kylebgorman

Yes, this is a discussion forum, so while you're welcome to pose ethical questions, you should also feel free to weigh in on them too ;) It's a challenging question.

Is it the mere fact of someone being aroused by children that makes child pornography offensive to society, or is it our belief that the creation of such materials necessary involves some degree of exploitation? If the latter, what about the well-documented risks of exploitation of consenting adult pornographic performers? Does access to pornography increase or decrease adults' tendencies to engage in unsanctioned sexual acts?

A speech recognizer is not an "AI".

The first author's an independent scholar so I don't think that's right.

I have the same set-up: SomaFM (and a Pi with an external hard disk) and three Squeezeboxes all in sync.

If I may, there's no real reason to break out ACL vs. NAACL vs. EMNLP, since they're all run by the ACL and one would be hard-pressed to say how the EMNLP community might differ from the ACL community at this point. And if you're doing NAACL you might want to do EACL and IJCNLP too.

I have been doing this for a long time, and FreeDB's metadata is way, way worse than MusicBrainz's, and has less coverage too.

If you find any issues in MusicBrainz metadata you can submit a correction, which will be reviewed and voted on by community members. I've done this maybe 4-5 times for my collection of about 2000 releases and in each case, my correction was quickly accepted.

Sorry but this person isn't nearly famous enough to have been "cancelled".

Get W. Sidney Allen's slim text _Vox Graeca_. It goes over not only how Ancient Greek was pronounced but also how we know. The evidence includes explicit notes from grammarians and other writers, graffiti, borrowings, puns, and later forms of Greek. Allen also has another book _Vox Latina_ which does the same thing for Latin.

"text normalization" just refers to that sort of thing. it's a subcomponent of the "front-end" of a speech synthesizer but it can also be used for speech recognition (if your training data contains things like "145" you may want to convert it to read "one hundred and forty five", for various reasons) and information extraction (perhaps you want to treat "145" and "one hundred and forty five" as the same).

Mechanical reproduction is a force multiplier allowing your art to touch millions or billions. (Mozart & Michelangelo didn't know this was coming, but I'm a consequentialist, so I give them credit for it.)

Far smaller is the audience for play-by-plays of competitive chess. I'd guess because the experience is less visceral? I've read a beginner-intermediate books with this kind of play-by-play and found it interesting enough, but I'd take my record collection over those books any day.

Author should have a native English speaker read this for fluency...

Activist investors who want changes in business model or cost are extremely common. Just about every publicly traded company has to deal with one at some point.

I am building libraries for predicting which pronunciation of a homograph to use for in a text-to-speech application. E.g., when we go to synthesize "bass", should we use the pronunciation that rhymes with "mace" or "pass"?

He'd led a violent (albeit somewhat pathetic, and suppressed) attempt at a coup---that's why he was in jail. Rough comparison points in recent US history would be the domestic terrorists like McVeigh and Kaczynski, I suppose.

Unless you know something I don't, the SOTA in ASR is a hybrid system that still has all the complexity of the traditional generative model plus NNs on top.

As a professional linguist, I think that Farmer et al. are probably right and it's almost surely not a "writing system" at all. (This is a term of art with a very specific sense.)

They have "limo licenses" but yeah, few if any of them are actual limousines in wheelbase / shape. Most are towncars or upscale SUVs.

I would not recommend NLTK (or its book) or Jurafsky & Martin, or Manning & Schuetze. All are insanely dated. Watch some Coursera lectures, check out a newer, non-academic, application-oriented text, or just build something.

This is not a question that really would be of much interest to most working linguists, who aren't really willing to make value judgments about language any more than chemists are to talk about their favorite molecules.

Constructed languages are also generally considered "mostly harmless" but not really an object of scientific inquiry.

If you're interested in languages, you may want what people who study languages for a living are interested in. Or you dive into conlanging...there's probably a subreddit for that.

Or:

    alias dronezone='mplayer http://somafm.com/dronezone130.pls'
    alias groovesalad='mplayer http://somafm.com/groovesalad130.pls'

Before Apple sold music, iTunes had an Internet Radio tab that listed several SomaFM stations.

That's an interesting problem. But I'm not seeing much upside for you and there's a high chance of being laid off before you get situated.

Anything that uses the standard decoding strategy (Viterbi) is quadratic in the number of "states" (i.e., tags), though, and that easily dominates.

(Nowadays, many people just use greedy---linear in the number of states---decoding because it rarely is much worse.)

Oh and of course, "gyspy"/"jitney" cabs/vans, which are quasi-legal $2 bus lines running bus routes in underserved communities.

So, for context, yellow taxis run "anywhere" but in practice, they are somewhat rare anywhere but lower Manhattan and at the airports.

Green taxis, a new category, are only permitted to pick up hails outside of lower Manhattan. They're useful for going between outer boroughs but are reluctant or unwilling to take you to lower Manhattan because they can't pick up a hail there.

There is also another, long-standing category known as "car service", where a central dispatcher reached by phone sends you a towncar/minivan/SUV either at a scheduled time (say, for an early flight or your regular work commute), or as soon as possible. Like Uber, they may have an "account" for you so you don't need to provide payment information to the driver, but nowadays many of them have Stripe too. Prices are arguably better for airport rides than cabs.

I live in Central Brooklyn and my impression is that there are far more Uber "U"s than any of the other categories, at least in my neighborhood.

Optimization is only a distant goal according to Guido---I think he said so in the PEP or the dev mailing list discussion thereof.

I feel the same way about C++ smart pointers, which are remarkably simple to use and understand.

I don't think we should incentivize moving to such places unless we plan to also (somehow) massively increase the opportunities available in such places.

I don't know how big "textfile.txt" is, but try it with a much larger body of text? Also you'll almost surely want a much better tokenization algorithm than that. Zipf's Law can only hold "at the limit" of an infinite stream of text, so as a consequence you get a better approximation with a larger sample.

I don't think "amount of shared machine code" (DNA) is a good metric for functional similarity? What percentage of machine code do two small UNIX binaries (statically) linked against libc share? Potentially a lot. But obviously they would have next to nothing in common, functionally?

That we share so much DNA with primitive creatures is just another way of saying that amount of shared DNA is not closely related to functionality similarity.