prev1
stringlengths 1
7.9k
| prev2
stringlengths 1
7.9k
⌀ | next1
stringlengths 1
7.9k
| next2
stringlengths 1
7.9k
⌀ | is_boundary
bool 2
classes | session_id
stringlengths 7
14
|
|---|---|---|---|---|---|
Wow.
|
You know. So it 's We 're we 're doing better.
|
This is this is really good.
|
I mean , we 're getting better recognition. I mean , I 'm sure other people working on this are not sitting still either , but
| false |
QMSum_194
|
You know. So it 's We 're we 're doing better.
|
This is this is really good.
|
I mean , we 're getting better recognition. I mean , I 'm sure other people working on this are not sitting still either , but
|
Yeah.
| false |
QMSum_194
|
This is this is really good.
|
I mean , we 're getting better recognition. I mean , I 'm sure other people working on this are not sitting still either , but
|
Yeah.
|
but but , uh Uh , I mean , the important thing is that we learn how to do this better , and , you know. So. Um , Yeah. So , our , um Yeah , you can see the kind of kind of numbers that we 're having , say , on SpeechDat - Car which is a hard task , cuz it 's really , um I think it 's just sort of sort of reasonable numbers , starting to be. I mean , it 's still terri
| false |
QMSum_194
|
I mean , we 're getting better recognition. I mean , I 'm sure other people working on this are not sitting still either , but
|
Yeah.
|
but but , uh Uh , I mean , the important thing is that we learn how to do this better , and , you know. So. Um , Yeah. So , our , um Yeah , you can see the kind of kind of numbers that we 're having , say , on SpeechDat - Car which is a hard task , cuz it 's really , um I think it 's just sort of sort of reasonable numbers , starting to be. I mean , it 's still terri
|
Mm - hmm. Yeah , even for a well - matched case it 's sixty percent error rate reduction ,
| false |
QMSum_194
|
Yeah.
|
but but , uh Uh , I mean , the important thing is that we learn how to do this better , and , you know. So. Um , Yeah. So , our , um Yeah , you can see the kind of kind of numbers that we 're having , say , on SpeechDat - Car which is a hard task , cuz it 's really , um I think it 's just sort of sort of reasonable numbers , starting to be. I mean , it 's still terri
|
Mm - hmm. Yeah , even for a well - matched case it 's sixty percent error rate reduction ,
|
Yeah.
| false |
QMSum_194
|
but but , uh Uh , I mean , the important thing is that we learn how to do this better , and , you know. So. Um , Yeah. So , our , um Yeah , you can see the kind of kind of numbers that we 're having , say , on SpeechDat - Car which is a hard task , cuz it 's really , um I think it 's just sort of sort of reasonable numbers , starting to be. I mean , it 's still terri
|
Mm - hmm. Yeah , even for a well - matched case it 's sixty percent error rate reduction ,
|
Yeah.
|
which is
| false |
QMSum_194
|
Mm - hmm. Yeah , even for a well - matched case it 's sixty percent error rate reduction ,
|
Yeah.
|
which is
|
Yeah. Probably half. Good !
| false |
QMSum_194
|
Yeah.
|
which is
|
Yeah. Probably half. Good !
|
Um , Yeah. So actually , this is in between what we had with the previous VAD and what Sunil did with an IDL VAD. Which gave sixty - two percent improvement , right ?
| false |
QMSum_194
|
which is
|
Yeah. Probably half. Good !
|
Um , Yeah. So actually , this is in between what we had with the previous VAD and what Sunil did with an IDL VAD. Which gave sixty - two percent improvement , right ?
|
Yeah , it 's almost that.
| false |
QMSum_194
|
Yeah. Probably half. Good !
|
Um , Yeah. So actually , this is in between what we had with the previous VAD and what Sunil did with an IDL VAD. Which gave sixty - two percent improvement , right ?
|
Yeah , it 's almost that.
|
So
| false |
QMSum_194
|
Um , Yeah. So actually , this is in between what we had with the previous VAD and what Sunil did with an IDL VAD. Which gave sixty - two percent improvement , right ?
|
Yeah , it 's almost that.
