AfterDawn: Tech news

Search within audio files

Written by Jari Ketola @ 22 Dec 2002 1:38 User comments (34)

Fast-Talk Communications is developing an interesting technology for searching text strings within audio files. Unlike audio search tools released so far, Fast-Talk's engine converts speech to phonemes, the basic units of sound in a language, and indexes them. A written search string is also converted to phonemes, then that string of phonemes is searched within the index, and the resulting timecodes are returned.
Fast-Talk offers an a trial evaluation kit of their engine. The company, however, is not looking to make a product for the end user, but instead offers the engine and the software development kit to companies willing to utilize the features in their own products.

Read this wonderful article on InfoWorld for a closer look at the technology.

Source:
Slashdot.org

Previous Next  

34 user comments

122.12.2002 23:42

I didn't delve too deeply into the text, but did you download the trial? Did you phoenetically search for any word(s) in a wave file? If so, did you find the audio portion(s) you were loking for? Does the program scan and then scoot off to that particular portion of the audio file and play back a sample? What are the limitations of the trial version? What colour is blue? Why is water wet? How many chucks could a wood-chuck chuck, if a wood-chuck could chuck wood? Tell me tell me, I have to know everything right now, immmediately. Please compile a complete, detailed analysis in response to the above questions, and leave the report on my desk. I'll pick it up first thing in the morning. -- Michael, the Insatiable but Lazy Klingon --

223.12.2002 06:54

Mike, No. I didn't. Awfully sorry, mate! *grin* But seriously.. It seems that there's no trial available for download. Atleast I couldn't see one with a quick peek at the site. As far as I could tell, it's only available on CD-R. I'm probably wrong, though, as always.. =)

323.12.2002 18:01

If they didn't require such a lot of user-info, (like exact mailing address, and intended uses of trial), I'd fill out the application page because they also mention that the software can be obtained via ftp as well as cd-rom. I.E. : >>> ..request...a trial....delivered via CD-ROM [[OR FTP dpwnload]]. <<< If I do get the thing though, do you want me to email you a copy? (Assuming the thing isn't megs and megs huge). Just a general perusal of the text tells me that the software can scan a wave file and seek out specifically-requested portions based on whatever phoenetic input you type. Worth the effort to pursue this, do you suppose? (The software authors may request or expect user-feedback from downloader by the looks of things). -- Mike --

423.12.2002 23:16

Thanks, but I guess I'll go ahead and request a trial copy of my own, if and when I have the time. I do find it rather an interesting concept, so I will try and find time to look into it further.

524.12.2002 00:42

(I wouldn't bother Jari, unless you want to be bugged). Mandatory fields included home phone number, snail-mail address, e-mail address, all of which I gave them. (And who said Klingons were bright?) Also, some fields gave limited choices, example: Possible Uses? [ ] Governmental [ ] Commercial. A third choice [ ] Simple Curiosity, not available; *only* those 2 choices available, and this field is mandatory too. (In other words, they don't want to hear from you unless you are either a government representative or potential commercial manufacturer). So, I told 'em "Government" anyway. :-) Should be interesting if they choose to call me via long distance voice. (Won't *they* be surprised!). Or else, an email, perhaps not leading to an ftp after all. Then, the last "killer" announcement right after you click the "Submit" button, "Your request has been sent. A Fast-Talk Representative [[let's hope he's not a 'fast-talking' rep ]] will contact you shortly with further details." (Oh, joy.) Like I say, you might wanna avoid this one, unless you're a sucker for punishment. :-) -- Klingy --

This message has been edited since its posting. Latest edit was made on 24 Dec 2002 @ 12:47

624.12.2002 07:31
8bit_bytr
Inactive

Got a message from their admin dept. saying I hadn't filled everything in, duh! I can't fill in everthing 'cos I live in the UK and not America. Told 'em I was a small comercial concern and am waiting for my link to be sent to me via email. If I ever get said link and If I eventually get the software I would gladly do a write-up for you :)

