elplaguister/DTS_pair_datasets · Datasets at Hugging Face

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
But I bu	It it seems that the VAD network doesn't Well , it doesn't drop , uh , too many frames because the dele the number of deletion is reasonable. But it 's just when we add the tandem , the final MLP , and then	Yeah. Now the only problem is you don't want to ta I guess wait for the output of the VAD before you can put something into the other system ,	cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?	false	QMSum_194
It it seems that the VAD network doesn't Well , it doesn't drop , uh , too many frames because the dele the number of deletion is reasonable. But it 's just when we add the tandem , the final MLP , and then	Yeah. Now the only problem is you don't want to ta I guess wait for the output of the VAD before you can put something into the other system ,	cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?	But	false	QMSum_194
Yeah. Now the only problem is you don't want to ta I guess wait for the output of the VAD before you can put something into the other system ,	cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?	But	Mm - hmm.	false	QMSum_194
cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?	But	Mm - hmm.	Yeah. Right.	false	QMSum_194
But	Mm - hmm.	Yeah. Right.	Yeah. So that 's maybe a problem with what I was just saying. But but I I guess	false	QMSum_194
Mm - hmm.	Yeah. Right.	Yeah. So that 's maybe a problem with what I was just saying. But but I I guess	But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?	false	QMSum_194
Yeah. Right.	Yeah. So that 's maybe a problem with what I was just saying. But but I I guess	But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?	Um , well. We w we don't have it , actually ,	false	QMSum_194
Yeah. So that 's maybe a problem with what I was just saying. But but I I guess	But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?	Um , well. We w we don't have it , actually ,	No.	false	QMSum_194
But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?	Um , well. We w we don't have it , actually ,	No.	because it 's it has a high rate energy	false	QMSum_194
Um , well. We w we don't have it , actually ,	No.	because it 's it has a high rate energy	Ah.	false	QMSum_194
No.	because it 's it has a high rate energy	Ah.	the VAD has a	false	QMSum_194
because it 's it has a high rate energy	Ah.	the VAD has a	Yeah.	false	QMSum_194
Ah.	the VAD has a	Yeah.	OK.	false	QMSum_194
the VAD has a	Yeah.	OK.	It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.	false	QMSum_194
Yeah.	OK.	It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.	Right.	false	QMSum_194
OK.	It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.	Right.	In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.	false	QMSum_194
It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.	Right.	In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.	Mm - hmm.	false	QMSum_194
Right.	In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.	Mm - hmm.	And then maybe it would just learn learn it better.	false	QMSum_194
In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.	Mm - hmm.	And then maybe it would just learn learn it better.	Mm - hmm.	false	QMSum_194
Mm - hmm.	And then maybe it would just learn learn it better.	Mm - hmm.	Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,	false	QMSum_194
And then maybe it would just learn learn it better.	Mm - hmm.	Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,	Mm - hmm.	false	QMSum_194
Mm - hmm.	Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,	Mm - hmm.	at some point where the VAD is saying it 's actually speech.	false	QMSum_194
Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,	Mm - hmm.	at some point where the VAD is saying it 's actually speech.	Yeah.	false	QMSum_194
Mm - hmm.	at some point where the VAD is saying it 's actually speech.	Yeah.	Which is probably true.	false	QMSum_194
at some point where the VAD is saying it 's actually speech.	Yeah.	Which is probably true.	So , m	false	QMSum_194
Yeah.	Which is probably true.	So , m	Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh	false	QMSum_194
Which is probably true.	So , m	Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh	Yeah.	false	QMSum_194
So , m	Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh	Yeah.	Hmm. Anyway. Might be.	false	QMSum_194
Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh	Yeah.	Hmm. Anyway. Might be.	Mm - hmm.	false	QMSum_194
Yeah.	Hmm. Anyway. Might be.	Mm - hmm.	Yeah. Well , we just started working with it. But these are these are some good ideas I think.	false	QMSum_194
Hmm. Anyway. Might be.	Mm - hmm.	Yeah. Well , we just started working with it. But these are these are some good ideas I think.	Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?	false	QMSum_194
Mm - hmm.	Yeah. Well , we just started working with it. But these are these are some good ideas I think.	Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?	For the tandem net you mean ?	false	QMSum_194