|
So
|
It 's almost an average somewhere around
| false |
QMSum_194
|
Yeah , it 's almost that.
|
So
|
It 's almost an average somewhere around
|
Yeah.
| false |
QMSum_194
|
So
|
It 's almost an average somewhere around
|
Yeah.
|
Yeah.
| false |
QMSum_194
|
It 's almost an average somewhere around
|
Yeah.
|
Yeah.
|
What was that ? Say that last part again ?
| false |
QMSum_194
|
Yeah.
|
Yeah.
|
What was that ? Say that last part again ?
|
So , if you use , like , an IDL VAD , uh , for dropping the frames ,
| false |
QMSum_194
|
Yeah.
|
What was that ? Say that last part again ?
|
So , if you use , like , an IDL VAD , uh , for dropping the frames ,
|
o o Or the best we can get.
| false |
QMSum_194
|
What was that ? Say that last part again ?
|
So , if you use , like , an IDL VAD , uh , for dropping the frames ,
|
o o Or the best we can get.
|
the best that we can get i That means that we estimate the silence probability on the clean version of the utterances. Then you can go up to sixty - two percent error rate reduction , globally.
| false |
QMSum_194
|
So , if you use , like , an IDL VAD , uh , for dropping the frames ,
|
o o Or the best we can get.
|
the best that we can get i That means that we estimate the silence probability on the clean version of the utterances. Then you can go up to sixty - two percent error rate reduction , globally.
|
Mmm.
| false |
QMSum_194
|
o o Or the best we can get.
|
the best that we can get i That means that we estimate the silence probability on the clean version of the utterances. Then you can go up to sixty - two percent error rate reduction , globally.
|
Mmm.
|
Mmm Yeah.
| false |
QMSum_194
|
the best that we can get i That means that we estimate the silence probability on the clean version of the utterances. Then you can go up to sixty - two percent error rate reduction , globally.
|
Mmm.
|
Mmm Yeah.
|
So that would be even That wouldn't change this number down here to sixty - two ?
| false |
QMSum_194
|
Mmm.
|
Mmm Yeah.
|
So that would be even That wouldn't change this number down here to sixty - two ?
|
Yeah.
| false |
QMSum_194
|
Mmm Yeah.
|
So that would be even That wouldn't change this number down here to sixty - two ?
|
Yeah.
|
Yeah. So you you were get
| false |
QMSum_194
|
So that would be even That wouldn't change this number down here to sixty - two ?
|
Yeah.
|
Yeah. So you you were get
|
If you add a g good v very good VAD , that works as well as a VAD working on clean speech ,
| false |
QMSum_194
|
Yeah.
|
Yeah. So you you were get
|
If you add a g good v very good VAD , that works as well as a VAD working on clean speech ,
|
Yeah. Yeah.
| false |
QMSum_194
|
Yeah. So you you were get
|
If you add a g good v very good VAD , that works as well as a VAD working on clean speech ,
|
Yeah. Yeah.
|
then you wou you would go
| false |
QMSum_194
|
If you add a g good v very good VAD , that works as well as a VAD working on clean speech ,
|
Yeah. Yeah.
|
then you wou you would go
|
So that 's sort of the best you could hope for.
| false |
QMSum_194
|
Yeah. Yeah.
|
then you wou you would go
|
So that 's sort of the best you could hope for.
|
Mm - hmm.
| false |
QMSum_194
|
then you wou you would go
|
So that 's sort of the best you could hope for.
|
Mm - hmm.
|
I see.
| false |
QMSum_194
|
So that 's sort of the best you could hope for.
|
Mm - hmm.
|
I see.
|
Probably. Yeah. So fi si fifty - three is what you were getting with the old VAD.
| false |
QMSum_194
|
Mm - hmm.
|
I see.
|
Probably. Yeah. So fi si fifty - three is what you were getting with the old VAD.
|
Yeah.
| false |
QMSum_194
|
I see.
|
Probably. Yeah. So fi si fifty - three is what you were getting with the old VAD.
|
Yeah.
|
And , uh and sixty - two with the the , you know , quote , unquote , cheating VAD. And fifty - seven is what you got with the real VAD.
