lab and Dr. Norio, Kogure Executive Director of Kogure Veterinary Hospital for Bowlingual, their dog-to-human translation device in "promoting peace between the species."

Unfortunately, neither Suzuki nor Kogure seems to have published anything on a relevant topic, at least in English. And the recent Molnár et al. paper does not appear to reference any work by Japanese researchers of any name.]

[Update #3 -- a bit of searching turns up some patents. One is " Apparatus for determining dog's emotions by vocal analysis of barking, which tells us that

emotions represented by the reference voice patterns include "loneliness", "frustration", "aggressiveness", "assertiveness", "happiness", and "wistfulness"

The author is Matsumi Suzuki. I'm glad to say that the patent contains quite a bit of detail about the features and their claimed meanings.

Another is " Device and Method for Judging Dog's Feeling form Cry Vocal Character Analysis", for which Suzuki is also the inventor. In this case,

The reference voice patterns by feelings correspond to the feelings of 'loneliness', 'frustration', 'threating', 'self−expression', 'delight' and 'demand'.


Posted by Mark Liberman at January 25, 2008 07:11 AM