Yeah. Well , we just started working with it. But these are these are some good ideas I think.	Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?	For the tandem net you mean ?	Well , I 'm Yeah.	false	QMSum_194
Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?	For the tandem net you mean ?	Well , I 'm Yeah.	Hmm.	false	QMSum_194
For the tandem net you mean ?	Well , I 'm Yeah.	Hmm.	I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.	false	QMSum_194
Well , I 'm Yeah.	Hmm.	I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.	Problem is , if you are going to run this on different m test sets , including large vocabulary ,	false	QMSum_194
Hmm.	I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.	Problem is , if you are going to run this on different m test sets , including large vocabulary ,	Yeah. Yeah.	false	QMSum_194
I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.	Problem is , if you are going to run this on different m test sets , including large vocabulary ,	Yeah. Yeah.	um ,	false	QMSum_194
Problem is , if you are going to run this on different m test sets , including large vocabulary ,	Yeah. Yeah.	um ,	Uh	false	QMSum_194
Yeah. Yeah.	um ,	Uh	I think	false	QMSum_194
um ,	Uh	I think	Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.	false	QMSum_194
Uh	I think	Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.	Yeah.	false	QMSum_194
I think	Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.	Yeah.	So. But I don't know.	false	QMSum_194
Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.	Yeah.	So. But I don't know.	Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,	false	QMSum_194
Yeah.	So. But I don't know.	Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,	Mm - hmm.	false	QMSum_194
So. But I don't know.	Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,	Mm - hmm.	but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.	false	QMSum_194
Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,	Mm - hmm.	but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.	Mm - hmm.	false	QMSum_194
Mm - hmm.	but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.	Mm - hmm.	That and the correlation between stuff.	false	QMSum_194
but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.	Mm - hmm.	That and the correlation between stuff.	So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?	false	QMSum_194
Mm - hmm.	That and the correlation between stuff.	So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?	Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i	false	QMSum_194
That and the correlation between stuff.	So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?	Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i	y Like sixty ?	false	QMSum_194
So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?	Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i	y Like sixty ?	Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.	false	QMSum_194
Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i	y Like sixty ?	Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.	Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.	false	QMSum_194
y Like sixty ?	Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.	Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.	Uh - huh. Yeah. So this would be sixty - two ?	false	QMSum_194
Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.	Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.	Uh - huh. Yeah. So this would be sixty - two ?	Sixty - two.	false	QMSum_194
Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.	Uh - huh. Yeah. So this would be sixty - two ?	Sixty - two.	Which is	false	QMSum_194
Uh - huh. Yeah. So this would be sixty - two ?	Sixty - two.	Which is	Yeah.	false	QMSum_194
Sixty - two.	Which is	Yeah.	Sixty - two , yeah.	false	QMSum_194
Which is	Yeah.	Sixty - two , yeah.	Somewhere around sixty , must be. Right ? Yeah.	false	QMSum_194
Yeah.	Sixty - two , yeah.	Somewhere around sixty , must be. Right ? Yeah.	Well , it 's around five percent , because it 's s Right ? If everything is five percent.	false	QMSum_194
Sixty - two , yeah.	Somewhere around sixty , must be. Right ? Yeah.	Well , it 's around five percent , because it 's s Right ? If everything is five percent.	Yeah. Yeah.	false	QMSum_194
Somewhere around sixty , must be. Right ? Yeah.	Well , it 's around five percent , because it 's s Right ? If everything is five percent.	Yeah. Yeah.	All the other ones were five percent ,	false	QMSum_194
Well , it 's around five percent , because it 's s Right ? If everything is five percent.	Yeah. Yeah.	All the other ones were five percent ,	Mm - hmm.	false	QMSum_194
Yeah. Yeah.	All the other ones were five percent ,	Mm - hmm.	the	false	QMSum_194
All the other ones were five percent ,	Mm - hmm.	the	Yeah.	false	QMSum_194
Mm - hmm.	the	Yeah.	I d I d I just have the SpeechDat - Car right now , so	false	QMSum_194
the	Yeah.	I d I d I just have the SpeechDat - Car right now , so	Yeah.	false	QMSum_194
