Datasets:
elplaguister
/
DTS_line_datasets

Dataset card Data Studio Files
xet
Community
topic_shift
bool
2 classes
utterance
stringlengths
1
7.9k
session_id
stringlengths
7
14
false
Mm - hmm.
QMSum_78
false
I think with
QMSum_78
false
the to speech - nonspeech as
QMSum_78
false
That 's right. But I think Javier 's
QMSum_78
false
it 's only speech or it 's it 's it 's nonspeech.
QMSum_78
false
Ah. Yeah.
QMSum_78
false
Mm - hmm.
QMSum_78
false
I think Javier 's might be able to.
QMSum_78
false
So.
QMSum_78
false
N n
QMSum_78
false
It doesn't have the same Gaus - uh , H M M modeling ,
QMSum_78
false
Yeah.
QMSum_78
false
which is I think a drawback.
QMSum_78
false
OK.
QMSum_78
false
But , uh
QMSum_78
false
Well , it 's sort of has a simple one.
QMSum_78
false
Mmm , yeah.
QMSum_78
false
Does it ?
QMSum_78
false
Right ? It 's it 's just it 's just a isn't it just a Gaussian
QMSum_78
false
Yeah.
QMSum_78
false
for each ?
QMSum_78
false
Yeah. And then he ch you choose optimal splitting.
QMSum_78
false
Hmm.
QMSum_78
false
Mm - hmm.
QMSum_78
false
Yeah. Oh , it doesn't have it doesn't have any temporal , uh ?
QMSum_78
false
Maybe I 'm misremembering , but I did not think it had a Markov
QMSum_78
false
I thought it Yeah. I gues I guess I don't remember either. Uh. It 's been a while.
QMSum_78
false
Yeah. Uh , I could have a look at it.
QMSum_78
false
Javier
QMSum_78
false
Uh.
QMSum_78
false
So.
QMSum_78
false
You mean Ja - eh , eh , Javier program ?
QMSum_78
false
Mm - hmm.
QMSum_78
false
No , Javier di doesn't worked with , uh , a Markov
QMSum_78
false
Yeah , I didn't think so.
QMSum_78
false
He on only train
QMSum_78
false
Oh , OK. So he 's just he just computes a Gaussian over potential
QMSum_78
false
Yep.
QMSum_78
false
Yeah. It was only Gaussian.
QMSum_78
false
Oh , I see. I see.
QMSum_78
false
And so I I think it would work fine for detecting overlap.
QMSum_78
false
This is the idea.
QMSum_78
false
And and
QMSum_78
false
It 's just , uh , that i it he has the two - pass issue that What he does is , as a first pass he he p he does , um , a guess at where the divisions might be and he overestimates. And that 's just a data reduction step , so that you 're not trying at every time interval.
QMSum_78
false
OK.
QMSum_78
false
And so those are the putative places where he tries.
QMSum_78
false
Yeah.
QMSum_78
false
Yeah. OK.
QMSum_78
false
And right now he 's doing that with silence and that doesn't work with the Meeting Recorder. So if we used another method to get the first pass , I think it would probably work.
QMSum_78
false
Yeah. Yeah. Sure. Yeah. Yeah , OK.
QMSum_78
false
It 's a good method. As long as the len as long the segments are long enough.
QMSum_78
false
Yeah.
QMSum_78
false
That 's the other problem.
QMSum_78
false
So
QMSum_78
false
O - k OK. So let me go back to what you had , though.
QMSum_78
false
Yeah.
QMSum_78
false
Um.
QMSum_78
false
Mm - hmm.
QMSum_78
false
The other thing one could do is Couldn't I mean , it 's So you have two categories
QMSum_78
false
Yeah.
QMSum_78
false
and you have Markov models for each. Couldn't you have a third category ? So you have , uh you have , uh , nonspeech , single - person speech , and multiple - person speech ?
QMSum_78
false
He has this on his board actually. Don't you have , like those those several different categories on the board ?
QMSum_78
false
Right ? And then you have a Markov model for each ?
QMSum_78
false
Um I 'm not sure. I I thought about , uh , adding , uh , uh , another class too. But it 's not too easy , I think , the the transition between the different class , to model them in in the system I have now. But it it it could be possible , I think ,
QMSum_78
false
I see. I see.
QMSum_78
false
in principle.
QMSum_78
false
Yeah , I mean , I This is all pretty gross.
QMSum_78
false
Yeah.
QMSum_78
false
I mean , the th the reason why , uh , I was suggesting originally that we look at features is because I thought , well , we 're doing something we haven't done before ,
QMSum_78
false
Yeah.
QMSum_78
false
we should at least look at the space and understand
QMSum_78
false
Yeah.
QMSum_78
false
Yeah.
QMSum_78
false
It seems like if two people two or more people talk at once , it should get louder ,
QMSum_78
false
Yeah.
QMSum_78
false
uh , and , uh , uh , there should be some discontinuity in pitch contours ,
QMSum_78
false
I had the impression.
QMSum_78
false
Yeah.
QMSum_78
false
Yeah.
QMSum_78
false
and , uh , there should overall be a , um , smaller proportion of the total energy that is explained by any particular harmonic sequence in the spectrum.
QMSum_78
false
Right.
QMSum_78
false
Yeah.
QMSum_78
false
So those are all things that should be there.
QMSum_78
false
Yeah.
QMSum_78
false
Mm - hmm.
QMSum_78
false
So far , um , uh , Jose has has been By the way , I was told I should be calling you Pepe , but
QMSum_78
false
Yeah.
QMSum_78
false
by your friends , but Anyway ,
QMSum_78
false
Yeah.
QMSum_78
false
um , uh , the has has , uh , been exploring , uh , e largely the energy issue and , um , as with a lot of things , it is not uh , like this , it 's not as simple as it sounds.
QMSum_78
false
Yeah.
QMSum_78
false
And then there 's , you know Is it energy ? Is it log energy ? Is it LPC residual energy ? Is it is it is it , uh , delta of those things ? Uh , what is it no Obviously , just a simple number absolute number isn't gonna work. So it should be with compared to what ? Should there be a long window for the normalizing factor and a short window for what you 're looking at ?
QMSum_78
false
Yeah.
QMSum_78
false
Or , you know , how b short should they be ? So ,
QMSum_78
false
Hmm.
QMSum_78
false
th he 's been playing around with a lot of these different things and and so far at least has not come up with any combination that really gave you an indicator.
QMSum_78
false
Yeah.
QMSum_78
false
So I I still have a hunch that there 's it 's in there some place , but it may be given that you have a limited time here , it it just may not be the best thing to to to focus on for the remaining of it.
QMSum_78
false
Yeah. To overrule , yeah.
QMSum_78
false
So pitch - related and harmonic - related , I 'm I 'm somewhat more hopeful for it.
QMSum_78