Datasets:
key large_stringlengths 44 49 | group large_stringclasses 137
values | parts large_stringlengths 116 3.85k | tail_s float64 0 0.5 | duration_s float64 0.39 29.9 | target large_stringlengths 10 904 | n_speakers int64 1 4 | n_parts int64 1 31 |
|---|---|---|---|---|---|---|---|
EN2001a:AMI_EN2001a_H04_MEO069_0000334_0000388 | EN2001a | [{"row": 299, "id": "AMI_EN2001a_H04_MEO069_0000334_0000388", "speaker": "MEO069", "offset_s": 0.21, "dur_s": 0.54}, {"row": 1591, "id": "AMI_EN2001a_H00_MEE068_0000557_0000594", "speaker": "MEE068", "offset_s": 2.25, "dur_s": 0.37}, {"row": 47, "id": "AMI_EN2001a_H04_MEO069_0000560_0000601", "speaker": "MEO069", "offs... | 0.379 | 21.529 | <SPK_1>Okay.<SPK_2>Okay.<SPK_1>Gosh. Okay.<SPK_2>Does anyone want to see Steve's feedback from the specification?<SPK_1>Is there much more in it than he?<SPK_3>I drive Reddit last time, so.<SPK_2>Right.<SPK_1>Is there much more in it than he said yesterday?<SPK_2>Not really, um, just what he was talking about like dupl... | 3 | 13 |
EN2001a:AMI_EN2001a_H00_MEE068_0002809_0003300 | EN2001a | [{"row": 1002, "id": "AMI_EN2001a_H00_MEE068_0002809_0003300", "speaker": "MEE068", "offset_s": 0.202, "dur_s": 4.91}, {"row": 1262, "id": "AMI_EN2001a_H00_MEE068_0003300_0003961", "speaker": "MEE068", "offset_s": 5.112, "dur_s": 6.61}, {"row": 666, "id": "AMI_EN2001a_H03_MEE067_0003878_0003940", "speaker": "MEE067", "... | 0.256 | 12.408 | <SPK_1>Like duplication of effort and stuff and yeah. Saying that we should maybe think about having a prototype for week six, which is next week.<SPK_2>Next week.<SPK_1>Yeah. | 2 | 4 |
EN2001a:AMI_EN2001a_H00_MEE068_0004224_0004746 | EN2001a | [{"row": 766, "id": "AMI_EN2001a_H00_MEE068_0004224_0004746", "speaker": "MEE068", "offset_s": 0.238, "dur_s": 5.22}, {"row": 1282, "id": "AMI_EN2001a_H04_MEO069_0004683_0005624", "speaker": "MEO069", "offset_s": 4.828, "dur_s": 9.41}, {"row": 1197, "id": "AMI_EN2001a_H04_MEO069_0005624_0005970", "speaker": "MEO069", "... | 0.152 | 22.56 | <SPK_1>So, we should probably prioritize our packages.<SPK_2>Yeah, no, I'd say if for the prototype, if you just like wherever possible, chunking the stuff that we have, um, pre annotated and stuff. And for the stuff that we don't have pre-annotated, write like a stupid baseline.<SPK_1>Hmm.<SPK_2>Then we should probabl... | 2 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0006441_0007098 | EN2001a | [{"row": 123, "id": "AMI_EN2001a_H04_MEO069_0006441_0007098", "speaker": "MEO069", "offset_s": 0.252, "dur_s": 6.57}, {"row": 948, "id": "AMI_EN2001a_H04_MEO069_0007098_0007829", "speaker": "MEO069", "offset_s": 6.822, "dur_s": 7.31}, {"row": 492, "id": "AMI_EN2001a_H04_MEO069_0007829_0007991", "speaker": "MEO069", "of... | 0.454 | 18.906 | <SPK_1>So that we we take the the ready made parts and just see how we get them work together in the interface the way we wanted, and then we have a working prototype. And then we can go back and replace pieces either by our own components or by more sophisticated components of our own. So it's probably feasible.<SPK_2... | 3 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0008261_0008508 | EN2001a | [{"row": 1446, "id": "AMI_EN2001a_H04_MEO069_0008261_0008508", "speaker": "MEO069", "offset_s": 0.309, "dur_s": 2.47}, {"row": 1528, "id": "AMI_EN2001a_H01_FEO066_0008438_0009432", "speaker": "FEO066", "offset_s": 2.079, "dur_s": 9.94}, {"row": 277, "id": "AMI_EN2001a_H04_MEO069_0009323_0009394", "speaker": "MEO069", "... | 0.378 | 12.397 | <SPK_1>So that, for me.