Yeah.	I d I d I just have the SpeechDat - Car right now , so	Yeah.	It 's running it shou we should have the results today during the afternoon ,	false	QMSum_194
I d I d I just have the SpeechDat - Car right now , so	Yeah.	It 's running it shou we should have the results today during the afternoon ,	Hmm.	false	QMSum_194
Yeah.	It 's running it shou we should have the results today during the afternoon ,	Hmm.	but Well.	false	QMSum_194
It 's running it shou we should have the results today during the afternoon ,	Hmm.	but Well.	Hmm. Well Um So I won't be here for	false	QMSum_194
Hmm.	but Well.	Hmm. Well Um So I won't be here for	When When do you leave ?	false	QMSum_194
but Well.	Hmm. Well Um So I won't be here for	When When do you leave ?	Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,	false	QMSum_194
Hmm. Well Um So I won't be here for	When When do you leave ?	Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,	But you 're	false	QMSum_194
When When do you leave ?	Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,	But you 're	so I	false	QMSum_194
Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,	But you 're	so I	are you you 're not gonna be around this afternoon ?	false	QMSum_194
But you 're	so I	are you you 're not gonna be around this afternoon ?	Yeah.	false	QMSum_194
so I	are you you 're not gonna be around this afternoon ?	Yeah.	Oh.	false	QMSum_194
are you you 're not gonna be around this afternoon ?	Yeah.	Oh.	Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.	false	QMSum_194
Yeah.	Oh.	Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.	Uh - huh.	false	QMSum_194
Oh.	Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.	Uh - huh.	This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.	false	QMSum_194
Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.	Uh - huh.	This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.	Ah , OK , OK.	false	QMSum_194
Uh - huh.	This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.	Ah , OK , OK.	Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,	false	QMSum_194
This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.	Ah , OK , OK.	Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,	What 's September sixth ?	false	QMSum_194
Ah , OK , OK.	Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,	What 's September sixth ?	and Uh , that 's during Eurospeech.	false	QMSum_194
Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,	What 's September sixth ?	and Uh , that 's during Eurospeech.	Oh , oh , right. OK.	false	QMSum_194
What 's September sixth ?	and Uh , that 's during Eurospeech.	Oh , oh , right. OK.	So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?	false	QMSum_194
and Uh , that 's during Eurospeech.	Oh , oh , right. OK.	So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?	So , uh , you 're gonna be gone for the next three weeks or something ?	false	QMSum_194
Oh , oh , right. OK.	So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?	So , uh , you 're gonna be gone for the next three weeks or something ?	I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.	false	QMSum_194
So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?	So , uh , you 're gonna be gone for the next three weeks or something ?	I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.	So that 's you won't be at the next three of these meetings. Is that right ?	false	QMSum_194
So , uh , you 're gonna be gone for the next three weeks or something ?	I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.	So that 's you won't be at the next three of these meetings. Is that right ?	Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.	false	QMSum_194
I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.	So that 's you won't be at the next three of these meetings. Is that right ?	Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.	Oh , right. Right.	false	QMSum_194
So that 's you won't be at the next three of these meetings. Is that right ?	Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.	Oh , right. Right.	Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,	false	QMSum_194
Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.	Oh , right. Right.	Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,	OK.	false	QMSum_194
Oh , right. Right.	Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,	OK.	but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.	false	QMSum_194
Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,	OK.	but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.	When do you go back ?	false	QMSum_194
OK.	but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.	When do you go back ?	So.	false	QMSum_194
but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.	When do you go back ?	So.	Thirty - first , August.	false	QMSum_194
When do you go back ?	So.	Thirty - first , August.	Yeah. Yeah. So. Cool.	false	QMSum_194
So.	Thirty - first , August.	Yeah. Yeah. So. Cool.	When is the evaluation ? November , or something ?	false	QMSum_194

