speech recognition equipment/software to transcribe recordings

[TreeTops]

I'm with Lordsmurf. You are trying to do something yourself that you know nothing about. Sure you can learn how to do it if you have the money and time. (Especially the time as you have stated several times that you don't have the funds for the viable suggestions offered.) Your project is a big one and you should hire a pro to do it.

[lordsmurf]

Indeed, it is my PR background that causes me to advise an organization to not do a cheap slipshod in-house hack job, but rather seek the professional quality needed by any group that respects its own work. If the sessions are important enough to record and transcribe, it's probably worth doing correctly.

Not knowing the equipment will make you look stupid to your internal demographics. And then the likely-lousy output will make you look amateur to external demographics.

And since you seem intent on focusing on use of language, know that the word "****" does not make angels cry and kittens die. If anything, being so easily offended at common language would suggest you're the one who's not reading. Pick up any number of magazines, and you'll see "curse words" in use, because they have power to convey where passive "clean" language does not. I'm not talking Playboy here, but magazines like Harper's and The Atlantic, or even many technology magazines. Ironically, I don't think Playboy uses many colorful words in articles -- been a while since I've seen an issue, though.

Seeing how I'm not paid to post here, and there's no copy editor on staff, I'm allowed to have typos. Don't try to school me in writing.

[Cornucopia]

Ok, now I'm a little confused...

You seem to be trying to put the cart before the horse here.

My take on the overall point to all this is to get a CLEAN and CLEAR set of recordings to enable transcription.
The first items needed for this is a high quality recording & playback chain from start to finish. This includes:
1. Quiet and non-echoey Room (and environment beyond) where the recording will take place
2. Talent & their habits & circumstances (and if they can take turns, loudness levels, etc..) that record well without interference
3. Mike quality, placement and specs
4. Mike pre-amp quality
5. Lack of a bunch of intervening equipment which can get in the way
6. Quality A-to-D converter
7. Appropriate buffering and storage pipeline
8. Decent recording software (that's the easiest part)

Then:
9. Decent playback software (also fairly common)
10. Quality D-to-A card
11. Quality Amplifier
12. Quality Speakers/Headphones
13. A listening space with little to no distractions

What I was assuming was that you understood & had accounted for items 1 & 2 & 5 & 13, and that you could fairly easily acquire (or already had) 9 & 10 & 11 & 12. The usage of such a device as the Zoom R16 should have covered items 4 & 6 & 7 & 8. So all you needed then was 3. WHICH HAPPENS TO OFTEN BE THE MOST IMPORTANT LINK IN THE CHAIN.

You are probably going to have speakers which DO overlap and which DON'T have time for 5-10 minute training speeches. It is kind of unreasonable to expect them to strictly follow turn-taking (it isn't natural in most societies when there is an ANIMATED or IMPORTANT impromptu discussion). So one of your HIGHEST PRIORITIES is sound clarity and isolation.

For clarity, you need fairly high quality mikes (condensers, not dynamics). For isolation, you need the placement and mike specs to support this. Lavs help with placement (and proximity effect) - my faves in this area are Sennheiser & Sony wired lavs, Cardioid polar pattern enhances this isolation capability. Omni polar pattern is common in lavs, but won't ENHANCE the isolation that you sorely need. THIS IS NOT TO SAY THEY ARE OF LOWER QUALITY, just not best for the job at hand.

To further enhance isolation, I've suggested Multitrack recording. This gives each speaker their own DEDICATED channel (it also allows for a certain amount of "training" accommodation and allow the possibility of multiple simultaneous transcribers). The idea of using low quality voice recorders both DEFEATS the purpose of multitrack AND the isolation & clarity factors.

I'm beginning to think that you haven't thought this through. There are many options here, but you are the one who is going to have to decide WHAT the priorities are. I know what I would do (and probably how much it would cost). You sound like you still haven't decided which path to take because you talk about taking one path and then refer to things that exist on a different path from it.

WHAT IS YOUR TIMELINE?
WHAT IS YOUR BUDGET?
WHAT ARE YOUR EXISTING RESOURCES (in hardware, software, manpower, skillsets, etc)?
WHICH ARE YOUR PRIORITIES?
(these are the kinds of things that are best laid out at the outset, so you and we will know what arena you are working in)

It's getting close to decision time, and to make the best decision requires some understanding of pros and cons, advantages and pitfalls, rules of thumb, etc. If those aren't already under your belt, it might be better to hire someone else to run this, like as has been mentioned before...