<SPK_2>I mean, if we just want to have some data for the user face, it could even be random data.<SPK_1>Oh yeah, yeah.<SPK_3>Yeah.<SPK_1>No. | 3 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0009419_0010003 | EN2001a | [{"row": 1438, "id": "AMI_EN2001a_H04_MEO069_0009419_0010003", "speaker": "MEO069", "offset_s": 0.491, "dur_s": 5.84}, {"row": 745, "id": "AMI_EN2001a_H01_FEO066_0009432_0009471", "speaker": "FEO066", "offset_s": 0.621, "dur_s": 0.39}, {"row": 1291, "id": "AMI_EN2001a_H04_MEO069_0010003_0010426", "speaker": "MEO069", "... | 0.455 | 11.336 | <SPK_1>But also like the, the similarity thing, like my just my matrix itself, for my stuff.<SPK_2>嗯。<SPK_1>I I think I can do that fairly quickly because I have the algorithms.<SPK_2>Yeah, I'm. | 2 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0010426_0011043 | EN2001a | [{"row": 989, "id": "AMI_EN2001a_H04_MEO069_0010426_0011043", "speaker": "MEO069", "offset_s": 0.155, "dur_s": 6.17}, {"row": 562, "id": "AMI_EN2001a_H03_MEE067_0010977_0011022", "speaker": "MEE067", "offset_s": 5.665, "dur_s": 0.45}, {"row": 555, "id": "AMI_EN2001a_H04_MEO069_0011043_0011289", "speaker": "MEO069", "of... | 0.451 | 22.786 | <SPK_1>Yeah, I think today's meeting is really the one where we, where we sort of settle down the data structure. And as soon as we have that.<SPK_2>Yeah.<SPK_1>So probably like after today's meeting, we then actually need to. Well, go back first of all and look at Night XML to see how far that that which we want is co... | 2 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0012911_0013629 | EN2001a | [{"row": 1241, "id": "AMI_EN2001a_H04_MEO069_0012911_0013629", "speaker": "MEO069", "offset_s": 0.342, "dur_s": 7.18}, {"row": 319, "id": "AMI_EN2001a_H04_MEO069_0013629_0013961", "speaker": "MEO069", "offset_s": 7.522, "dur_s": 3.32}, {"row": 763, "id": "AMI_EN2001a_H04_MEO069_0013961_0014534", "speaker": "MEO069", "o... | 0.449 | 23.091 | <SPK_1>So, I think if today we decide on what data we want to have, now and and later maybe even today. We go and look at night XML, or some of us look at night XML in a bit more detail. Just trying to make some sense of that code, and see how does the representation work in their system. And then sort of with that kno... | 1 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0015141_0015737 | EN2001a | [{"row": 1444, "id": "AMI_EN2001a_H04_MEO069_0015141_0015737", "speaker": "MEO069", "offset_s": 0.217, "dur_s": 5.96}, {"row": 879, "id": "AMI_EN2001a_H00_MEE068_0015547_0015635", "speaker": "MEE068", "offset_s": 4.277, "dur_s": 0.88}, {"row": 759, "id": "AMI_EN2001a_H00_MEE068_0015635_0016140", "speaker": "MEE068", "o... | 0.05 | 10.547 | <SPK_1>And this is how everyone can access it. And then we should be able to go from there.<SPK_2>嗯。 Has anyone actually looked at the Java code for the XML?<SPK_1>No.<SPK_3>No.