| false |
QMSum_194
|
Probably. Yeah. So fi si fifty - three is what you were getting with the old VAD.
|
Yeah.
|
And , uh and sixty - two with the the , you know , quote , unquote , cheating VAD. And fifty - seven is what you got with the real VAD.
|
Mm - hmm.
| false |
QMSum_194
|
Yeah.
|
And , uh and sixty - two with the the , you know , quote , unquote , cheating VAD. And fifty - seven is what you got with the real VAD.
|
Mm - hmm.
|
That 's great.
| false |
QMSum_194
|
And , uh and sixty - two with the the , you know , quote , unquote , cheating VAD. And fifty - seven is what you got with the real VAD.
|
Mm - hmm.
|
That 's great.
|
Uh , yeah , the next thing is , I started to play Well , I don't want to worry too much about the delay , no. Maybe it 's better to wait
| false |
QMSum_194
|
Mm - hmm.
|
That 's great.
|
Uh , yeah , the next thing is , I started to play Well , I don't want to worry too much about the delay , no. Maybe it 's better to wait
|
OK.
| false |
QMSum_194
|
That 's great.
|
Uh , yeah , the next thing is , I started to play Well , I don't want to worry too much about the delay , no. Maybe it 's better to wait
|
OK.
|
for the decision
| false |
QMSum_194
|
Uh , yeah , the next thing is , I started to play Well , I don't want to worry too much about the delay , no. Maybe it 's better to wait
|
OK.
|
for the decision
|
Yeah.
| false |
QMSum_194
|
OK.
|
for the decision
|
Yeah.
|
from the committee. Uh , but I started to play with the , um , uh , tandem neural network. Mmm I just did the configuration that 's very similar to what we did for the February proposal. And Um. So. There is a f a first feature stream that use uh straight MFCC features.
| false |
QMSum_194
|
for the decision
|
Yeah.
|
from the committee. Uh , but I started to play with the , um , uh , tandem neural network. Mmm I just did the configuration that 's very similar to what we did for the February proposal. And Um. So. There is a f a first feature stream that use uh straight MFCC features.
|
Mm - hmm.
| false |
QMSum_194
|
Yeah.
|
from the committee. Uh , but I started to play with the , um , uh , tandem neural network. Mmm I just did the configuration that 's very similar to what we did for the February proposal. And Um. So. There is a f a first feature stream that use uh straight MFCC features.
|
Mm - hmm.
|
Well , these features actually. And the other stream is the output of a neural network , using as input , also , these , um , cleaned MFCC. Um
| false |
QMSum_194
|
from the committee. Uh , but I started to play with the , um , uh , tandem neural network. Mmm I just did the configuration that 's very similar to what we did for the February proposal. And Um. So. There is a f a first feature stream that use uh straight MFCC features.
|
Mm - hmm.
|
Well , these features actually. And the other stream is the output of a neural network , using as input , also , these , um , cleaned MFCC. Um
|
Those are th those are th what is going into the tandem net ?
| false |
QMSum_194
|
Mm - hmm.
|
Well , these features actually. And the other stream is the output of a neural network , using as input , also , these , um , cleaned MFCC. Um
|
Those are th those are th what is going into the tandem net ?
|
I don't have the comp Mmm ?
| false |
QMSum_194
|
Well , these features actually. And the other stream is the output of a neural network , using as input , also , these , um , cleaned MFCC. Um
|
Those are th those are th what is going into the tandem net ?
|
I don't have the comp Mmm ?
|
Those two ?
| false |
QMSum_194
|
Those are th those are th what is going into the tandem net ?
|
I don't have the comp Mmm ?
|
Those two ?
|
So there is just this feature stream , the fifteen MFCC plus delta and double - delta.
| false |
QMSum_194
|
I don't have the comp Mmm ?
|
Those two ?
|
So there is just this feature stream , the fifteen MFCC plus delta and double - delta.
|
No.
| false |
QMSum_194
|
Those two ?
|
So there is just this feature stream , the fifteen MFCC plus delta and double - delta.
|
No.
|
Yeah ?
| false |
QMSum_194
|
So there is just this feature stream , the fifteen MFCC plus delta and double - delta.