Good luck,

Scott

[chowmein]

Also, how many in your group discussion?

If you're going to have each speaker on their own track you need an appropriate A/D interface. Over 8 kinda gets expensive.

If you have all the speakers on a single track, that's fine if you can easily identify their voice I suppose. The main issue would be good mic'ing (and mixing), as has been said. Actually, the main issue is transcribing the entire discussion (yuck!)... How long is the discussion going for?

[p_s_92]

My take on the overall point to all this is to get a CLEAN and CLEAR set of recordings to enable transcription.
The first items needed for this is a high quality recording & playback chain from start to finish. This includes:
1. Quiet and non-echoey Room (and environment beyond) where the recording will take place
2. Talent & their habits & circumstances (and if they can take turns, loudness levels, etc..) that record well without interference
3. Mike quality, placement and specs
4. Mike pre-amp quality
5. Lack of a bunch of intervening equipment which can get in the way
6. Quality A-to-D converter
7. Appropriate buffering and storage pipeline
8. Decent recording software (that's the easiest part)

Then:
9. Decent playback software (also fairly common)
10. Quality D-to-A card
11. Quality Amplifier
12. Quality Speakers/Headphones
13. A listening space with little to no distractions

What I was assuming was that you understood & had accounted for items 1 & 2 & 5 & 13,

Yes, I did.

and that you could fairly easily acquire (or already had) 9 & 10 & 11 & 12.

I don't have 10 and 11. For 9, can I use Videolan, Media Player Classic? For 12, we have the speaker set which come with most desktops and we do have a headphone too.

For 10 and 11, what would you suggest?

For clarity, you need fairly high quality mikes (condensers, not dynamics). For isolation, you need the placement and mike specs to support this. Lavs help with placement (and proximity effect) - my faves in this area are Sennheiser & Sony wired lavs, Cardioid polar pattern enhances this isolation capability. Omni polar pattern is common in lavs, but won't ENHANCE the isolation that you sorely need. THIS IS NOT TO SAY THEY ARE OF LOWER QUALITY, just not best for the job at hand.

To further enhance isolation, I've suggested Multitrack recording. This gives each speaker their own DEDICATED channel (it also allows for a certain amount of "training" accommodation and allow the possibility of multiple simultaneous transcribers). The idea of using low quality voice recorders both DEFEATS the purpose of multitrack AND the isolation & clarity factors.

Ok, now I get it. I am completely new to this so did not know the reason for each equipment. Now, I know why you suggested something and why that is essential for this project.

I'm beginning to think that you haven't thought this through. There are many options here, but you are the one who is going to have to decide WHAT the priorities are. I know what I would do (and probably how much it would cost). You sound like you still haven't decided which path to take because you talk about taking one path and then refer to things that exist on a different path from it.

I haven't thought this through completely because until now I was unaware of what equipment would I need, how to configure and use it. So, I was confused and referred to paths in different directions.

WHAT IS YOUR TIMELINE?

We have about 2 months.

WHAT IS YOUR BUDGET?

Not fixed yet, but around 2500(give or take). I assume with the R16 costing some 400, eight lav mikes some 200 meaning 1600 for mikes. For 10. Quality D-to-A card and 11. Quality Amplifier what would you suggest as some models I can look into and how much would they cost? We have about 500 left out of the 2500 so if we can get 10 and 11 using that, that is great.

WHAT ARE YOUR EXISTING RESOURCES (in hardware, software, manpower, skillsets, etc)?

Hardware we have laptops, desktops, desktop stereo speakers, headphones, wireless headset.
For software we have/can get the free/open source tools like VideoLan, Media Player Classic, Audacity, n Track(we have to purchase if it will be required for 64), Windows Media Player(which may not be needed)
Manpower: We have 5 people who will be helping us out who go to college full time and they work for our office as interns. They are computer engineering majors so are fairly tech savvy and skilled with computers. Even, if they have not used a software, they can learn it quickly. Setting up/Configuring hardware they can learn, once they have been shown or by reading manuals.

WHICH ARE YOUR PRIORITIES?