<SPK_2>Hmm. | 3 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0016157_0017488 | EN2001a | [{"row": 796, "id": "AMI_EN2001a_H04_MEO069_0016157_0017488", "speaker": "MEO069", "offset_s": 0.483, "dur_s": 13.31}, {"row": 488, "id": "AMI_EN2001a_H04_MEO069_0017488_0017761", "speaker": "MEO069", "offset_s": 13.793, "dur_s": 2.73}, {"row": 960, "id": "AMI_EN2001a_H04_MEO069_0017761_0018162", "speaker": "MEO069", "... | 0.457 | 20.99 | <SPK_1>I've looked at the documentation and like seen enough to make me think that we want to use the NADX model framework because, um, they have a good event model that synchronizes sort of the data and, in every display element. So that takes a lot of work away from us. Sort of that would be a reason for staying with... | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0018162_0018710 | EN2001a | [{"row": 1534, "id": "AMI_EN2001a_H04_MEO069_0018162_0018710", "speaker": "MEO069", "offset_s": 0.13, "dur_s": 5.48}, {"row": 871, "id": "AMI_EN2001a_H00_MEE068_0018680_0018752", "speaker": "MEE068", "offset_s": 5.31, "dur_s": 0.72}, {"row": 940, "id": "AMI_EN2001a_H04_MEO069_0018710_0019626", "speaker": "MEO069", "off... | 0.433 | 16.513 | <SPK_1>But beyond that, I haven't looked at it at all. Which is something we should really do.<SPK_2>Yeah.<SPK_1>Who actually like for this whole discussion? I mean, who of us, is doing stuff that, is happening online and who of us is doing stuff that's happening offline. Like my data is coming. | 2 | 4 |
EN2001a:AMI_EN2001a_H02_FEO065_0021133_0021442 | EN2001a | [{"row": 257, "id": "AMI_EN2001a_H02_FEO065_0021133_0021442", "speaker": "FEO065", "offset_s": 0.007, "dur_s": 3.09}, {"row": 1612, "id": "AMI_EN2001a_H04_MEO069_0021316_0021363", "speaker": "MEO069", "offset_s": 1.837, "dur_s": 0.47}, {"row": 661, "id": "AMI_EN2001a_H02_FEO065_0021442_0022058", "speaker": "FEO065", "o... | 0.274 | 25.461 | <SPK_1>The basic word importance is offline as well.<SPK_2>Yeah.<SPK_1>The combined measure might not be if we want to wait, what a user has typed into the search.<SPK_2>Okay. Okay.<SPK_3>The mine's gonna be mostly using the offline. But the actual stuff is doing will be online. But it won't be very processor intensive... | 3 | 10 |
EN2001a:AMI_EN2001a_H04_MEO069_0023651_0024471 | EN2001a | [{"row": 1265, "id": "AMI_EN2001a_H04_MEO069_0023651_0024471", "speaker": "MEO069", "offset_s": 0.412, "dur_s": 8.2}, {"row": 67, "id": "AMI_EN2001a_H04_MEO069_0024471_0025049", "speaker": "MEO069", "offset_s": 8.612, "dur_s": 5.78}, {"row": 1051, "id": "AMI_EN2001a_H04_MEO069_0025225_0025273", "speaker": "MEO069", "of... | 0.199 | 22.202 | <SPK_1>We don't have an extremely high degree of interaction between, sort of, our modules that create the stuff and the interface. So, the interface is mainly while it's running, just working on data that's just loaded from a file, I guess. There isn't.<SPK_2>Yeah I don't know about the search functionality that might... | 3 | 7 |
EN2001a:AMI_EN2001a_H04_MEO069_0025836_0026585 | EN2001a | [{"row": 53, "id": "AMI_EN2001a_H04_MEO069_0025836_0026585", "speaker": "MEO069", "offset_s": 0.247, "dur_s": 7.49}, {"row": 98, "id": "AMI_EN2001a_H00_MEE068_0026583_0026620", "speaker": "MEE068", "offset_s": 7.717, "dur_s": 0.37}, {"row": 1509, "id": "AMI_EN2001a_H04_MEO069_0026585_0026933", "speaker": "MEO069", "off... | 0.001 | 20.137 | <SPK_1>Yeah, the search is I guess the search is sort of a strange beast anyway because for the search we're leaving the Night XML framework.<SPK_2>Hmm.