elplaguister/DTS_line_datasets · Datasets at Hugging Face

topic_shift bool 2 classes	utterance stringlengths 1 7.9k	session_id stringlengths 7 14
false	So , yeah.	QMSum_120
false	Mm - hmm.	QMSum_120
false	Oh. Oh , so this i is this the difference here , for that ?	QMSum_120
false	No. This is the signal. This is the signal.	QMSum_120
false	I see that. Oh , yeah.	QMSum_120
false	The frame.	QMSum_120
false	Oh , that 's the f the original.	QMSum_120
false	Yeah.	QMSum_120
false	This is the fra the original frame.	QMSum_120
false	So with a short frame basically you have only two periods	QMSum_120
false	Yeah.	QMSum_120
false	and it 's not not enough to to have this kind of neat things.	QMSum_120
false	Mm - hmm.	QMSum_120
false	Mm - hmm.	QMSum_120
false	Yeah.	QMSum_120
false	But	QMSum_120
false	And here No , well.	QMSum_120
false	Yeah. So probably we 'll have to use , like , long f long frames. Mm - hmm.	QMSum_120
false	Mm - hmm.	QMSum_120
false	Hmm.	QMSum_120
false	Oh.	QMSum_120
false	Mmm.	QMSum_120
false	That 's interesting.	QMSum_120
false	Yeah , maybe. Well , I mean it looks better , but , I mean , the thing is if if , uh if you 're actually asking you know , if you actually j uh , need to do place along an FFT , it may be it may be pushing things.	QMSum_120
false	Yeah.	QMSum_120
false	And and , uh	QMSum_120
false	Would you would you wanna do this kind of , uh , difference thing after you do spectral subtraction ?	QMSum_120
false	Uh , maybe.	QMSum_120
false	No. Maybe we can do that.	QMSum_120
false	Mmm.	QMSum_120
false	Hmm. The spectral subtraction is being done at what level ? Is it being done at the level of FFT bins or at the level of , uh , mel spectrum or something ?	QMSum_120
false	Um , I guess it depends.	QMSum_120
false	I mean , how are they doing it ?	QMSum_120
false	How they 're doing it ? Yeah. Um , I guess Ericsson is on the , um , filter bank ,	QMSum_120
false	FFT. Filter bank ,	QMSum_120
false	no ? It 's on the filter bank ,	QMSum_120
false	yeah.	QMSum_120
false	so. So , yeah , probably I i it Yeah.	QMSum_120
false	So in that case , it might not make much difference at all.	QMSum_120
false	Seems like you 'd wanna do it on the FFT bins.	QMSum_120
false	Maybe. I mean , certainly it 'd be better.	QMSum_120
false	I I mean , if you were gonna uh , for for this purpose , that is.	QMSum_120
false	Mm - hmm.	QMSum_120
false	Yeah.	QMSum_120
false	Mm - hmm.	QMSum_120
false	Yeah. OK.	QMSum_120
false	Mmm.	QMSum_120
false	What else ?	QMSum_120
false	Uh. Yeah , that 's all. So we 'll perhaps try to convince OGI people to use the new the new filters and Yeah.	QMSum_120
false	OK. Uh , has has anything happened yet on this business of having some sort of standard , uh , source ,	QMSum_120
false	Uh , not yet	QMSum_120
false	or ?	QMSum_120
false	but I wi I will call them and	QMSum_120
false	OK.	QMSum_120
false	now they are I think they have more time because they have this well , Eurospeech deadline is over	QMSum_120
true	When is the next , um , Aurora deadline ?	QMSum_120
false	and It 's , um , in June. Yeah.	QMSum_120
false	June.	QMSum_120
false	Early June , late June , middle June ?	QMSum_120
false	I don't know w	QMSum_120
false	Hmm.	QMSum_120
false	Hmm.	QMSum_120
false	OK. Um , and he 's been doing all the talking but but these he 's he 's , uh	QMSum_120
false	Yeah.	QMSum_120
false	This is this by the way a bad thing. We 're trying to get , um , m more female voices in this record as well. So. Make sur make sure Carmen talks as well. Uh , but has he pretty much been talking about what you 're doing also , and ?	QMSum_120
false	Oh , I I am doing this.	QMSum_120
false	Yes.	QMSum_120
false	Yeah , yeah. I don't know. I 'm sorry , but I think that for the recognizer for the meeting recorder that it 's better that I don't speak.	QMSum_120
false	Yeah , well.	QMSum_120
false	Because	QMSum_120
false	You know , uh , we 'll get we 'll get to , uh , Spanish voices sometime , and we do we want to recognize , uh , you too.	QMSum_120
false	After the after , uh , the result for the TI - digits on the meeting record there will be foreigns people.	QMSum_120
false	Yeah , but	QMSum_120
false	Oh , no.	QMSum_120
false	We like we we 're we 're w we are we 're in the , uh , Bourlard - Hermansky - Morgan , uh , frame of mind. Yeah , we like high error rates. It 's	QMSum_120
false	Yeah.	QMSum_120
false	That way there 's lots of work to do. So it 's Uh , anything to talk about ?	QMSum_120
false	N um , not not not much is new. So when I talked about what I 'm planning to do last time , I said I was , um , going to use Avendano 's method of , um , using a transformation , um , to map from long analysis frames which are used for removing reverberation to short analysis frames for feature calculation. He has a trick for doing that involving viewing the DFT as a matrix. Um , but , uh , um , I decided not to do that after all because I I realized to use it I 'd need to have these short analysis frames get plugged directly into the feature computation somehow	QMSum_120
false	Mm - hmm.	QMSum_120
false	and right now I think our feature computation is set to up to , um , take , um , audio as input , in general. So I decided that I I 'll do the reverberation removal on the long analysis windows and then just re - synthesize audio and then send that.	QMSum_120
false	This is in order to use the SRI system or something. Right ?	QMSum_120
false	Um , or or even if I 'm using our system , I was thinking it might be easier to just re - synthesize the audio ,	QMSum_120
false	Yeah ?	QMSum_120
false	because then I could just feacalc as is and I wouldn't have to change the code.	QMSum_120
false	Oh , OK. Yeah. I mean , it 's um , certainly in a short short - term this just sounds easier.	QMSum_120
false	Uh - huh.	QMSum_120
false	Yeah. I mean , longer - term if it 's if it turns out to be useful , one one might want to do something else ,	QMSum_120
false	Right. That 's true.	QMSum_120
false	but Uh , uh , I mean , in in other words , you you may be putting other kinds of errors in from the re - synthesis process.	QMSum_120
false	But e u From the re - synthesis ? Um ,	QMSum_120
false	Yeah.	QMSum_120
false	O - OK. I don't know anything about re - synthesis. Uh , how likely do you think that is ?	QMSum_120
false	Uh , it depends what you what you do. I mean , it 's it 's it 's , uh , um Don't know. But anyway it sounds like a reasonable way to go for a for an initial thing , and we can look at at exactly what you end up doing and and then figure out if there 's some something that could be be hurt by the end part of the process.	QMSum_120
false	OK.	QMSum_120
false	OK. So that 's That was it , huh ?	QMSum_120
false	That Yeah , e That 's it , that 's it.	QMSum_120
false	OK. OK.	QMSum_120
false	Uh - huh.	QMSum_120
false	Um , anything to add ?	QMSum_120
false	Um. Well , I 've been continuing reading. I went off on a little tangent this past week , um , looking at , uh , uh , modulation s spectrum stuff , um , and and learning a bit about what what , um what it is , and , uh , the importance of it in speech recognition. And I found some some , uh , neat papers , um , historical papers from , um , Kanedera , Hermansky , and Arai.	QMSum_120