To get this recording of multiple people done smoothly so that less amount of work needs to be done while transcribing. That is why I was thinking of Dragon. So, if 7 people are in a discussion and we have recorded them on separate tracks, if we can get separate recordings for each individual, Dragon(after the 10 minute training with each speaker's voice/accent) can get us some 60-70% accuracy in transcribing their voice to text. The rest 30-40% has to be done manually using the software you suggested which involves foot/pedal to control the recording speed.

(these are the kinds of things that are best laid out at the outset, so you and we will know what arena you are working in)

It's getting close to decision time, and to make the best decision requires some understanding of pros and cons, advantages and pitfalls, rules of thumb, etc. If those aren't already under your belt, it might be better to hire someone else to run this, like as has been mentioned before...

Now, I hope our needs are a bit clearer.

Hiring someone is ruled out for few reasons. I don't know what would be a reasonable rate for people who do such recordings as I am new to this area, but the recording people we have talked to(I live near Jacksonville, Florida) quoted 100$/hour or for a two hour recording 200. We need to do such a recording every month. So, in a year it would be 2400. They were willing to reduce the price only if we did 5 hour recordings every month with them for which they were ok with 80$/hour. We don't need 5 hour recordings every month. Also, some months we may need a extra couple of hours so if we assume 30 hours/year instead of 24hours/year counting 2hrs/month it would be around 3000. The cost would remain same every year or increase if the recording people hike their rates. Buying our own equipment and getting our staff trained would involve a one time expense, but that is something which we won't have to incur every year. We don't need transcription in real time so as you suggested we don't need to pay the rates of court transcribers.

Next issue is timing. With professional recorders, we may have to book them weeks in advance or they may not be available. If we are doing it in house, we have more flexibility.

Yes, we at present don't have the equipment or skill to do it, but thanks to you I now have a much better idea on what equipment is needed, how it will fit in the scenario. Once, we get the equipment, reading manuals/using this forum or web we can learn how to do it. Our first recordings may not be stellar and we may have to do some trials before we get a good quality, but we are OK with that.

Good luck,

Scott

Thanks a lot for the wishes, help and time. That has been very useful indeed.

[Cornucopia]

10 & 11 can often depend on who your transcriber is, how good their hearing is, and in what environment they're listening. If these are bad, you'll need a separate high-quality playback chain (using the output end of a number of those multi-track interface units), hooked up to a true HiFi stereo Amp/Reciever & Musician/Engineer-quality speakers (or a PREMIUM set of over-the ear headphones). If these are all good, you can probably get by with your PC's regular soundcard and a set of DECENT On-The-Ear headphones.

For 9, you could use those apps, but they DON'T specifically support multiple speeds and the use of a footpedal, etc. I'm sure they don't support speed changes with NO pitch change. So they aren't the best app for the job. Usable, but not optimal.

I'll RARELY recommend a particular brand or model of physical equipment, because I don't know if you could get it in your area for your budget and in your timeline. Also, there are a range of pieces of equipment for a lot of this stuff, but knowing why something is good or not has to do with understanding all the parameters of your situations requirements and how they interact.
You seem to be in a little over your head, so I've suggested you HIRE someone who can set this up for you (whether doing it themselves, or setting it up so your people can do it), someone who has a background in audio recording and also in transcription. You need someone local/regional also.

BTW, unless you have some STRONG REQUIREMENT that things be mobile or not showing wires, I recommend you AVOID anything wireless (whether in the recording or playback chain). That adds extra complexity and possibly noise+distortion, and you don't need that.

Your quoted rate of $100/hr isn't too far off. You might get someone down to $50/hr, or as high as $150-200/hr. Lower tells me they might not fully know what they're doing or what they're getting themselves into. Higher is overkill. But that's for SETTING UP & RUNNING the recording/editing. Unless they're familiar with transcribing, they shouldn't be quoting the same rate for both things as they're 2 different beasts.

I don't know what kind of event(s) you're expecting to be capturing, but in all my years, I've NEVER seen participants spend their time at the beginning training speech recog software. If you need ~10 min. each, that's 40-80 minutes alone (depending on how many participants you've got) where ALL but one will be waiting around. Talk about a lesson in frustration!

