Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
Technology

Full-Text Audio Search 135

Posted by timothy from the information-awareness dept.
Captain Chad writes "The latest print edition (12/16/2002) of InfoWorld has an interesting article about an audio search program by Fast-Talk Communications. (The article is not yet available on the InfoWorld web site, but the Fast-Talk site has some good info, including a downloadable trial version.) The product works by breaking the audio stream into phonemes, which are the 'basic units of sound in a language.' The search is then performed for a specific sequence of phonemes. This method is faster and far superior to traditional audio searches which convert to text and then perform a normal text search. The author of the Infoworld article, Jon Udell, tried a variety of searches that were surpisingly successful. If this technology is as good as he claims, there is a reasonable chance it will revolutionize the way we store data. Maybe there will even be an 'Audio' tab on Google." Here's the Infoworld article.
This discussion has been archived. No new comments can be posted.

Full-Text Audio Search More

Full-Text Audio Search

Comments Filter:

  • How long... (Score:2, Interesting)

    by slashuzer ( 580287 )
    How long before this technolog is actually banned under DMCA? After all, if you *need* to search music by electronic means then you are obviously a *thief*.....

    /sarcasm

  • Just one more step on the road to TIA (Score:5, Insightful)

    by Grapes4Buddha ( 32825 ) on Sunday December 22, 2002 @05:48PM (#4941626) Journal
    How long before the feds start digitizing all of our telephone conversations and using this technology to google our private conversations?

    Yay!

    • Re:Just one more step on the road to TIA (Score:5, Insightful)

      by m.lemur ( 618095 ) on Sunday December 22, 2002 @05:56PM (#4941667)
      You think its not happening already?

      • Re:Just one more step on the road to TIA (Score:3, Interesting)

        by Anonymous Coward
        You think its not happening already?

        Yes and no; more no than yes:

        Until about two years ago, speaker-independent telephone speech dictation for accurate word-spotting wasn't good enough to run on mass volumes of calls.

        So, the story goes that it finally got tested in 2000. Even with fairly high accuracy rates, the utility of the extracted data was near or below zero, in that the amount of time a human agent would spend reviewing the tagged calls (international calls can be eavesdropped without a warrant) was less effective by some measure than if the same agent had been following other readily available leads.

        The whole problem is that most of the people who talk about {drugs, bombs, nerve gas, etc.} are not the people engaged in the manufacture and smuggling of those contraband; usually such people use code words. A code book with coordinates, maps, and timetables sent by FAX between anonymous hotel business centers can completely confound even the most concerted traditional eavesdropping scheme, let alone an automated word-spotting system. So, the agents ended up reviewing hundreds of calls between people talking about cocaine, but not one call between people talking about shipping or producing cocaine.

        Speech recognition has a place in law enforcement by mass-eavesdropping, but I don't think that place is found, yet. I predict it will probably end up being used more to ferret gays out of the military than anything else.

    • My understanding is that most phone calls are digitised already. The connections between exchanges are (mostly) fibre-optic, so the traffic is digital. I can't comment on who's listening in. Surely that doesn't happen without due cause and the appropriate courtroom approvals?

      </wide-eyed innocence>

      Ironically, I'm posting this over a dial-up which modulates my digital data to analogue. The signal is then digitally encoded at the telephone exchange, with the whole process being reversed as the signal reaches my ISP.

    • Re:Just one more step on the road to TIA (Score:4, Interesting)

      by RDPIII ( 586736 ) on Sunday December 22, 2002 @09:17PM (#4942339) Journal
      How long before the feds start digitizing all of our telephone conversations and using this technology to google our private conversations?

      Let's see, given 5000 billion dial equipment minutes in 2001, we'd have around 150 trillion seconds of conversations. Assuming you could code everything at a bitrate of 8kbps, this would mean roughly 150 terabytes of compressed data for 2001 alone. Presumably the storage would be distributed at the switches where you record the conversations. So the problem is now to compress, transcribe, index, search, decompress, and access 150 terabytes of distributed storage.

      And keep in mind that doing a phoneme transcription rather than full-blown speech-to-text is likely to generate a whole lot of nonsense transcriptions, precisely because you don't have any guiding information from the words in the conversation.

      While I enjoy Popular Paranoia as much as the next guy, the whole TIA thing does not really get to me. My reaction is mostly: bring it on, if you really think you can convince yourself that it can be done.