<SPK_1>Um, but that's still sort of that's good. That means that at least like we don't have the type of situation where somebody has to do like a billion calculations ... | 2 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0027825_0028518 | EN2001a | [{"row": 1557, "id": "AMI_EN2001a_H04_MEO069_0027825_0028518", "speaker": "MEO069", "offset_s": 0.163, "dur_s": 6.93}, {"row": 801, "id": "AMI_EN2001a_H04_MEO069_0028518_0028968", "speaker": "MEO069", "offset_s": 7.093, "dur_s": 4.5}, {"row": 1006, "id": "AMI_EN2001a_H04_MEO069_0028968_0029312", "speaker": "MEO069", "o... | 0.122 | 22.085 | <SPK_1>Because that would make it a lot more like that would mean that our interface for the data would have to be a lot more careful about how it performs and and everything. And nobody is modifying that data at online time at all. It seems. Nobody is making any changes to the actual data online.<SPK_2>I think so.<SPK... | 2 | 7 |
EN2001a:AMI_EN2001a_H04_MEO069_0030029_0030614 | EN2001a | [{"row": 420, "id": "AMI_EN2001a_H04_MEO069_0030029_0030614", "speaker": "MEO069", "offset_s": 0.284, "dur_s": 5.85}, {"row": 976, "id": "AMI_EN2001a_H03_MEE067_0031094_0031358", "speaker": "MEE067", "offset_s": 7.634, "dur_s": 2.64}, {"row": 20, "id": "AMI_EN2001a_H03_MEE067_0031358_0031560", "speaker": "MEE067", "off... | 0.096 | 26.939 | <SPK_1>Which isn't doing much with the data except for sort of selecting a piece of it and displaying it.<SPK_2>Are we still gonna go for dumping it into a database? Are we still gonna dump it into a database?<SPK_1>Will some parts relevant for the search? Yes. I'd say so.<SPK_2>Because if we are, I reckon we should al... | 2 | 9 |
EN2001a:AMI_EN2001a_H03_MEE067_0033015_0033618 | EN2001a | [{"row": 326, "id": "AMI_EN2001a_H03_MEE067_0033015_0033618", "speaker": "MEE067", "offset_s": 0.402, "dur_s": 6.03}, {"row": 1514, "id": "AMI_EN2001a_H03_MEE067_0033618_0034110", "speaker": "MEE067", "offset_s": 6.432, "dur_s": 4.92}, {"row": 783, "id": "AMI_EN2001a_H03_MEE067_0034110_0034202", "speaker": "MEE067", "o... | 0.484 | 12.755 | <SPK_1>We might as well dump all the fields we want into the database, calculate everything from there. Then we don't even have to worry that much about the underlying XML representation. We can just query it. | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0034338_0034715 | EN2001a | [{"row": 858, "id": "AMI_EN2001a_H04_MEO069_0034338_0034715", "speaker": "MEO069", "offset_s": 0.16, "dur_s": 3.77}, {"row": 121, "id": "AMI_EN2001a_H04_MEO069_0034882_0035121", "speaker": "MEO069", "offset_s": 5.43, "dur_s": 2.39}, {"row": 130, "id": "AMI_EN2001a_H04_MEO069_0035121_0035759", "speaker": "MEO069", "offs... | 0.04 | 14.24 | <SPK_1>Yeah, but nobody of us is doing much of searching from the data in the online stage. And for altogether, like the display itself. I think we are easier if we if it's sitting on the XML than if it's sitting on the SQL stuff because if it's sitting on the XML. | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0035759_0036968 | EN2001a | [{"row": 193, "id": "AMI_EN2001a_H04_MEO069_0035759_0036968", "speaker": "MEO069", "offset_s": 0.055, "dur_s": 12.09}, {"row": 1640, "id": "AMI_EN2001a_H04_MEO069_0036968_0037277", "speaker": "MEO069", "offset_s": 12.145, "dur_s": 3.09}, {"row": 917, "id": "AMI_EN2001a_H04_MEO069_0037277_0037573", "speaker": "MEO069", ... | 0.466 | 18.661 | <SPK_1>We have the the Night XML framework with all its functionality for synchronizing, through all the different levels, whenever there's a change, whenever some things moving forward and stuff. And we can just more or less look at their code, like how their