false

So , yeah.

QMSum_120

false

Mm - hmm.

QMSum_120

false

Oh. Oh , so this i is this the difference here , for that ?

QMSum_120

false

No. This is the signal. This is the signal.

QMSum_120

false

I see that. Oh , yeah.

QMSum_120

false

The frame.

QMSum_120

false

Oh , that 's the f the original.

QMSum_120

false

Yeah.

QMSum_120

false

This is the fra the original frame.

QMSum_120

false

So with a short frame basically you have only two periods

QMSum_120

false

Yeah.

QMSum_120

false

and it 's not not enough to to have this kind of neat things.

QMSum_120

false

Mm - hmm.

QMSum_120

false

Mm - hmm.

QMSum_120

false

Yeah.

QMSum_120

false

But

QMSum_120

false

And here No , well.

QMSum_120

false

Yeah. So probably we 'll have to use , like , long f long frames. Mm - hmm.

QMSum_120

false

Mm - hmm.

QMSum_120

false

Hmm.

QMSum_120

false

Oh.

QMSum_120

false

Mmm.

QMSum_120

false

That 's interesting.

QMSum_120

false

Yeah , maybe. Well , I mean it looks better , but , I mean , the thing is if if , uh if you 're actually asking you know , if you actually j uh , need to do place along an FFT , it may be it may be pushing things.

QMSum_120

false

Yeah.

QMSum_120

false

And and , uh

QMSum_120

false

Would you would you wanna do this kind of , uh , difference thing after you do spectral subtraction ?

QMSum_120

false

Uh , maybe.

QMSum_120

false

No. Maybe we can do that.

QMSum_120

false

Mmm.

QMSum_120

false

Hmm. The spectral subtraction is being done at what level ? Is it being done at the level of FFT bins or at the level of , uh , mel spectrum or something ?

QMSum_120

false

Um , I guess it depends.

QMSum_120

false

I mean , how are they doing it ?

QMSum_120

false

How they 're doing it ? Yeah. Um , I guess Ericsson is on the , um , filter bank ,

QMSum_120

false

FFT. Filter bank ,

QMSum_120

false

no ? It 's on the filter bank ,

QMSum_120

false

yeah.

QMSum_120

false

so. So , yeah , probably I i it Yeah.

QMSum_120

false

So in that case , it might not make much difference at all.

QMSum_120

false

Seems like you 'd wanna do it on the FFT bins.

QMSum_120

false

Maybe. I mean , certainly it 'd be better.