But I bu

It it seems that the VAD network doesn't Well , it doesn't drop , uh , too many frames because the dele the number of deletion is reasonable. But it 's just when we add the tandem , the final MLP , and then

Yeah. Now the only problem is you don't want to ta I guess wait for the output of the VAD before you can put something into the other system ,

cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?

false

QMSum_194

It it seems that the VAD network doesn't Well , it doesn't drop , uh , too many frames because the dele the number of deletion is reasonable. But it 's just when we add the tandem , the final MLP , and then

Yeah. Now the only problem is you don't want to ta I guess wait for the output of the VAD before you can put something into the other system ,

cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?

But

false

QMSum_194

Yeah. Now the only problem is you don't want to ta I guess wait for the output of the VAD before you can put something into the other system ,

cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?

But

Mm - hmm.

false

QMSum_194

cuz that 'll shoot up the latency a lot , right ? Am I missing something here ?

But

Mm - hmm.

Yeah. Right.

false

QMSum_194

But

Mm - hmm.

Yeah. Right.

Yeah. So that 's maybe a problem with what I was just saying. But but I I guess

false

QMSum_194

Mm - hmm.

Yeah. Right.

Yeah. So that 's maybe a problem with what I was just saying. But but I I guess

But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?

false

QMSum_194

Yeah. Right.

Yeah. So that 's maybe a problem with what I was just saying. But but I I guess

But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?

Um , well. We w we don't have it , actually ,

false

QMSum_194

Yeah. So that 's maybe a problem with what I was just saying. But but I I guess

But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?

Um , well. We w we don't have it , actually ,

No.

false

QMSum_194

But if you were gonna put it in as a feature it means you already have it by the time you get to the tandem net , right ?

Um , well. We w we don't have it , actually ,

No.

because it 's it has a high rate energy

false

QMSum_194

Um , well. We w we don't have it , actually ,

No.

because it 's it has a high rate energy

Ah.

false

QMSum_194

No.

because it 's it has a high rate energy

Ah.

the VAD has a

false

QMSum_194

because it 's it has a high rate energy

Ah.

the VAD has a

Yeah.

false

QMSum_194

Ah.

the VAD has a

Yeah.

OK.

false

QMSum_194

the VAD has a

Yeah.

OK.

It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.

false

QMSum_194

Yeah.

OK.

It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.

Right.

false

QMSum_194

OK.

It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.

Right.

In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.

false

QMSum_194

It 's kind of done in I mean , some of the things are , not in parallel , but certainly , it would be in parallel with the with a tandem net.

Right.

In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.

Mm - hmm.

false

QMSum_194

Right.

In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.

Mm - hmm.

And then maybe it would just learn learn it better.

false

QMSum_194

In time. So maybe , if that doesn't work , um But it would be interesting to see if that was the problem , anyway. And and and then I guess another alternative would be to take the feature that you 're feeding into the VAD , and feeding it into the other one as well.

Mm - hmm.

And then maybe it would just learn learn it better.

Mm - hmm.

false

QMSum_194

Mm - hmm.

And then maybe it would just learn learn it better.

Mm - hmm.

Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,

false

QMSum_194

And then maybe it would just learn learn it better.

Mm - hmm.

Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,

Mm - hmm.

false

QMSum_194

Mm - hmm.

Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,

Mm - hmm.

at some point where the VAD is saying it 's actually speech.

false

QMSum_194

Um But that 's Yeah , that 's an interesting thing to try to see , if what 's going on is that in the highly mismatched condition , it 's , um , causing deletions by having this silence probability up up too high ,

Mm - hmm.

at some point where the VAD is saying it 's actually speech.