724.12.2002 10:39

Hi 8-bit. Well, I must say, this thing *does* look like it might be fun to play with! I just got a polite email thanking me for my interest (no pesky phone call or requests on their part). Included was a direct link that started my download right away, but *here's* the killer: I only have a slow, pokey, dial-up modem. The 'trial vrsion' of this search software is a single .exe file. Simple enough, but it is one-hundred and fifteen megabytes in size!!! (Oh geeze....) It comes with a built-in 11-minute 'travel video' of New York. Apparently, one can search through video/audio files as well as simple audio files. It would take me over 6 hours (at bst) to get this thing. (!!) Now would have been a great time to have that DSL connection I've always wanted. Since I 'plug-in' to the common household telephone here, I'll probably have to wait some night until everybody has sacked out for the night before I attempt the download. (Would have been simpler for them to keep the application and the video separate). I *may* attempt a download later, but with 4 copies of TMP running in the background right now, it's asking for trouble. <gg> If I get the file, (won't be right away) and it proves interesting enough, I'd be happy to snail-mail it on a cd to both of you. Partial Blurb: Thank you for your interest in Fast Talk. I am pleased to present you the Fast Talk Evaluation Kit link including a fully searchable 11 minute travel video featuring New York City. Fast Talk has created the world's first, fast and accurate, telephony/audio/video search engine. With our technology you can now monitor and/or search for _any_ spoken word, without having to first convert audio to text. This is not speech-to-text software, it is a phonetic search engine. Therefore, it is not restricted to a specific dictionary, voice, phrases, keywords, or dialect. With this technology you can now positively identify people's names, places, or even highly technical terms. (And it goes on a bit....) This applies to searching hundreds, or even thousands of hours of archived telephony, audio, or video. We have recently debuted the ability to monitor for a list of 25,000 words or phrases......We currently have the capability to search/monitor in English and Spanish, and we are currently porting to additional languages....... (etc. etc.) Sound interesting enough to pursue??? Cheers. -- Klingy --

824.12.2002 12:38
8bit_bytr
Inactive

Oh yes, :-) Playing with sound is a hobby of mine, I think a look at what they have done would be very interesting indeed. I was a bit dubious when I looked at the TV News Report that I downloaded, they claimed they could search ANY language. Even a savvy geek like me can see that the problem lays with pheonetics. Every language is different and they will have to create quite a library to make this thing work. Of course If I can get a look at the software there is a chance of understanding how it basicly works. Still waiting for my reply, got a similar email as you Klingy. Alas only got 56k modem as well. I live too far away from a Broadband Connection. Still, lets see who gets it first. If I manage to download it and it 'actually' works, I would be glad to snail mail you a copy on CD-Rom. Thanks :-)

924.12.2002 14:02

After reading these posts, I have decided to attempt to recieve the program. I have a cable modem so (IF I get the program) I will be able to give it to any who may desire. Perhaps through an P2P service, or I may just give the address and password of the email account I created to recieve the file, It will be of no use to me afterwards, so anyone could use it with no problem POSTED LATER: I just had a thought, they may find it odd that a senior marketing director was working on Christmas Eve

This message has been edited since its posting. Latest edit was made on 24 Dec 2002 @ 2:07

1025.12.2002 23:47

Jari or 8bit_bytr or msb5150 : I got the download - InstallSearchDemo_2_0.exe - 121,293,045 bytes. Haven't had much opportunity to play with it though. Installs quickly - no rebooting necessary. Includes searchable .mpg video file, "Destination New York", 11 mins :20 secs. Movie appears in it's own window. You can enter phoenetic search phrases in box beneath viewing screen, and below that, a scrollable window returns your 'hits', and their timed-positions within the video file (mins/secs). Clicking on a 'hit' will take you directly to that part of the video, then begin playback. Every search I performed returned several (many) hits, but I didn't seem to _hear_ the phrases/words I was looking for, however, I am unaccustomed to this program and so, must refine my search technique. My search 'words' were very brief, and the actual word(s) probably flew past so quickly that I didn't recognize them. Longer phrases (2 or more phoenetic 'words') will produce more accurate results, e.g. : "wurrled trayed senturr" (or something like that) :-) There are built-in 'Tips' and user hints, both simple and advanced. That's about as far as I got for now. Have not tried any external, independent audio or video files. If any of you three want to PM me your snail mail, I'll burn a copy of the demo for you. (Please, no other user-requests just for now, ok?) Thankee. -- "myke tha kleen gonn" --

1126.12.2002 04:26
8bit_bytr
Inactive

Sorry, I'm a bit thick :( how would you like me to send you my address?