      • 150 terabytes to manage is not really all that much if you consider the budget that this program will have. Even if it was, the ability to store and process data will increase at a much faster rate than the amount of data generated by telephone conversations, so your argument about it not being entirely feasable from a technical standpoint would only hold water for a few years at most.
        • Sure, you can easily store 150 TB of data per year. But the problem is: how are you going to get anything out of it? You'll need an index, but what would you put in the index if all you have is phonemes? Similarly, you can search for suspicious words, but you can't easily build word-based concordances from just phonemes. So the phoneme-based word spotting technology (which, by the way, is not all that new) is definitely not something that any form of TIA should be based on, if it's supposed to have any practical utility.
      • Comment removed based on user account deletion
        • Granted, you can go on KaZaA and search through 4 petabytes of data fairly quickly, but keep in mind, a large portion of that data is redundant, and the data is not what you are searching. In the situation of TIA, the software would have to analyze every portion of the 150 tb of data to perform a search, whereas KaZaA has a set of meta information for each file. For example, for a 1 GB copy of the Matrix (not that anyone would have an illegal copy of a movie on KaZaA), there is probably 1 kb at most of data to search.
      • "bring it on, if you really think you can convince yourself that it can be done."

        Maybe the idea that they CAN'T do it right is the exact idea that should worry you. Think about this: They can't execute the war on (some) drugs correctly because it's a bullshit idea, same as TIA. Now we have jack-boot thugs busting in on completely innocent 70 year old couples and scaring them (literally) to death... And we have agents (of Matrix quality) imprisoning, denying medication, and consequently killing medical MJ patients who are using the drug legitmately (Peter McWilliams [hazlitt.org]).

        In short, things the government does that it is not supposed to do or cannot do correctly are the exact things you should worry about.
    • Is probably already using something like it.

  • How long... (Score:2, Interesting)

    by ejdmoo ( 193585 )
    before we have a "video" tab on google? :)

    • Re:How long... (Score:1, Informative)

      by Anonymous Coward
      Actually, there is already an equivalent to the proposed Audio tab on Google. It is an HP Research project, SpeechBot [compaq.com]

      Another similar product is already on the market from Scansoft [scansoft.com] (formerly Dragon Systems) but it uses a complete different approach than the fast-talk product. We are actually using Scansoft product where I work to index all of the media (audio and video) files on the corprate lan.
    • before we have a "video" tab on google? :)
      With the way Bush is going how long before he gets a "homeland" tab on his own version of Google :P

      -Jason

  • Point? (Score:1, Interesting)

    by Anonymous Coward
    I can't help but wonder what the point of a "Full text audio search" would be. Most songs have their lyrics online, most speaches are already in document form, etc etc. Plus, wouldn't searching through audio files be incredibly computer-taxing? What about the services like AOL that cache almost everything? Would they start having to cache 300meg wav files?

    Now, if I could hum a tune into my computer and have it find what song I was humming (for those songs you just can't remember the lyrics too), i would be much happier.
    • and there are several research prototypes that can do it (check out a paper from Cornell [cornell.edu], or just google for query-by-humming)
    • I can see a good use for it, and that's taking notes. Imagine carrying around a microphone with you 24/7 and recording everything you hear. We've got the space to store it all, after all. Then you can go back and check "When did he say that meeting was?" or whatever. And those with significant others know how often you'll end up arguing over who said what.
    • 1 - real-time key-word alerting - i.e. having software listen out for key-words on an audio/visual source, and alert someone appropriately when they have detected. TIA, anyone ?

      2 - data retrieval - a phonetic query language - cool !!!

      3 - pr0n video spider - just listen out for lots of Ooohs, Aaahs, and the like (sorry, could not resist ;o))

    • Re:Point? (Score:2, Insightful)

      by lux55 ( 532736 )
      Aside from searching for music, I can see this being really useful in web conferencing software. Consider this:

      You hold a meeting where each person's channel was recorded and stored as part of the meeting info. Upon saving the meeting minutes, the software builds a phonetic index of the entire conversation.

      Searches later on would be no more taxing on the server than a fulltext search in MySQL is now.

      Useful? Definitely. And that's just one possibility.

    • Re:Point? (Score:3, Insightful)

      by Chasing Amy ( 450778 )
      Well, I'd personally love to have an audio search tool to comb through all the mp3 files of talk radio programs such as *Loveline*, *Opie & Anthony*, and *The Greaseman*, which I have. Sometimes I think, "Now which show had that cool bit about..." and I have no hope of finding it.

      For a professional rather than personal use, imagine how useful this could be to radio stations if they keep digital archives of their programs--if someone wanted to look up a particular program based on a vague memory of some of the text, a tool like this would be invaluable.
    • Well if your the govt, then the implications are huge. If they had a way of effieciently "keyword" scanning spoken conversations (esp. phone), this would help intelligence gathering tremendously. If this company can make their stuff work as advertised, they have huge upside potential with the likes of the NSA, DOD, CIA, etc.
  • Do the songs need to be converted to this new phoenic format or can you just search the audio? Wouldn't this use a tremendous amount of computing power?
    • 'cmon, would it kill you to RTFA?

      oh, what am I saying, of course it would!

      and in answer to your question, you index it once which can be done in pretty much 1:1 time, then you save the index files and just search *those* to find things - the index files tell you which recording and timecode the result is found at, then you playback from there.