Yeah.

false

QMSum_194

Mm - hmm.

at some point where the VAD is saying it 's actually speech.

Yeah.

Which is probably true.

false

QMSum_194

at some point where the VAD is saying it 's actually speech.

Yeah.

Which is probably true.

So , m

false

QMSum_194

Yeah.

Which is probably true.

So , m

Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh

false

QMSum_194

Which is probably true.

So , m

Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh

Yeah.

false

QMSum_194

So , m

Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh

Yeah.

Hmm. Anyway. Might be.

false

QMSum_194

Cuz Well , the V A if the VAD said since the VAD is is is right a lot , uh

Yeah.

Hmm. Anyway. Might be.

Mm - hmm.

false

QMSum_194

Yeah.

Hmm. Anyway. Might be.

Mm - hmm.

Yeah. Well , we just started working with it. But these are these are some good ideas I think.

false

QMSum_194

Hmm. Anyway. Might be.

Mm - hmm.

Yeah. Well , we just started working with it. But these are these are some good ideas I think.

Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?

false

QMSum_194

Mm - hmm.

Yeah. Well , we just started working with it. But these are these are some good ideas I think.

Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?

For the tandem net you mean ?

false

QMSum_194

Yeah. Well , we just started working with it. But these are these are some good ideas I think.

Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?

For the tandem net you mean ?

Well , I 'm Yeah.

false

QMSum_194

Mm - hmm. Yeah , and the other thing Well , there are other issues maybe for the tandem , like , uh , well , do we want to , w uh n Do we want to work on the targets ? Or , like , instead of using phonemes , using more context dependent units ?

For the tandem net you mean ?

Well , I 'm Yeah.

Hmm.

false

QMSum_194

For the tandem net you mean ?

Well , I 'm Yeah.

Hmm.

I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.

false

QMSum_194

Well , I 'm Yeah.

Hmm.

I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.

Problem is , if you are going to run this on different m test sets , including large vocabulary ,

false

QMSum_194

Hmm.

I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.

Problem is , if you are going to run this on different m test sets , including large vocabulary ,

Yeah. Yeah.

false

QMSum_194

I 'm thinking , also , a w about Dan 's work where he he trained a network , not on phoneme targets but on the HMM state targets. And it was giving s slightly better results.

Problem is , if you are going to run this on different m test sets , including large vocabulary ,

Yeah. Yeah.

um ,

false

QMSum_194

Problem is , if you are going to run this on different m test sets , including large vocabulary ,

Yeah. Yeah.

um ,

Uh

false

QMSum_194

Yeah. Yeah.

um ,

Uh

I think

false

QMSum_194

um ,

Uh

I think

Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.

false

QMSum_194

Uh

I think

Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.

Yeah.

false

QMSum_194

I think

Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.

Yeah.

So. But I don't know.

false

QMSum_194

Mmm. I was just thinking maybe about , like , generalized diphones , and come up with a a reasonable , not too large , set of context dependent units , and and Yeah. And then anyway we would have to reduce this with the KLT.

Yeah.

So. But I don't know.

Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,

false

QMSum_194

Yeah.

So. But I don't know.

Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,

Mm - hmm.

false

QMSum_194

So. But I don't know.

Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,

Mm - hmm.

but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.

false

QMSum_194

Yeah. Well , maybe. But I d I d it it i it 's all worth looking at ,

Mm - hmm.

but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.

Mm - hmm.

false

QMSum_194

Mm - hmm.

but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.

Mm - hmm.

That and the correlation between stuff.

false

QMSum_194

but it sounds to me like , uh , looking at the relationship between this and the speech noise stuff is is is probably a key thing.

Mm - hmm.

That and the correlation between stuff.

So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?

false

QMSum_194

Mm - hmm.

That and the correlation between stuff.

So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?

Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i

false

QMSum_194

That and the correlation between stuff.

So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?

Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i

y Like sixty ?

false

QMSum_194

So if , uh if the , uh , high mismatch case had been more like the , uh , the other two cases in terms of giving you just a better performance , how would this number have changed ?

Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i

y Like sixty ?

Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.

false

QMSum_194

Mm - hmm. Oh , it would be Yeah. Around five percent better , I guess. If if i

y Like sixty ?

Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.

Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.

false

QMSum_194

y Like sixty ?

Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.

Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.

Uh - huh. Yeah. So this would be sixty - two ?

false

QMSum_194

Well , we don't know what 's it 's gonna be the TI - digits yet. He hasn't got the results back yet.

Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.

Uh - huh. Yeah. So this would be sixty - two ?

Sixty - two.

false

QMSum_194

Yeah. If you extrapolate the SpeechDat - Car well - matched and medium - mismatch , it 's around , yeah , maybe five.

Uh - huh. Yeah. So this would be sixty - two ?

Sixty - two.

Which is

false

QMSum_194

Uh - huh. Yeah. So this would be sixty - two ?

Sixty - two.

Which is

Yeah.

false

QMSum_194

Sixty - two.

Which is

Yeah.

Sixty - two , yeah.

false

QMSum_194

Which is

Yeah.

Sixty - two , yeah.

Somewhere around sixty , must be. Right ? Yeah.

false

QMSum_194

Yeah.

Sixty - two , yeah.

Somewhere around sixty , must be. Right ? Yeah.

Well , it 's around five percent , because it 's s Right ? If everything is five percent.

false

QMSum_194

Sixty - two , yeah.

Somewhere around sixty , must be. Right ? Yeah.

Well , it 's around five percent , because it 's s Right ? If everything is five percent.

Yeah. Yeah.

false

QMSum_194

Somewhere around sixty , must be. Right ? Yeah.

Well , it 's around five percent , because it 's s Right ? If everything is five percent.

Yeah. Yeah.

All the other ones were five percent ,

false

QMSum_194

Well , it 's around five percent , because it 's s Right ? If everything is five percent.

Yeah. Yeah.

All the other ones were five percent ,

Mm - hmm.

false

QMSum_194

Yeah. Yeah.

All the other ones were five percent ,

Mm - hmm.

the

false

QMSum_194

All the other ones were five percent ,

Mm - hmm.

the

Yeah.

false

QMSum_194

Mm - hmm.

the

Yeah.

I d I d I just have the SpeechDat - Car right now , so

false

QMSum_194

the

Yeah.

I d I d I just have the SpeechDat - Car right now , so

Yeah.

false

QMSum_194

Yeah.

I d I d I just have the SpeechDat - Car right now , so

Yeah.

It 's running it shou we should have the results today during the afternoon ,

false

QMSum_194

I d I d I just have the SpeechDat - Car right now , so

Yeah.

It 's running it shou we should have the results today during the afternoon ,

Hmm.

false

QMSum_194

Yeah.

It 's running it shou we should have the results today during the afternoon ,

Hmm.

but Well.

false

QMSum_194

It 's running it shou we should have the results today during the afternoon ,

Hmm.

but Well.

Hmm. Well Um So I won't be here for

false

QMSum_194

Hmm.

but Well.

Hmm. Well Um So I won't be here for

When When do you leave ?

false

QMSum_194

but Well.

Hmm. Well Um So I won't be here for

When When do you leave ?

Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,

false

QMSum_194

Hmm. Well Um So I won't be here for

When When do you leave ?

Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,

But you 're

false

QMSum_194

When When do you leave ?

Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,

But you 're

so I

false

QMSum_194

Uh , I 'm leaving next Wednesday. May or may not be in in the morning. I leave in the afternoon. Um ,

But you 're

so I

are you you 're not gonna be around this afternoon ?

false

QMSum_194

But you 're

so I

are you you 're not gonna be around this afternoon ?

Yeah.

false

QMSum_194

so I

are you you 're not gonna be around this afternoon ?