|
No.
|
Yeah ?
|
Um , so it 's makes forty - five features that are used as input to the HTK. And then , there is there are more inputs that comes from the tandem MLP.
| false |
QMSum_194
|
No.
|
Yeah ?
|
Um , so it 's makes forty - five features that are used as input to the HTK. And then , there is there are more inputs that comes from the tandem MLP.
|
Oh , oh. OK. I see.
| false |
QMSum_194
|
Yeah ?
|
Um , so it 's makes forty - five features that are used as input to the HTK. And then , there is there are more inputs that comes from the tandem MLP.
|
Oh , oh. OK. I see.
|
Yeah , h he likes to use them both ,
| false |
QMSum_194
|
Um , so it 's makes forty - five features that are used as input to the HTK. And then , there is there are more inputs that comes from the tandem MLP.
|
Oh , oh. OK. I see.
|
Yeah , h he likes to use them both ,
|
Uh - huh.
| false |
QMSum_194
|
Oh , oh. OK. I see.
|
Yeah , h he likes to use them both ,
|
Uh - huh.
|
cuz then it has one part that 's discriminative ,
| false |
QMSum_194
|
Yeah , h he likes to use them both ,
|
Uh - huh.
|
cuz then it has one part that 's discriminative ,
|
Yeah. Um
| false |
QMSum_194
|
Uh - huh.
|
cuz then it has one part that 's discriminative ,
|
Yeah. Um
|
one part that 's not.
| false |
QMSum_194
|
cuz then it has one part that 's discriminative ,
|
Yeah. Um
|
one part that 's not.
|
Right. OK.
| false |
QMSum_194
|
Yeah. Um
|
one part that 's not.
|
Right. OK.
|
So , um , uh , yeah. Right now it seems that i I just tested on SpeechDat - Car while the experiment are running on your on TI - digits. Well , it improves on the well - matched and the mismatched conditions , but it get worse on the highly mismatched. Um ,
| false |
QMSum_194
|
one part that 's not.
|
Right. OK.
|
So , um , uh , yeah. Right now it seems that i I just tested on SpeechDat - Car while the experiment are running on your on TI - digits. Well , it improves on the well - matched and the mismatched conditions , but it get worse on the highly mismatched. Um ,
|
Compared to these numbers ?
| false |
QMSum_194
|
Right. OK.
|
So , um , uh , yeah. Right now it seems that i I just tested on SpeechDat - Car while the experiment are running on your on TI - digits. Well , it improves on the well - matched and the mismatched conditions , but it get worse on the highly mismatched. Um ,
|
Compared to these numbers ?
|
Compared to these numbers , yeah. Um ,
| false |
QMSum_194
|
So , um , uh , yeah. Right now it seems that i I just tested on SpeechDat - Car while the experiment are running on your on TI - digits. Well , it improves on the well - matched and the mismatched conditions , but it get worse on the highly mismatched. Um ,
|
Compared to these numbers ?
|
Compared to these numbers , yeah. Um ,
|
like , on the well - match and medium mismatch , the gain is around five percent relative , but it goes down a lot more , like fifteen percent on the HM case.
| false |
QMSum_194
|
Compared to these numbers ?
|
Compared to these numbers , yeah. Um ,
|
like , on the well - match and medium mismatch , the gain is around five percent relative , but it goes down a lot more , like fifteen percent on the HM case.
|
You 're just using the full ninety features ?
| false |
QMSum_194
|
Compared to these numbers , yeah. Um ,
|
like , on the well - match and medium mismatch , the gain is around five percent relative , but it goes down a lot more , like fifteen percent on the HM case.
|
You 're just using the full ninety features ?
|
The
| false |
QMSum_194
|
like , on the well - match and medium mismatch , the gain is around five percent relative , but it goes down a lot more , like fifteen percent on the HM case.
|
You 're just using the full ninety features ?
|
The
|
Y you have ninety features ?
| false |
QMSum_194
|
You 're just using the full ninety features ?
|
The
|
Y you have ninety features ?
|
i I have , um From the networks , it 's twenty - eight. So
| false |
QMSum_194
|
The
|
Y you have ninety features ?