  • Converting to Text (Score:1, Insightful)

    by TrekkieGod ( 627867 )
    I don't know much about the subject, but isn't this the method used to convert speech to text? Sounds to me like it's the only way to do it...comparison of a sequence of phonemes to another, except that the each word in the dictionary is associated with a sequence of phonemes. And that's why you're required to "train" the software with your own voice/accent.

    Somebody who knows about the subject, please post and explain the process.
    • Looking at it, it seems essentially the same, except that this will remove the most error prone part. Conversion of phonemes to proper sentences is not easy. there are too many homonomes. If you were to talk about a hare, it could easily be translated as a hair. Therefore a search for hare will nto produce a match.

      If we convert the text to phonemes instead, hare and hair resolve to the same result. So a search for either of those words will produce a match.

      In hindsight, this is an obvious idea. Like many obvious ideas, the person who spotted it was a genius.
    • Most speech to text conversion software has to initially bind the word it thinks it hears to a word in the dictionary. And it could be the wrong word - therefore you will never find the word you are looking for (also you would have to re-index if you are even able to change the word to the correct word)... This Fast Talk technology does not do that, all it tries to do is match the phonemes it hears initially to the phonemes from the query screen and as long as the words sound similar, it will find it. Also dictionaries that the speech to text software's use are not good with proper names, acronyms, and slang words, etc.
    • The Fast-Talk technology doesn't convert Speech to Text. It indexes audio (thats where we apply the secret sauce to phonemes) and then provide the capability for you to search for words, terms, or phrases by applying that secret sauce to your search query and matching that to the indexed audio. The audio that is indexed is already trained to deal with various accents and dialects so it is not necessary to train it for your own voice or accent.
  • Actually, Google already has a voice search [google.com], albeit in beta form.

  • Woohoo. (Score:2, Funny)

    by Nordberg ( 218317 )
    Now I can finally search for the Free Radio Linux kernel reading by phonemes!
  • eat wheel nibble fork,

    linx ear [hazardfactory.org]

  • Yes, but... (Score:5, Funny)

    by Erpo ( 237853 ) on Sunday December 22, 2002 @05:55PM (#4941657)
    ...can it decode rap and/or reggae? I swear I can't understand 3/4 of those lyrics. Songs could start with

    -----BEGIN PGP MESSAGE-----

    and I wouldn't be able to tell the difference.
    • Forget rap and reggae lyrics. How about technical terms using letters and numbers, especially since they often are only intended to have one pronunciation of several possibilities (See Roman Numerals, MAC OSX(="ten"), etc.). I'm wondering how it would deal with those.
    • Italys nose probes them et al. fingering our hose
      the fee Cult to longer stained syrups and Hussein marmot pervert sucks eggs rat. Intact, eye amusing into dick tape his pest of flash snot.

      - - -

      It has no problems at all firguring out those difficult to understand lyrics and has an almost perfect success rate. In fact, I am using it to dictate this post to Slashdot.

  • Patches... (Score:4, Funny)

    by Rhubarb Crumble ( 581156 ) <r_crumble@hotmail.com> on Sunday December 22, 2002 @05:57PM (#4941671) Homepage
    Combine this with the Streamed Audio Kernel source [slashdot.org], and it's only a matter of time before people leave patches on Linus' voicemail! The great thing is that to patch an audio kernel, you only need a tape recorder.... :P
    • Now what would be cool is a C virtual machine running in a tape recorder, so you could just get one of those larger-capacity audio tapes, record the linux kernel broadcast for the next 600-something days, and then boot up linux on a tape recorder! Hah! NetBSD, beat that!

      • and then boot up linux on a tape recorder!

        And party like it's 1986! Don't go there! (nightmare visions of a zx spectrum enter my mind...."look, it's the yellow and blue pattern and the random static noise! that means it's loading the program!") ;-)

        • I keep trying to sample my collection of BBC Micro software on tape and find a program to convert it to files. Apparently it is possible to read the sound in software and write a dump suitable for an emulator. I just need to figure out how to record sound under Linux on my hardware (builtin sound on A7V333).
  • First I'd like to say that this would be wonderful for NPR to use. *drool*

    On a serious note. I really didn't think NLP software was to the point to make this plausible. I've never actually used NLP tools, but what I've heard in the main stream is that while they work they aren't perfect. This is fine for someone starting at a screen while talking or someone who is going to review the transcription, but it seems like it would break any automated system when there is not system of checks in place, since this involves a human.
    • The nice thing about using speech recognition for an application is that the reco rate doesn't HAVE to be 100% accurate. After all, YOUR speech recognition system (the one in wetware, between your ears) isn't 100% accurate, and that doesn't stop you from understanding people, right?