Scott

[neomaine]

... Hiring someone is ruled out for few reasons. I don't know what would be a reasonable rate for people who do such recordings as I am new to this area, but the recording people we have talked to(I live near Jacksonville, Florida) quoted 100$/hour or for a two hour recording 200. We need to do such a recording every month. So, in a year it would be 2400. They were willing to reduce the price only if we did 5 hour recordings every month with them for which they were ok with 80$/hour. We don't need 5 hour recordings every month. Also, some months we may need a extra couple of hours so if we assume 30 hours/year instead of 24hours/year counting 2hrs/month it would be around 3000. The cost would remain same every year or increase if the recording people hike their rates. Buying our own equipment and getting our staff trained would involve a one time expense, but that is something which we won't have to incur every year. We don't need transcription in real time so as you suggested we don't need to pay the rates of court transcribers.

Next issue is timing. With professional recorders, we may have to book them weeks in advance or they may not be available. If we are doing it in house, we have more flexibility.

Yes, we at present don't have the equipment or skill to do it, but thanks to you I now have a much better idea on what equipment is needed, how it will fit in the scenario. Once, we get the equipment, reading manuals/using this forum or web we can learn how to do it. Our first recordings may not be stellar and we may have to do some trials before we get a good quality, but we are OK with that.

Thanks a lot for the wishes, help and time. That has been very useful indeed.

I'm going to reiterate something I recommended towards the beggining, especially with your cost estimates and budget laid out:

1 - $100/hr for this work is dirt cheap and very cost effective.

2 - This is going to cost you A LOT more than $2500/yr.

I'm not sure if you're the one setting your budget but its not adequate. VERY close if you use the pros that you've mentioned but not even close (I'd estimate at least double for equipment/software) for the road you're going down.

Appearantly you have no cost associated with training, using, updating, adjusting the potential equipment you're going to buy. Don't know of any manager who doesn't associate employee time with a cost. There is NO SUCH THING as a one time cost. You actually have people who have 20-40 hours available per month?

My recommendation: Hire the professional for 2 to 3 months to get things going. A very low cost POC (proof of concept) before you jump into something you're not experienced with. You may, no, WILL find that this is outside of your experience and the learning curve, upfront and ongoing cost in both materials and employee time. Have you (can you?) estimate the employee time needed if you do this in house? Can you estimate it? What happens if your transcribing need goes away and you're stuck with the equipment?

I'm hoping that you're starting to see the picture.

Case in point: I recently replaced all (Ok, still two rooms left...) the carpeting in my home. Zero experiance. But I had two things in my favor:

1 - I'm good with my hands, I'm generally a detailed person and I like to learn.

2 - When pricing flooring, we wanted hard wood not the laminate, we also priced installation. Bam, 2x the price for installers.

3 - I had plenty of time. Not selling my home. Just got a bigger than expected check from Uncle Sam and decided to help my ashtma out and replace the carpet. (The difference is huge btw...)

However, we have a very good friend in the flooring business. He found us great quality prefinished flooring at a good price. Roughly $900/room on average. Perfect fit for the budget. He also said he'd help us start out.

Now the basic installation was actually pretty easy. Borrowed a compressor and nailer from a fellow referee and we went to town. He stayed with me a few hours giving me a heads up on where to start, using splines to reverse if we needed, how to do corners and closets, spacing, how/when to use glue and/or nails instead of the nailer. All things I had no idea and would have to learn the hard way.

When asking him how much the installation would be thourgh him he said to essentially double the cost. And when I asked him how long it would take them, he said two days ... for all 5 rooms!?! Well, we're 5 months into it now and we don't mind it for the cost its saved us and the reward its given us.

The moral of the story: Make sure you understand what your undertaking and what the skill sets are that are involved. My wife and I where ready in a heartbeat to have the installers come do a room at time as we could afford it IF we found it was something we couldn't handle.

Get the pros, see what's entailed, make an informed decision.

(Sorry, kinda rambled on there...)

[p_s_92]

10 & 11 can often depend on who your transcriber is, how good their hearing is, and in what environment they're listening. If these are all good, you can probably get by with your PC's regular soundcard and a set of DECENT On-The-Ear headphones.

Yes, they are all good. Now, can you suggest something?

For 9, you could use those apps, but they DON'T specifically support multiple speeds and the use of a footpedal, etc. I'm sure they don't support speed changes with NO pitch change. So they aren't the best app for the job. Usable, but not optimal.