1226.12.2002 07:17

Do I assume that I am one of the three. If so, thankyou

1326.12.2002 09:54

Hello again, 8-bit. :) When I say "PM me a message", that means you can send me a <P>rivate <M>essage right here in the forum. You have to be logged-in first. Look up above near the top of the screen, and you will see a link, "Private Messages". (Took me a while to get used to it too). <gg> Anyway, now that I have had a better chance to play with the program, I have found that you *cannot* use any of your own audio and/or audio/video files for searching. The file they include is "Discover New York", a PAL .mpg video, and it is the *only* file you can search through. This truly is a DEMO program only! The file resides in the c:\\Program Files\\FastTalk\\SearchDemo\\ folder, and is called, "DNY_clips.mpg". I deleted the file, then substituted one of my own vcd files, then renamed it to the original file's name. Didn't work. Got the error message, "Invalid PKG file". (Bummer, bummer). I'll hang on to the file for you if you still want to send me a PM, but otherwise I intend to delete it. Let me know. -- Mike --

1426.12.2002 14:39
8bit_bytr
Inactive

'tis done, sent a mail direct via PM, you should have it as I write this. Cheers

1526.12.2002 15:39

Got it. Responded! -- Mike --

1627.12.2002 17:35

I got the email, very interesting, I could post the download link if anyone might like it.

1727.12.2002 17:45

Here's the dl link: ne 1 w/dialup will need 2 pull an allniter ftp://fasttalkdemo:fe214@www.fast-talk.com/demo/InstallSearchDemo_2_0.exe -jonselpsy

1827.12.2002 17:50

Nevermind, jonselpsy beat me to. Ce la vie

1927.12.2002 19:27

Well, if anyone has a 56k modem connection, they're looking at anywhere between 6 and 8 hours download time (I suggest using some form of file-download resumer!!) just to get a demo that won't let you search through any of your *own* files! (Unless someone can think of a crack). :-) -- Mike --

2029.12.2002 15:12

The email stated that a better trial version was available that would allow someone to use the product with their own media, perhaps someone could sweettalk the guy into sending them a copy

2129.12.2002 15:17

This is the email that was sent to me, some information has been left out(ie: the name I'm using) It has been brought to my attention to advise users, before you experience our technology, of the following; The Evaluation Kit will return "hits" even if you have chosen a word or phrase that does not exist in the audio. These "hits" on incorrect phrases is simply due to our Evaluation Kit having no minimum confidence level preset. Please be aware that our SDK can be fully configured to not return sub-optimum confidence scores. Thank you. Mr. (My fake name), Thank you for your interest in Fast-Talk. I am pleased to present you the Fast-Talk Evaluation Kit link including a fully searchable 11 minute travel video featuring New York City. Fast-Talk has created the world's first, fast and accurate, telephony/audio/video search engine. With our technology you can now monitor and/or search for any spoken word, without having to first convert audio to text. This is not speech-to text software, it is a phonetic search engine. Therefore, it is not restricted to a specific dictionary, voice, phrases, keywords, or dialect. With this technology you can now positively identify people's names, places, or even highly technical terms. This applies to searching hundreds, or even thousands of hours of archived telephony, audio, or video. We also have recently debuted the ability to monitor for a list of 25,000 words or phrases in real-time alerting fashion with our new Monitoring Services Framework. We currently have the capability to search/monitor in English and Spanish, and we are currently porting to additional languages. Please take a moment to visit our website for our latest news and technologies. A Fast-Talk Technology link to the Evaluation Kit is below. Here are some areas where Fast-Talk technology can be applied. surveillance knowledge management real-time monitoring and alerting call center management audio data mining searchable voice mail digital asset management distance learning training auditing and rights compliance non-linear editing depositions transcript synchronization If you believe you have an application or use for Fast-Talk, we do offer a more comprehensive evaluation kit. It will allow you to search your own audio to further test our capabilities. Please feel free to contact me, by phone or email, to discuss your application and market potential. I will be happy to deliver the more interactive, advanced evaluation kit to qualified applicants. Click the below link to download Fast-Talk's audio search engine. In addition, I have attached a Quick Start Guide to help you navigate your way through the kit. SEARCH DEMO(the link can be found above) Best regards, Jason Jason B. Witzke Sales Associate Fast-Talk "The Power To Search Audio" One Buckhead Plaza 3060 Peachtree Rd. NW Suite 730 Atlanta, Georgia 30305 Direct: 404-495-7234 Fax: 404-495-7221 Cell: 770-722-1017 http://www.Fast-Talk.com