      Dictation (the only application where you need/want 100% accuracy) is only one small application for speech recognition.
      • How can you reliably search for something if the automatic translation can be wrong?

        the one in wetware, between your ears) isn't 100% accurate, and that doesn't stop you from understanding people, right?

        The wetware has knowledge of context which helps reduce the choices of a potential vocal sound. This is one of the hardest things to build into a NLP system and afaik it currenlty isn't anywhere near perfect, otherwise we'd have voice interfaces to a lot more things in the common world
        • You can search reliably because of the redundancy of language. You don't need 100% accuracy because people repeat themselves, names are more distinct than common words, etc.

          It really depends on the problem. If you want to guarantee that if a person says "frobnitz" that you're going to find it, yes, you need 100% accuracy. But even human listners aren't going to give that to you. If you want to find people mentioning words associated with terrorism, your accuracy rate can drop somewhat. If you want to find people talking about the *topic* of terrorism, your accuracy can drop even more.

          Your comment about context is very perceptive, however -- I would say that NO current ASR system has essentially ANY real-world context, and as you said this is a tremendous boost in how humans interpret speech. Once that breakthrough is made, NLP in general will take a quantum step forward.
          • Here's a real world problem. I was listening to NPR last weekend and only caught part of a program. I was able to track it down rather easily from the stations schedule of programs, but what if I couldn't? What if all I remember from the show is the line "and then she turned away from me and walked away. I was destroyed" how do you search for such a thing in the escentially fuzzy system you have described?

            I see how your description would work in a key-word environment, but what about a phrasal aspect? Or is our best bet ever going to be a fuzzy system?
            • Well, that's a bit of a contrived problem. It's unlikely you'd ever only remember that about the show. When was the last time you searched for a web page only knowing a single phrase with almost no content words or names, out of context, without any idea of the context?

              In the case of the NPR show, it's likely you'd know the general topic, or a speaker, or some proper names that were mentioned. All of these can be used to augment your search, and all of them can contribute to the accuracy of your results.
              • In that case you're be correct most of the time you have enough context.

                However I've seen other threads where they have I guess assumed that this technology would be used for music. I think that remembering a single line would be more apt and example in such a case.
            • Actually you're both right. This technology has the capability of searching for an individual term or for a phrase or line. When speed is important you can use both the time/channel to limit the search and then apply the quote to go directly to that part of the program. When you use phrases or lines of speech the search accuracy actually improves.
  • Now you can pretend you care about what they are saying and not just listening for the sweet intro music.

  • Dialects and Foreign Language application? (Score:3, Interesting)

    by beanerspace ( 443710 ) on Sunday December 22, 2002 @06:04PM (#4941704) Homepage
    So does anyone out there know how well this technology deals with accents and dialects? If so, perhaps we could finally see that 'Star Trek' like universal translator - or at least translate on a large scale media works from the past century into other languages.

    Of course, noble thoughts aside, I keep thinking how useful it would have been to have such technology in college when I had to transpose long lectures from my chicken scratch.

    Hmmm ... does this spell the end to stenography as we know it?
    • perhaps we could finally see that 'Star Trek' like universal translator

      It's been done already. [ectaco.com]

    • These are totally seperate issues. All this does is pattern match on a string of phonemes, it makes no claim to understand the meaning behind the sounds, which is what's necessary for a universal translator. In fact, this is actually farther from a translator than the old-style, convert-to-words-then-string-match methods, since that one cares about what the words the sounds make up. This is, however, more efficient, and more versatile, since it needs no dictionary, for this one particular application.

      Phonemes are pretty much language independent. One particular sound sounds the same in different languages. It might be spelled differently, and it certainly falls into different places in different words, but it's made the same way in the mouth, and it has the same acoustic pattern (there are some variations, since some languages make distinctions between different sounds, and others don't, like [l] and [r] are the same phoneme in Japanese, but not in English, and [p] and [p^h] (aspirated [p]) are the same in English, but not in Hindi, but this mostly doesn't matter, since in a given language each form tends to be used in the same words regardless of the speaker). Converting an audio stream to a sequence of phonemes is basically a solved problem (given lack of inflection/emphasis, no background noise, etc.), this is just a new, useful application of an established technique. The problem of translation lies in finding the meaning behind the phonemic sequences.
    • This speech search tool works because phonemes are simple.
      Cognitition and translation are VERY complex.