Yeah.

Oh.

false

QMSum_194

are you you 're not gonna be around this afternoon ?

Yeah.

Oh.

Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.

false

QMSum_194

Yeah.

Oh.

Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.

Uh - huh.

false

QMSum_194

Oh.

Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.

Uh - huh.

This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.

false

QMSum_194

Oh , well. I 'm talking about next week. I 'm leaving leaving next Wednesday.

Uh - huh.

This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.

Ah , OK , OK.

false

QMSum_194

Uh - huh.

This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.

Ah , OK , OK.

Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,

false

QMSum_194

This afternoon uh Oh , right , for the Meeting meeting ? Yeah , that 's just cuz of something on campus.

Ah , OK , OK.

Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,

What 's September sixth ?

false

QMSum_194

Ah , OK , OK.

Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,

What 's September sixth ?

and Uh , that 's during Eurospeech.

false

QMSum_194

Yeah. But , um , yeah , so next week I won't , and the week after I won't , cuz I 'll be in Finland. And the week after that I won't. By that time you 'll be Uh , you 'll both be gone from here. So there 'll be no definitely no meeting on on September sixth. Uh ,

What 's September sixth ?

and Uh , that 's during Eurospeech.

Oh , oh , right. OK.

false

QMSum_194

What 's September sixth ?

and Uh , that 's during Eurospeech.

Oh , oh , right. OK.

So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?

false

QMSum_194

and Uh , that 's during Eurospeech.

Oh , oh , right. OK.

So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?

So , uh , you 're gonna be gone for the next three weeks or something ?

false

QMSum_194

Oh , oh , right. OK.

So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?

So , uh , you 're gonna be gone for the next three weeks or something ?

I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.

false

QMSum_194

So , uh , Sunil will be in Oregon. Uh , Stephane and I will be in Denmark. Uh Right ? So it 'll be a few weeks , really , before we have a meeting of the same cast of characters. Um , but , uh I guess , just I mean , you guys should probably meet. And maybe Barry Barry will be around. And and then uh , uh , we 'll start up again with Dave and Dave and Barry and Stephane and us on the , uh , twentieth. No. Thirteenth ? About a month ?

So , uh , you 're gonna be gone for the next three weeks or something ?

I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.

So that 's you won't be at the next three of these meetings. Is that right ?

false

QMSum_194

So , uh , you 're gonna be gone for the next three weeks or something ?

I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.

So that 's you won't be at the next three of these meetings. Is that right ?

Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.

false

QMSum_194

I 'm gone for two and a half weeks starting starting next Wed - late next Wednesday.

So that 's you won't be at the next three of these meetings. Is that right ?

Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.

Oh , right. Right.

false

QMSum_194

So that 's you won't be at the next three of these meetings. Is that right ?

Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.

Oh , right. Right.

Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,

false

QMSum_194

Uh , I won't it 's probably four because of is it three ? Let 's see , twenty - third , thirtieth , sixth. That 's right , next three. And the the third one won't probably won't be a meeting , cuz cuz , uh , Su - Sunil , Stephane , and I will all not be here.

Oh , right. Right.

Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,

OK.

false

QMSum_194

Oh , right. Right.

Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,

OK.

but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.

false

QMSum_194

Um Mmm. So it 's just , uh , the next two where there will be there , you know , may as well be meetings ,

OK.

but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.

When do you go back ?

false

QMSum_194

OK.

but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.

When do you go back ?

So.

false

QMSum_194

but I just won't be at them. And then starting up on the thirteenth , uh , we 'll have meetings again but we 'll have to do without Sunil here somehow.

When do you go back ?

So.

Thirty - first , August.

false

QMSum_194

When do you go back ?

So.

Thirty - first , August.

Yeah. Yeah. So. Cool.

false

QMSum_194

So.

Thirty - first , August.

Yeah. Yeah. So. Cool.

When is the evaluation ? November , or something ?

false

QMSum_194