player moves forward. And how that moving forward is repres... | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0037573_0038440 | EN2001a | [{"row": 751, "id": "AMI_EN2001a_H04_MEO069_0037573_0038440", "speaker": "MEO069", "offset_s": 0.274, "dur_s": 8.67}, {"row": 614, "id": "AMI_EN2001a_H04_MEO069_0038440_0039285", "speaker": "MEO069", "offset_s": 8.944, "dur_s": 8.45}] | 0.353 | 17.747 | <SPK_1>So I think in the actual browser itself, I don't want to sit on the SQL. If we can sit on the XML, because sitting on the XML, we have all we have so much help. And for for like the the calculations that we're doing, apart from the search, it seems everyone needs some special representations anyway. | 1 | 2 |
EN2001a:AMI_EN2001a_H03_MEE067_0039366_0039887 | EN2001a | [{"row": 346, "id": "AMI_EN2001a_H03_MEE067_0039366_0039887", "speaker": "MEE067", "offset_s": 0.482, "dur_s": 5.21}, {"row": 634, "id": "AMI_EN2001a_H04_MEO069_0040012_0040165", "speaker": "MEO069", "offset_s": 6.942, "dur_s": 1.53}, {"row": 601, "id": "AMI_EN2001a_H03_MEE067_0040119_0040152", "speaker": "MEE067", "of... | 0.27 | 22.582 | <SPK_1>Well if we're gonna do that, we should try and store everything in an XML format as well.<SPK_2>You mean our results.<SPK_1>Yeah.<SPK_2>Yeah, in in the night XML XML format. So with the timestamps and stuff, so that it's easy to to tie together things.<SPK_3>Yes.<SPK_1>Yeah.<SPK_2>What I'm like, what we have to ... | 3 | 7 |
EN2001a:AMI_EN2001a_H04_MEO069_0041711_0042569 | EN2001a | [{"row": 1341, "id": "AMI_EN2001a_H04_MEO069_0041711_0042569", "speaker": "MEO069", "offset_s": 0.222, "dur_s": 8.58}, {"row": 733, "id": "AMI_EN2001a_H04_MEO069_0042569_0043255", "speaker": "MEO069", "offset_s": 8.803, "dur_s": 6.86}, {"row": 761, "id": "AMI_EN2001a_H04_MEO069_0043255_0043779", "speaker": "MEO069", "o... | 0.294 | 21.196 | <SPK_1>Like this idea that sort of if you start with the whole meeting series as one entity, as one thing that you display, as one whole sort of. That then, the individual chunks are the individual meetings, whereas and then you can click on a meeting. And then sort of the meeting is the whole thing, and the chunks are... | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0043779_0044242 | EN2001a | [{"row": 16, "id": "AMI_EN2001a_H04_MEO069_0043779_0044242", "speaker": "MEO069", "offset_s": 0.288, "dur_s": 4.63}, {"row": 1413, "id": "AMI_EN2001a_H04_MEO069_0044388_0044746", "speaker": "MEO069", "offset_s": 6.378, "dur_s": 3.58}, {"row": 19, "id": "AMI_EN2001a_H04_MEO069_0044746_0044987", "speaker": "MEO069", "off... | 0.192 | 12.56 | <SPK_1>That means sort of we have multiple levels of representation, which we probably. If we if we do it this way, like we we have to discuss that if we do it this way. Then we should probably find some abstraction models. | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0044987_0046015 | EN2001a | [{"row": 787, "id": "AMI_EN2001a_H04_MEO069_0044987_0046015", "speaker": "MEO069", "offset_s": 0.093, "dur_s": 10.28}, {"row": 1148, "id": "AMI_EN2001a_H04_MEO069_0046015_0046186", "speaker": "MEO069", "offset_s": 10.373, "dur_s": 1.71}, {"row": 376, "id": "AMI_EN2001a_H00_MEE068_0046098_0046122", "speaker": "MEE068", ... | 0.095 | 12.178 | <SPK_1>So that the interface, in the sense like, deals with it as if it's the same, so that the interface doesn't really have to worry whether it's a meeting, in the whole meeting series or a segment within a meeting. You know what I mean.