      Phonemes are a unifying constant of speech, not cognition of language. By definition, words are converted to audible phoneme, not spelling with etemological, gramatical and syntactical meanings.

      Any words used to search are converted to phonemes and then searched against a phoneme trascript of human speech, simplifying and broadening the chance of a match.

      This is MUCH easier than the reverse, of converting phonemes into words, particulaarly homonyms, e.g. "pair" and "pear", "two", "to", and "too". The complexity of extracting correct words, grammar and syntax makes understanding the orignial spoken message VERY hard. We have yet to reliably solve computer cognition of (machine readable) written or spoken language.

      We often use proper nouns such as people and place names. Is "Victor" some guy or the battle winner?
      What would a universal translater think of a spoken question "Do you listen to Phish?"

      How about the amazing French subtitles for "Pulp Fiction"? to paraphrase Travolta's pun joke... A family of tomatos are walking down the street, and the baby Tomato keeps dawdling and getting distracted, and after two warnings to keep up with the family from the Daddy tomato, the Daddy finally just looses his composure on the third time and pounds the pulp out of the baby tomato and yells 'Ketchup!'. A pun for catch-up and tomato ketchup.
      The French subtitle change the joke from Tomatos to Lemons, err... rather "Citron" and on doling out punishment yells the pun "Citron presse!" meaning "Lemon hurry up!" and "lemonade!"

      Universal translators are years off. High level translations will require humans for the forseeable future.

      Mac Refugee, Paper MCSE, Linux wanna be

  • Imagine... (Score:4, Insightful)

    by tjamme ( 601294 ) on Sunday December 22, 2002 @06:09PM (#4941719) Homepage
    ... Or imagine Google recording all possible audio streams (TV, radio, ... streets?) and allowing us to search those? All it takes is enough procesors, a bit of wiring...

    Now if you record street conversations or all types of public conversations... Do a search on 'bomb'... How appealing is that to big brother.

    All right... I'm learning sign language. Now.
    • Vendor : Look at it's sleek lines, it's a real goer this one. Client : Looks like a bit of an old bomb to me. CIA : Down on the ground everybody !
    • All right... I'm learning sign language. Now.

      Too bad ... computers already know sign language [mit.edu]
    • I dunno, why don't you try it. Go to Google and do a search on 'bomb'. Was that very useful or appealing to you? I imagine not! :-)

      Now, try that same tactic on every conversation in America. The utility would be some order of magnitude LESS than the crap you got back from google! (if you can have utility less than zero, that is!)
    • Just googled for "bomb". Got 5,430,000 results. Imagine being Big Brother and having to sort through THAT.

      • Re:Imagine... (Score:3, Insightful)

        by Idarubicin ( 579475 )
        Just googled for "bomb". Got 5,430,000 results.

        I just did the same. Got 5,580,000 results, only three hours later.

        At that rate of growth, (50,000 bombs per hour, or about 14 bombs per second) there's going to be an awful lot of poor bastards at the FBI/CIA/NSA chasing noise...

  • Exciting Implications (Score:3, Insightful)

    by flopsy mopsalon ( 635863 ) on Sunday December 22, 2002 @06:09PM (#4941722)
    By focusing on phonemes rather than syllables or whole words, this software can operate independent of any one languange. This has exciting implications not just for audio searching, but implies a strong beginning for voice-recognition and even speech translation software.

    I just hope one of those nuisance lawsuits from Tzsvestaeya Zolskovova, the eccentric widow of Sergei Zolskovova, (Russian lunguist who coined the word phoneme) over the use of the term "phoneme" doesn't hobble progress in this fascinating area.

  • Phoenetic search engines (Score:3, Interesting)

    by Moonshadow ( 84117 ) on Sunday December 22, 2002 @06:09PM (#4941723)
    I once wrote a phoenetic search engine for a site that took keywords and broke them down into their soundex phonemes, then stored those. Then, when a search was executed, it would convert the text words into phoenetic pieces and search the database for matches. It was quite accurate, actually. For example, one could search for "olif ghardin" and one of the returned results would be "olive garden".

    I guess this is a similar idea. Pretty cool tech.
  • For editing films. For documentaries in particular, this would be a godsend. Imagine if, in addition to video/audio tracks, you had a simple 'text' track with which you could easily assemble your cuts.

    If nothing else, putting the computer to work on the 'condense 100 hours of footage into pieces of paper' stage would be a nice step.

    If it prevented just one assitant editor from going insane, it'd be worth it. Do it for the children.