Yes, I realize I have to use some software which has Dictaphone/foot pedal.

I'll RARELY recommend a particular brand or model of physical equipment, because I don't know if you could get it in your area for your budget and in your timeline. Also, there are a range of pieces of equipment for a lot of this stuff, but knowing why something is good or not has to do with understanding all the parameters of your situations requirements and how they interact.

The situation would be recording of a discussion done by 7, 8 people. There won't be any technical/scientific jargon used. The area where the discussion would be done would be a quiet place without any disturbance/background noise. This discussion will be archived later on so it is not that this transcription/recording will go live(in which case it needs to be of highest quality). We need quality, but not like that of a theater level.

Now, can you please suggest something for 10 & 11. We may or may not be able to get that, but we know what is a good brand, price range for it and so on which gives us some idea.

You seem to be in a little over your head, so I've suggested you HIRE someone who can set this up for you (whether doing it themselves, or setting it up so your people can do it), someone who has a background in audio recording and also in transcription. You need someone local/regional also.

I don't make those decisions(whether to hire someone or not). The office does not want to hire someone because of cost reasons and scheduling issues(recording people need to be booked well in advance and cannot come at a short notice). One of the college interns has a friend who used to work in the recording field at some point. The college intern has mentioned that he will try his best to have that friend of his come to set it up for us the first day and show how it is done.

BTW, unless you have some STRONG REQUIREMENT that things be mobile or not showing wires, I recommend you AVOID anything wireless (whether in the recording or playback chain). That adds extra complexity and possibly noise+distortion, and you don't need that.

Thanks, will keep that tip in mind.

I don't know what kind of event(s) you're expecting to be capturing, but in all my years, I've NEVER seen participants spend their time at the beginning training speech recog software. If you need ~10 min. each, that's 40-80 minutes alone (depending on how many participants you've got) where ALL but one will be waiting around. Talk about a lesson in frustration!
Scott

We just need to arrange it so that other participants don't wait idly when one person trains the speech recognition software with their voice for 10mins. The others could do something like read the topic which they are expected to discuss, rules on how to discuss, introduce themselves to others and so on.

Thanks a lot for your advice and time. That is very helpful.

[p_s_92]

I'm going to reiterate something I recommended towards the beggining, especially with your cost estimates and budget laid out:

1 - $100/hr for this work is dirt cheap and very cost effective.

2 - This is going to cost you A LOT more than $2500/yr.

I'm not sure if you're the one setting your budget but its not adequate.

I don't set the budget.

VERY close if you use the pros that you've mentioned but not even close (I'd estimate at least double for equipment/software) for the road you're going down.

Yes, but with pros you face the expense every year whereas getting equipment/software is done once. Even if it is 5K as you estimate, it would be recovered in 2 years. Plus, pros need to be booked well in advance else we may not get them.

Appearantly you have no cost associated with training, using, updating, adjusting the potential equipment you're going to buy. Don't know of any manager who doesn't associate employee time with a cost. There is NO SUCH THING as a one time cost. You actually have people who have 20-40 hours available per month?

The college interns are working for college course credit. They are not compensated by us and they are part of another office and so are available to us for this project without any cost to us.

My recommendation: Hire the professional for 2 to 3 months to get things going. A very low cost POC (proof of concept) before you jump into something you're not experienced with. You may, no, WILL find that this is outside of your experience and the learning curve, upfront and ongoing cost in both materials and employee time. Have you (can you?) estimate the employee time needed if you do this in house? Can you estimate it? What happens if your transcribing need goes away and you're stuck with the equipment?

I'm hoping that you're starting to see the picture.

I am seeing the picture, but the person who run the office are not convinced. The need will not go away as things stand at present. We will try to get the college intern's friend for the first couple of recordings just so that he helps us do it well for the first 2-3 recordings.

The moral of the story: Make sure you understand what your undertaking and what the skill sets are that are involved. My wife and I where ready in a heartbeat to have the installers come do a room at time as we could afford it IF we found it was something we couldn't handle.

Get the pros, see what's entailed, make an informed decision.

I get the idea. If the college intern's friend does not come for some reason, I can request to get a pro for first couple of recordings.

Thanks for your tips.