|
i I have , um From the networks , it 's twenty - eight. So
|
And from the other side it 's forty - five.
| false |
QMSum_194
|
Y you have ninety features ?
|
i I have , um From the networks , it 's twenty - eight. So
|
And from the other side it 's forty - five.
|
So , d i It 's forty - five.
| false |
QMSum_194
|
i I have , um From the networks , it 's twenty - eight. So
|
And from the other side it 's forty - five.
|
So , d i It 's forty - five.
|
So it 's you have seventy - three features ,
| false |
QMSum_194
|
And from the other side it 's forty - five.
|
So , d i It 's forty - five.
|
So it 's you have seventy - three features ,
|
Yeah.
| false |
QMSum_194
|
So , d i It 's forty - five.
|
So it 's you have seventy - three features ,
|
Yeah.
|
and you 're just feeding them like that.
| false |
QMSum_194
|
So it 's you have seventy - three features ,
|
Yeah.
|
and you 're just feeding them like that.
|
Yeah.
| false |
QMSum_194
|
Yeah.
|
and you 're just feeding them like that.
|
Yeah.
|
There isn't any KLT or anything ?
| false |
QMSum_194
|
and you 're just feeding them like that.
|
Yeah.
|
There isn't any KLT or anything ?
|
Mm - hmm. There 's a KLT after the neural network , as as before.
| false |
QMSum_194
|
Yeah.
|
There isn't any KLT or anything ?
|
Mm - hmm. There 's a KLT after the neural network , as as before.
|
That 's how you get down to twenty - eight ?
| false |
QMSum_194
|
There isn't any KLT or anything ?
|
Mm - hmm. There 's a KLT after the neural network , as as before.
|
That 's how you get down to twenty - eight ?
|
Yeah.
| false |
QMSum_194
|
Mm - hmm. There 's a KLT after the neural network , as as before.
|
That 's how you get down to twenty - eight ?
|
Yeah.
|
Why twenty - eight ?
| false |
QMSum_194
|
That 's how you get down to twenty - eight ?
|
Yeah.
|
Why twenty - eight ?
|
I don't know.
| false |
QMSum_194
|
Yeah.
|
Why twenty - eight ?
|
I don't know.
|
Oh.
| false |
QMSum_194
|
Why twenty - eight ?
|
I don't know.
|
Oh.
|
Uh. It 's i i i It 's because it 's what we did for the first proposal. We tested , uh , trying to go down
| false |
QMSum_194
|
I don't know.
|
Oh.
|
Uh. It 's i i i It 's because it 's what we did for the first proposal. We tested , uh , trying to go down
|
Ah.
| false |
QMSum_194
|
Oh.
|
Uh. It 's i i i It 's because it 's what we did for the first proposal. We tested , uh , trying to go down
|
Ah.
|
It 's a multiple of seven.
| false |
QMSum_194
|
Uh. It 's i i i It 's because it 's what we did for the first proposal. We tested , uh , trying to go down
|
Ah.
|
It 's a multiple of seven.
|
and Yeah.
| false |
QMSum_194
|
Ah.
|
It 's a multiple of seven.
|
and Yeah.
|
Yeah.
| false |
QMSum_194
|
It 's a multiple of seven.
|
and Yeah.
|
Yeah.
|
So Um.
| false |
QMSum_194
|
and Yeah.
|
Yeah.
|
So Um.
|
Yeah.
| false |
QMSum_194
|
Yeah.
|
So Um.
|
Yeah.
|
I wanted to do something very similar to the proposal as a first first try.
| false |
QMSum_194
|
So Um.
|
Yeah.
|
I wanted to do something very similar to the proposal as a first first try.
|
Yeah.
| false |
QMSum_194
|
Yeah.
|
I wanted to do something very similar to the proposal as a first first try.
|
Yeah.
|
I see.
| false |
QMSum_194
|
I wanted to do something very similar to the proposal as a first first try.
|
Yeah.
|
I see.
|
Yeah.
| false |
QMSum_194
|
Yeah.
|
I see.
|
Yeah.
|
Yeah. That makes sense.
| false |
QMSum_194
|
I see.
|
Yeah.