<SPK_2>Uh huh.<SPK_3>Yes. | 3 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0046186_0046500 | EN2001a | [{"row": 1412, "id": "AMI_EN2001a_H04_MEO069_0046186_0046500", "speaker": "MEO069", "offset_s": 0.238, "dur_s": 3.14}, {"row": 481, "id": "AMI_EN2001a_H04_MEO069_0046500_0047294", "speaker": "MEO069", "offset_s": 3.378, "dur_s": 7.94}, {"row": 1042, "id": "AMI_EN2001a_H04_MEO069_0047294_0047426", "speaker": "MEO069", "... | 0.328 | 19.107 | <SPK_1>And that's probably stuff that we have to sort of like process twice. Like, for example, that like the summary of a meeting within the whole meeting corpus or meeting series. Is meeting series a good word for that? I don't really know what how to call it. You know what I mean? Like not not the whole corpus, but ... | 2 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0048009_0048470 | EN2001a | [{"row": 1344, "id": "AMI_EN2001a_H04_MEO069_0048009_0048470", "speaker": "MEO069", "offset_s": 0.438, "dur_s": 4.61}, {"row": 465, "id": "AMI_EN2001a_H04_MEO069_0048661_0048916", "speaker": "MEO069", "offset_s": 6.548, "dur_s": 2.55}] | 0.379 | 9.477 | <SPK_1>So in the meeting series, so that a summary for a meeting within the meeting series, a sort of. Compiled offline by a summary module. | 1 | 2 |
EN2001a:AMI_EN2001a_H04_MEO069_0048916_0049376 | EN2001a | [{"row": 471, "id": "AMI_EN2001a_H04_MEO069_0048916_0049376", "speaker": "MEO069", "offset_s": 0.449, "dur_s": 4.6}, {"row": 1551, "id": "AMI_EN2001a_H04_MEO069_0049376_0049466", "speaker": "MEO069", "offset_s": 5.049, "dur_s": 0.9}, {"row": 1617, "id": "AMI_EN2001a_H01_FEO066_0049430_0049647", "speaker": "FEO066", "of... | 0.421 | 22.5 | <SPK_1>And that is separate from a summary of a segment within a meeting. Because I don't think we can.<SPK_2>I'm not sure.<SPK_3>Well, we don't even need to do that because we got our information density calculated offline. So all we do is treat the whole lot as one massive document. I mean, it's not going to be so bi... | 3 | 7 |
EN2001a:AMI_EN2001a_H01_FEO066_0051155_0052234 | EN2001a | [{"row": 1416, "id": "AMI_EN2001a_H01_FEO066_0051155_0052234", "speaker": "FEO066", "offset_s": 0.196, "dur_s": 10.79}, {"row": 1199, "id": "AMI_EN2001a_H01_FEO066_0052234_0052845", "speaker": "FEO066", "offset_s": 10.986, "dur_s": 6.11}, {"row": 1045, "id": "AMI_EN2001a_H04_MEO069_0052921_0053066", "speaker": "MEO069"... | 0.27 | 27.506 | <SPK_1>I thought we would just have like um one big summary, um with only different importance levels, um, displayed on it. And depending on what our zoom level is, we just display a part of it.<SPK_2>So are we doing that at all levels? Are we?<SPK_1>I mean, we would have one very big thing offline. And from that, we w... | 2 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0053867_0054209 | EN2001a | [{"row": 99, "id": "AMI_EN2001a_H04_MEO069_0053867_0054209", "speaker": "MEO069", "offset_s": 0.406, "dur_s": 3.42}, {"row": 901, "id": "AMI_EN2001a_H01_FEO066_0054204_0054285", "speaker": "FEO066", "offset_s": 3.776, "dur_s": 0.81}, {"row": 427, "id": "AMI_EN2001a_H01_FEO066_0054285_0055276", "speaker": "FEO066", "off... | 0.138 | 23.514 | <SPK_1>And just have different like fine grainness levels, sort of.<SPK_2>Yes. So for example, you would give a high value to those sequences you want to display in the meeting series summary.<SPK_1>Hmm. Okay. So the only thing that yeah, so the only thing that would happen basically if I double click, let's say from t... | 2 | 8 |