    -Brett

  • What we need now is (Score:3, Funny)

    by Mr. Shiny And New ( 525071 ) on Sunday December 22, 2002 @06:14PM (#4941745) Homepage Journal
    A search engine that lets you hum a song and it figures out which one it is.

  • Wow, cool idea (Score:3, Insightful)

    by Drakonian ( 518722 ) on Sunday December 22, 2002 @06:15PM (#4941751) Homepage
    I've always thought that audio/video is one huge information bank that has never been easily accessed. If you know of something textual, you go to Google to find it. But what if you wanted to read a Steve Jobs keynote from a couple years back? It's not particularly likely that anyone transcribed it for you. The video stream is probably long gone. But with this technology, you can have a searchable record of that fairly easily. Brilliant stuff.

    Someone mentioned it can be used by the government for TIA stuff - agreed, but same with any technology. It has its positive and negative uses. I don't think we are all going to revert to cavemen to get away from it.

    • One thing not obvious from the InfoWorld article is that the FastTalk system does NOT leave you with a transcribed version of the speech. It is *SEARCH ONLY* -- meaning, it can tell you that the word "frobnitz" was spoken at 1'30 into a particular article, and then you have to go and listen.

      This is of somewhat limited utility. Perhaps if you had 2000 hours of John Gotti talking about lasagne, and one minute where he was talking about rubbing out Sammy the Bull, you could speed forward to find that bit. But if he talks about Sammy a whole lot, it's going to be easier to *read* and skim than listen.
    • I've always thought that audio/video is one huge information bank that has never been easily accessed.

      I consider audio/video data "fuzzy" -- there is no clear cut method of intrepreting such data. Here's a real world example: tell a computer to determine if a movie is pornographic or better still find out if a picture is showing a vagina, penis, etc.

  • If you had a database big enough, you could almost make your own audio episode of any tv/radio show there is out there. With the right equipment, you could have a celebrity say anything, anything at all you want. This sort of program was coming, but with it brings cool and at the same time terrifying feelings from me...
  • There's one kind of "audio search" I'd really like to see: searches for a song by tune.

    I've seen a couple of web sites which offer tune searches, but they all work on the index system used in fake-books: start from the first note, and then from there, say whether the next note is higher, lower, or the same. But this system has problems: a reasonably short search will match a whole lot of songs; it's often hard to tell whether certain extra notes are considered part of the tune; and some songs have an obscure beginning and an easily recognizable theme farther in, and you don't know which one is indexed.

    These sites have also tended to only index very well-known tunes - usually, folk songs, show tunes, and a few jazz standards.

    One site allows you to send them a recording of you whistling the tune, which seems like an improvement, but it actually just translates it into the up-down-repeat notation.

    My ideal music search would be something that would take large quantities of music (let's ignore for the moment where it gets the large quantities of music without pissing off the RIAA) and scan each song for prominent tunes. You could then search these with perhaps the up-down-repeat notation, but also by inputting music notation, for people who know it. The search would have to be key-insensitive, and allow fairly fuzzy matches.

    If it could give me the name of that pop song/jazz tune/classical piece I just heard on the radio, it'd be pretty good.

    But if it works really well, it'd be a blessing for music composers - they could just search for that tune that just popped into their head, instead of worrying over whether they're subconsciously ripping off another song.
    • If it could give me the name of that pop song/jazz tune/classical piece I just heard on the radio

      It would be better if the radio station just transmitted the name and other relevent information along with the tune itself. Digital radio must be able to do this, surely? And if not, why not?
    • I thnk that would be a nightmare for composers - too many are worried that they are subconciously ripping off a song.

      I can just see composers checking the database everytime they 'hear' something familiar, and never ignoring the fact that 'I just ripped off a Elvis Costello chorus. Oh, well. It worked for him.' Just write the damn song; if it is good it will stand on it's own merits. Everything old is new again! The public has a ever-shorter memory, and when you got just 12 notes, you like it that way.

      Besides, every musican has a friend who will pipe up and say, "That sounds just like that Archies song!" Don't be afraid to rip off a couple of notes, otherwise Dylan and the Beatles own the world.

      Bad artists borrow, good artists steal. (Thanks to whoever this sig belongs to..;)

  • We have three > 300,000 sqft underground facilities loaded with rows of 2U rack systems with eight 120gig hard drives in each. Every phone call you have had since 1994 is now stored in this massive datasystem. Transferring all the old calls from tapes to the new system was probably the most tedious job I have ever done.
    • Can we access these archives? I think that it would be pretty cool to hear some of my conversations from 1994. Or is this archive just for the government to search when it wants to?

