Language Log

« previous post | next post »

Google translate for animals:

I need to point out that this is not an entirely new development. Hearsay, CMU's early-1970s speech recognition system, was built using the best Artificial-Intelligence theories of the time:

The system consists of a set of cooperating independent processes, each representing a source of Knowledge. The knowledge is used either to predict what may appear in a given context or to verify hypotheses resulting from a prediction. The structure of the system is illustrated by considering its Operation in a particular task situation: Voice-Chess. The representation and use of various sources of knowledge are outlined. Preliminary results of the reduction in search resulting from the use of various sources of knowledge are given.

One of the "sources of Knowledge" was the state of the chess board and the list of currently-valid moves, arranged in rough order of plausibility. And the "source of Knowledge" representing the analysis of the acoustic input was somewhat, shall we say, open-textured. As a result, it was possible to play a decent game of chess against the machine by barking at it.

A similarly Bayesian approach to pattern recognition is clearly built into this new animal-communication app.  A cough played to the cat-to-English translator produces the output "Might go out on the town tonight, don't wait up!" The same sound, input to the dog-to-English engine, produces the output "Seriously, what's with the stick thing? It's getting old."

Some other prior art:

"The dogs of speech technology" (3/1/2005)
"It's a dog's life — 0.3 bits at a time" (1/19/2008)
"Dog language mailbag" (1/21/2008)
"Not Professor Milton, but Professor Schwartzman" (1/21/2008)

Those looking for a niche not yet occupied might consider the possibility of incorporating chemical sensors into cell phones, allowing them to deal with the sorts of animal communication involved in an interesting recent paper by Fiona Berry and Thomas Breithaupt ("To signal or not to signal? Chemical communication by urine-borne signals mirrors sexual conflict in crayfish", BMC Biology 8:25, 2010):

Sexual selection theory predicts that females, being the limiting sex, invest less in courtship signals than males. However, when chemical signals are involved it is often the female that initiates mating by producing stimuli that inform about sex and/or receptivity. This apparent contradiction has been discussed in the literature as 'the female pheromone fallacy'. Because the release of chemical stimuli may not have evolved to elicit the male's courtship response, whether these female stimuli represent signals remains an open question. […]

Urine-blocking experiments demonstrate that female urine contains sex-specific components that elicit male mating behaviour. The coincidence of chemical signalling and aggressive behaviour in both females and males suggests that urine release has evolved as an aggressive signal in both sexes of crayfish. By limiting urine release to aggressive behaviours in reproductive interactions females challenge their potential mating partners at the same time as they trigger a sexual response. These double messages should favour stronger males that are able to overcome the resistance of the female. We conclude that the difference between the sexes in disclosing urine-borne information reflects their conflicting interests in reproduction. Males discontinue aggressive urine signalling in order to increase their chances of mating. Females resume urine signalling in connection with aggressive behaviour, potentially repelling low quality or sexually inactive males while favouring reproduction with high quality males.

This strikes me as a worthy continuation of D. Barry's seminal work on gendered communication in herring.

April 1, 2010 @ 9:15 am · Filed by Mark Liberman under Animal communication

Permalink