|
Yeah. That makes sense.
|
But we have to for sure , we have to go down , because the limit is now sixty features.
| false |
QMSum_194
|
Yeah.
|
Yeah. That makes sense.
|
But we have to for sure , we have to go down , because the limit is now sixty features.
|
Yeah.
| false |
QMSum_194
|
Yeah. That makes sense.
|
But we have to for sure , we have to go down , because the limit is now sixty features.
|
Yeah.
|
So , uh , we have to find a way to decrease the number of features. Um
| false |
QMSum_194
|
But we have to for sure , we have to go down , because the limit is now sixty features.
|
Yeah.
|
So , uh , we have to find a way to decrease the number of features. Um
|
So , it seems funny that I don't know , maybe I don't u quite understand everything , but that adding features I guess I guess if you 're keeping the back - end fixed. Maybe that 's it. Because it seems like just adding information shouldn't give worse results. But I guess if you 're keeping the number of Gaussians fixed in the recognizer , then
| false |
QMSum_194
|
Yeah.
|
So , uh , we have to find a way to decrease the number of features. Um
|
So , it seems funny that I don't know , maybe I don't u quite understand everything , but that adding features I guess I guess if you 're keeping the back - end fixed. Maybe that 's it. Because it seems like just adding information shouldn't give worse results. But I guess if you 're keeping the number of Gaussians fixed in the recognizer , then
|
Well , yeah.
| false |
QMSum_194
|
So , uh , we have to find a way to decrease the number of features. Um
|
So , it seems funny that I don't know , maybe I don't u quite understand everything , but that adding features I guess I guess if you 're keeping the back - end fixed. Maybe that 's it. Because it seems like just adding information shouldn't give worse results. But I guess if you 're keeping the number of Gaussians fixed in the recognizer , then
|
Well , yeah.
|
Mmm.
| false |
QMSum_194
|
So , it seems funny that I don't know , maybe I don't u quite understand everything , but that adding features I guess I guess if you 're keeping the back - end fixed. Maybe that 's it. Because it seems like just adding information shouldn't give worse results. But I guess if you 're keeping the number of Gaussians fixed in the recognizer , then
|
Well , yeah.
|
Mmm.
|
But , I mean , just in general , adding information Suppose the information you added , well , was a really terrible feature and all it brought in was noise.
| false |
QMSum_194
|
Well , yeah.
|
Mmm.
|
But , I mean , just in general , adding information Suppose the information you added , well , was a really terrible feature and all it brought in was noise.
|
Yeah.
| false |
QMSum_194
|
Mmm.
|
But , I mean , just in general , adding information Suppose the information you added , well , was a really terrible feature and all it brought in was noise.
|
Yeah.
|
Right ? So so , um Or or suppose it wasn't completely terrible , but it was completely equivalent to another one feature that you had , except it was noisier.
| false |
QMSum_194
|
But , I mean , just in general , adding information Suppose the information you added , well , was a really terrible feature and all it brought in was noise.
|
Yeah.
|
Right ? So so , um Or or suppose it wasn't completely terrible , but it was completely equivalent to another one feature that you had , except it was noisier.
|
Uh - huh.
| false |
QMSum_194
|
Yeah.
|
Right ? So so , um Or or suppose it wasn't completely terrible , but it was completely equivalent to another one feature that you had , except it was noisier.
|
Uh - huh.
|
Right ? In that case you wouldn't necessarily expect it to be better at all.
| false |
QMSum_194
|
Right ? So so , um Or or suppose it wasn't completely terrible , but it was completely equivalent to another one feature that you had , except it was noisier.
|
Uh - huh.
|
Right ? In that case you wouldn't necessarily expect it to be better at all.
|
Oh , yeah , I wasn't necessarily saying it should be better. I 'm just surprised that you 're getting fifteen percent relative worse on the wel
| false |
QMSum_194
|
Uh - huh.
|
Right ? In that case you wouldn't necessarily expect it to be better at all.
|
Oh , yeah , I wasn't necessarily saying it should be better. I 'm just surprised that you 're getting fifteen percent relative worse on the wel
|
Uh - huh.
| false |
QMSum_194
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.