QMSum_120

false

I I mean , if you were gonna uh , for for this purpose , that is.

QMSum_120

false

Mm - hmm.

QMSum_120

false

Yeah.

QMSum_120

false

Mm - hmm.

QMSum_120

false

Yeah. OK.

QMSum_120

false

Mmm.

QMSum_120

false

What else ?

QMSum_120

false

Uh. Yeah , that 's all. So we 'll perhaps try to convince OGI people to use the new the new filters and Yeah.

QMSum_120

false

OK. Uh , has has anything happened yet on this business of having some sort of standard , uh , source ,

QMSum_120

false

Uh , not yet

QMSum_120

false

or ?

QMSum_120

false

but I wi I will call them and

QMSum_120

false

OK.

QMSum_120

false

now they are I think they have more time because they have this well , Eurospeech deadline is over

QMSum_120

true

When is the next , um , Aurora deadline ?

QMSum_120

false

and It 's , um , in June. Yeah.

QMSum_120

false

June.

QMSum_120

false

Early June , late June , middle June ?

QMSum_120

false

I don't know w

QMSum_120

false

Hmm.

QMSum_120

false

Hmm.

QMSum_120

false

OK. Um , and he 's been doing all the talking but but these he 's he 's , uh

QMSum_120

false

Yeah.

QMSum_120

false

This is this by the way a bad thing. We 're trying to get , um , m more female voices in this record as well. So. Make sur make sure Carmen talks as well. Uh , but has he pretty much been talking about what you 're doing also , and ?

QMSum_120

false

Oh , I I am doing this.

QMSum_120

false

Yes.

QMSum_120

false

Yeah , yeah. I don't know. I 'm sorry , but I think that for the recognizer for the meeting recorder that it 's better that I don't speak.

QMSum_120

false

Yeah , well.

QMSum_120

false

Because

QMSum_120

false

You know , uh , we 'll get we 'll get to , uh , Spanish voices sometime , and we do we want to recognize , uh , you too.

QMSum_120

false

After the after , uh , the result for the TI - digits on the meeting record there will be foreigns people.

QMSum_120

false

Yeah , but

QMSum_120

false

Oh , no.

QMSum_120

false

We like we we 're we 're w we are we 're in the , uh , Bourlard - Hermansky - Morgan , uh , frame of mind. Yeah , we like high error rates. It 's

QMSum_120

false

Yeah.

QMSum_120

false

That way there 's lots of work to do. So it 's Uh , anything to talk about ?

QMSum_120

false

N um , not not not much is new. So when I talked about what I 'm planning to do last time , I said I was , um , going to use Avendano 's method of , um , using a transformation , um , to map from long analysis frames which are used for removing reverberation to short analysis frames for feature calculation. He has a trick for doing that involving viewing the DFT as a matrix. Um , but , uh , um , I decided not to do that after all because I I realized to use it I 'd need to have these short analysis frames get plugged directly into the feature computation somehow

QMSum_120

false

Mm - hmm.

QMSum_120

false

and right now I think our feature computation is set to up to , um , take , um , audio as input , in general. So I decided that I I 'll do the reverberation removal on the long analysis windows and then just re - synthesize audio and then send that.

QMSum_120

false

This is in order to use the SRI system or something. Right ?

QMSum_120

false

Um , or or even if I 'm using our system , I was thinking it might be easier to just re - synthesize the audio ,

QMSum_120

false

Yeah ?

QMSum_120

false

because then I could just feacalc as is and I wouldn't have to change the code.

QMSum_120

false

Oh , OK. Yeah. I mean , it 's um , certainly in a short short - term this just sounds easier.

QMSum_120

false

Uh - huh.

QMSum_120

false

Yeah. I mean , longer - term if it 's if it turns out to be useful , one one might want to do something else ,

QMSum_120

false

Right. That 's true.

QMSum_120

false

but Uh , uh , I mean , in in other words , you you may be putting other kinds of errors in from the re - synthesis process.

QMSum_120

false

But e u From the re - synthesis ? Um ,

QMSum_120

false

Yeah.

QMSum_120

false

O - OK. I don't know anything about re - synthesis. Uh , how likely do you think that is ?

QMSum_120

false

Uh , it depends what you what you do. I mean , it 's it 's it 's , uh , um Don't know. But anyway it sounds like a reasonable way to go for a for an initial thing , and we can look at at exactly what you end up doing and and then figure out if there 's some something that could be be hurt by the end part of the process.

QMSum_120

false

OK.

QMSum_120

false

OK. So that 's That was it , huh ?

QMSum_120

false

That Yeah , e That 's it , that 's it.

QMSum_120

false

OK. OK.

QMSum_120

false

Uh - huh.

QMSum_120

false

Um , anything to add ?

QMSum_120

false

Um. Well , I 've been continuing reading. I went off on a little tangent this past week , um , looking at , uh , uh , modulation s spectrum stuff , um , and and learning a bit about what what , um what it is , and , uh , the importance of it in speech recognition. And I found some some , uh , neat papers , um , historical papers from , um , Kanedera , Hermansky , and Arai.

QMSum_120