  • Refined twist on an old idea (Score:3, Informative)

    by pongo000 ( 97357 ) on Sunday December 22, 2002 @06:41PM (#4941847)
    Soundex [archives.gov], which uses the way words sound rather than the way they are spelled, has been widely used by the government and genealogy researchers for the past 60 years. This isn't exactly "new" technology.

    Why are more and more /. articles starting to sound like corporate press releases?

  • So... (Score:5, Funny)

    by dacarr ( 562277 ) on Sunday December 22, 2002 @06:46PM (#4941859) Homepage Journal
    Does it recognize speech, or does it wreck a nice peach?
  • Direct search using various phonetic representations has been around for many years. All things being equal, it's known to be somewhat better than searching the output of speech recognizer using approximate string matching. But you have to weigh against that that both speech recognition and approximate string matching are being pushed much harder than this kind of search, so you may end up getting better performance using speech recognition and string searching anyway.
  • The folks down at Streamsage [streamsage.com] have been working on this for a while now. They are working on an index of NPR, last I heard. They do video and audio; a search retrieves the relevant clips of video. It works really well, apparently. This will be a fantastic boon for universities who have all kinds of lectures on video with no way of knowing where to find the information a student needs.

  • InfoWorld article (Score:3, Informative)

    by prostoalex ( 308614 ) on Sunday December 22, 2002 @06:58PM (#4941904) Homepage Journal
    The article is not yet available on the InfoWorld web site

    Actually it is. InfoWorld: The Power of Voice [infoworld.com].

  • I looked into this recently (Score:4, Informative)

    by RebornData ( 25811 ) on Sunday December 22, 2002 @07:17PM (#4941997)
    There are a few papers available for download from their website, but you have to register. Basically, traditional voice recognition parses the audio stream into some meta-form, usually representing phonemes (the low-level "atomic" sounds that your speech consists of). These phonemes are then matched against a dictionary of known words (and the phonemes they consist of) and text is produced.

    Because phoneme recognition is not particularly accurate (for example, it's hard to tell the difference between "hard d" as in "Dan" and "hard b" as in "Ban" over a noisy phone line), traditional speech to text systems use several approaches to improve accuracy. One is to improve the accuracy of the basic phoneme recognition by "training" it for a specific voice. Another is to use all sorts of hairy-language-specific grammar / syntax algorithms.

    Computationally, it's the matching of the phonemes against the dictionary that's the most difficult, and the larger the dictionary, the less accurate and more CPU-chomping it becomes. In addition, searching the resulting text for specific matches grows less accurate as the search string increases in length, due to the likelihood of a transcription errors.

    The cool thing that Fast Talk has done is to store and index the phoneme meta-data, rather than complete the recognition to text. When you enter search words, they break the search string into phonemes and look for matches that way. This has several positive benefits:

    1. Computational resources are dramatically lessened, since the "phoneme recognition algorithms" are fast and there's no dictionary matching.
    2. The matching doesn't depend on having the right words in the dictionary at input time. It works just as well for unusual proper names and technical jargon as it does for common words, since they're all formed from the same basic phonemes.
    3. The longer the search string, the greater probability of an accurate match.
    4. No need for accurate search string spelling. It doesn't matter if you know how to spell a word, as long as you can write it down phonetically.

    In theory, the system should work for any language, but reality is that different languages do have different sets of phonemes, and I think Fast Talk has only really worked on English. So languages like Spanish that are fairly similar phonetically to English would probably work pretty well, but tonal languages like Mandarin Chinese or those with non-vocal sounds like the clicks and pops of the African Bushmen would require a rework of the phoneme recognition code.

    The main downside of their system is that it doesn't actually produce text... which means that you'd need another speech-to-text system if you wanted transcripts, or want the data to be searchable with whatever standard text-based search engine you are using on your intranet. But they appear to be aiming at applications where that's not necessary. One of my favorite ideas is integrating it with a video editing suite and being able to jump to different cues in your video clip library simply by stating the dialogue that's found there.

    Of course, one of the most obvious applications is for intelligence and security. So far it doesn't appear that the company is pushing too hard in that direction -- it was founded by an academic group that originally developed the technology for a library project at Georgia Tech. However, I'm betting that's where the real money is, and it's only a matter of time before their ideas are found in your favorite national department of big-brotherhood.

    -R

  • Actually, (Score:2, Informative)

    by voodoo1man ( 594237 )
    phonemes are abstract groupings of phones (the most basic units of human-made sound) that are differentiated from each other by the environment they occur in. Allophone (grouping of phones in a phoneme) distribution and even the basic phone set differ from language to language and even dialect to dialect rather significantly, and so this system would actually be pretty inefficient as a replacement for regular text searches (where, for the most part, spelling is pretty standardized across dialects). It is faster and more accurate because it bypasses text recognition, but existing text would have to be converted to phone representation with several dialects/pronounciations stored, or existing audio is going to have to undergo text/word recognition and the different pronounciations generated, for it to be used in a general purpose search engine.