This message has been edited since its posting. Latest edit was made on 29 Dec 2002 @ 3:22

2229.12.2002 15:19

Any ideas of what I could say to him are accepted, if I can get the more addvanced trial version I will provide it to all in this forum who ask(somehow, for my cd burner is down now)

2329.12.2002 15:24

I hadn't heard of a 'more advanced' version of Fast Talk - my email was identical to the one above. But if you're still out there 8-bit, I posted your copy off last Saturday (cd-rom); it should arrive next week. -- Mike --

2429.12.2002 15:57
8bit_bytr
Inactive

Thanks :) careful how we go for this one, they are bound to get what we are doing sooner or later, then we will be fu****. From what I understand, they provide a software engine to be intergrated into actual software, so their own product is going to be quite a bit cut-down. Had no reply myself yet since, damn web page would not work properly, maybe their server was playing-up?, still I will see how I get on with the demo, you never know I may even find a proper use for it ;) Depending on how good these people are they may have the demo file well stitched into the demo, the fact that it is a separate file however leaves it open to be examined and I will draw my own conclusions from what I find. Rest assured what ever I find I will either post here, or start a new thread with the info.

2529.12.2002 18:46

Klingon: the 'more advanced version' is refered to here in the above copy of the email I recieved: "If you believe you have an application or use for Fast-Talk, we do offer a more comprehensive evaluation kit.It will allow you to search your own audio to further test our capabilities." I would like to try to get that version.

2629.12.2002 19:35

Just asking does the product in this link http://www.scansoft.com/audiomining/ use the same or similar technology as fast-talk?

2729.12.2002 21:02

Oops. Right you are, msb; I must have missed that. I bet, though, unless you can provide them with explicit details of your 'proposed application', they'll be unlikely or unwilling to give you the other version. (Seems to me they are looking for potential business). -- Mike --

2830.12.2002 08:42

I've got nothing to loose, unless TRYING to get software can be construed as illegal, besides the only thing tying me to them is maybe an IP address for the email service I used, everything else I made up.

296.1.2003 14:14
8bit_bytr
Inactive

Got my CD from Klingy today. Had a quick look, it would appear that there is a file called 'SearchableCD.pkg'. This is a key-generated file and any alteration to it stops the prog in it's tracks. Likewise if you change the name of the file they supply. This would suggest that there IS a workaround if I can reverse-engineer the key. This could of course be half the story and they may have embedded more code into the soft itself. I will take a look at the prog as well and see if I can bypass the 'lock'. No promisies though :(

306.1.2003 15:33

-->>> ... this would suggest that there is a workaround ... <<<-- Ouuuuu... I detect a little 'hacker' in your blood. (Hackers are my *favourite* people!) :-) Sherlock Holmes would have been proud of you, Matt. ("Elementary, my dear Klingon"). There are, fortunately, a lot of individual files created, so perhaps something can be done to mod one or more of them. (Got your PM; & responded). -- Klingy --

3112.1.2003 01:07
8bit_bytr
Inactive

Hi, I have found another signature file *inside* the .EXE main program file *SearchableCD.exe* when I alter it the program still runs but will not load the Mpeg file, it says that it is not present. But the software still loads, so I must be on the right track? I will keep you gus posted :)

3212.1.2003 09:36

Take your time. The proggy might not even work all that well anyway. Did you do any searches with the default file? It <sorta> works, but the part in the video is found so fast (and then keeps on going), that I'm not really sure if my spot was found or not!. (weird). -- Mike --

3312.1.2003 10:23
8bit_bytr
Inactive

yes it does work, quite well in fact. Longer phrases work best, but only a few words at a time, like *New York* or *Good Morning*, that sort of thing. Must sort my typing out, no-one called gus here ;) should read guys in last post.

3412.1.2003 19:04

nOttowoRRy.i hAvebUilT-iNerRoR-cOrrEctIonWHenReADiNGtEXtfIleS. -- akLIngOn --

Comments have been disabled for this article.

News archive