    And this doesn't even begin to deal with "Engrish" speakers =]

  • How is this different from soundex [reference.com]? For decades, databases of names have been stored in soundex. If your driver's license number begins with letter-number-number-number, it is probably soundex. If you have done any ancestry searching, as I have, you have encountered soundex; this way, if you search for John Smith in 1732, you will find records for Juan Smyth, Jon Smythe, John Smitt, etc.

    The benefits of having actual sound? If it's just going to use a soundex-type formula in the core functioning, the sound would just be a gimmick, and a storage-taking one at that. Sure, compression has gotten amazing, but will the sound of Smith really take anything near the same 4 bytes as "S720" ??
  • Fast-Talk assumes you know the word you are looking for before you search, which is not super useful, except for google style searches, also it is extraordinarily slow (only searches about 10 hours of audio a second, think about how slow that is for say a year of NPR). Check out CallMiner [callminer.com] for a much cooler use of speech to text technology. CallMiner uses trending to find trends in large volumes of calls. a real business use.

  • potential for medical applications is exciting (Score:3, Interesting)

    by ColMstrd ( 103170 ) on Sunday December 22, 2002 @09:42PM (#4942401) Homepage Journal
    As 90% of the data for diagnosis comes from the history-taking (interviewing) the patient, the potential for automating/supporting diagnosis is exciting.

    Imagine a system that listens to a consultation in real time, making helpful suggestions for diagnosis based on analysis of the patient and the doctor's phoneme streams! And no tedious data entry, just an unobtrusive microphone.

    I've been waiting for this.

  • wordspotting (Score:2, Insightful)

    by GoBears ( 86303 )
    The basic idea of using audio similarity to "grep" short sounds out of audio streams (as opposed to using ASR and text-matching) is quite old - some classic papers based on dynamic timewarping date back to 1977, and HMMs became popular for this application about ten years after that. Papers on this kind of thing appear in conferences like ICASSP - look for keywords like "keyword spotting" or "wordspotting." The phone company wanted to do this for obvious reasons.

    Note that I'm not saying the GATech technology used by this company is derivative - I haven't looked at the specifics of this approach.
  • After reading a posting that someone had probably typed in silence and submitted to slashdot, I posted this reply silently and now you're reading it. Chances are you aren't reading this out loud. Nobody said a word, or even had to hear one. Reading can be an astoundingly efficient way to transfer information.

    I have yet to meet anyone in good health who prefers getting ten voice mails over ten emails.

    What the world needs is fewer karma whores and more good friends.
    Go ahead, friend. :) Click that white button and turn it green.

  • Old News (Score:2, Interesting)

    by annodomini ( 544503 )
    I saw a demonstration essentially the same technology at Compaq's CRL [compaq.com] about two and a half years ago (formerly DEC CRL, or Cambridge Research Lab, the guys who did research for AltaVista). It did exactly the same thing. It broke sound files down into phonemes, then searched based on the phoneme. It was mostly used for finding a clip on the web rather than a specific place in a long file, but it was the same idea. The nice thing was that it was OK for its application if it missed once or twice. If the audio file was relevant, the word or phrase was probably used multiple times in the clip. It was pretty good at finding NPR stories about certain events. In fact, you can try it out for yourself at an online demo [compaq.com].
  • Maybe there will even be an 'Audio' tab on Google.

    I think there should be three tabs instead of one 'Audio' one:

    • Speach - to search a text pattern in recognized speech;
    • Melody - to search among music records using querying by accords or by notes;
    • Noise - to search by comparing to audio fragment/pattern;
  • Possibly more effective would be to convert to Soundex. This is an old technique where similar phonemes are assigned the same "soundex" code. It can be used for comparing strings where you may not know the correct spelling. I suspect it could help in cases where the audio-to-phoneme conversion isn't quite perfect too.

    Paul

  • The wise programmer is told about the Tao and follows it. The average
    programmer is told about the Tao and searches for it. The foolish programmer
    is told about the Tao and laughs at it. If it were not for laughter, there
    would be no Tao.
    The highest sounds are the hardest to hear. Going forward is a way to
    retreat. Greater talent shows itself late in life. Even a perfect program
    still has bugs.
    -- Geoffrey James, "The Tao of Programming"

    - this post brought to you by the Automated Last Post Generator...

Slashdot Top Deals

The Tao is like a glob pattern: used but never used up. It is like the extern void: filled with infinite possibilities.

Close