EN2001a:AMI_EN2001a_H04_MEO069_0056164_0056566 | EN2001a | [{"row": 1223, "id": "AMI_EN2001a_H04_MEO069_0056164_0056566", "speaker": "MEO069", "offset_s": 0.295, "dur_s": 4.02}, {"row": 1146, "id": "AMI_EN2001a_H01_FEO066_0056607_0056837", "speaker": "FEO066", "offset_s": 4.725, "dur_s": 2.3}, {"row": 221, "id": "AMI_EN2001a_H03_MEE067_0056786_0056925", "speaker": "MEE067", "o... | 0.448 | 25.722 | <SPK_1>Like the start and the end position changes, and the zoom level changes.<SPK_2>That was what I I thought.<SPK_3>I think you can do it online.<SPK_1>I, I thought we couldn't do.<SPK_2>Okay.<SPK_1>Like I was under the impression that we couldn't do that because we couldn't load the data for all that. But I don't k... | 4 | 9 |
EN2001a:AMI_EN2001a_H03_MEE067_0058716_0058885 | EN2001a | [{"row": 1103, "id": "AMI_EN2001a_H03_MEE067_0058716_0058885", "speaker": "MEE067", "offset_s": 0.225, "dur_s": 1.69}, {"row": 1046, "id": "AMI_EN2001a_H03_MEE067_0058885_0059458", "speaker": "MEE067", "offset_s": 1.915, "dur_s": 5.73}, {"row": 1247, "id": "AMI_EN2001a_H04_MEO069_0059647_0059865", "speaker": "MEO069", ... | 0.29 | 27.935 | <SPK_1>And that's all calculated offline. So all you're really doing is sorting a list. It's the computationally hard part of it.<SPK_2>So I'm not sure if I got it. Oh.<SPK_1>Well, like the idea is we're calculating our information density all offline first for every utterance, in the whole corpus, right.<SPK_2>Uh huh.... | 2 | 9 |
EN2001a:AMI_EN2001a_H03_MEE067_0061497_0062017 | EN2001a | [{"row": 1580, "id": "AMI_EN2001a_H03_MEE067_0061497_0062017", "speaker": "MEE067", "offset_s": 0.458, "dur_s": 5.2}, {"row": 1053, "id": "AMI_EN2001a_H04_MEO069_0061788_0061816", "speaker": "MEO069", "offset_s": 3.369, "dur_s": 0.28}, {"row": 623, "id": "AMI_EN2001a_H03_MEE067_0062017_0062382", "speaker": "MEE067", "o... | 0.498 | 17.007 | <SPK_1>And then all you have to do is sort them by information density.<SPK_2>Mhm.<SPK_1>Like maybe weighted with the search terms, and then extract them.<SPK_2>Okay. So.<SPK_1>I don't think it's too slow to do online to be honest.<SPK_2>Okay. I was just worried about the total memory complexity of it. | 2 | 8 |
EN2001a:AMI_EN2001a_H04_MEO069_0063102_0063796 | EN2001a | [{"row": 799, "id": "AMI_EN2001a_H04_MEO069_0063102_0063796", "speaker": "MEO069", "offset_s": 0.306, "dur_s": 6.94}, {"row": 1091, "id": "AMI_EN2001a_H04_MEO069_0063796_0063925", "speaker": "MEO069", "offset_s": 7.246, "dur_s": 1.29}, {"row": 1378, "id": "AMI_EN2001a_H04_MEO069_0063925_0064019", "speaker": "MEO069", "... | 0.041 | 25.228 | <SPK_1>But I completely admit. I mean, I just sort of like, took that from something that Jonathan once said about not loading everything. But maybe I was just wrong about it. How many attributes.<SPK_2>Is that.<SPK_3>But I think the difference might be that we just want to have the words. And that's not so much, what ... | 3 | 9 |
EN2001a:AMI_EN2001a_H01_FEO066_0065545_0065707 | EN2001a | [{"row": 1095, "id": "AMI_EN2001a_H01_FEO066_0065545_0065707", "speaker": "FEO066", "offset_s": 0.423, "dur_s": 1.62}, {"row": 1242, "id": "AMI_EN2001a_H00_MEE068_0065547_0065568", "speaker": "MEE068", "offset_s": 0.442, "dur_s": 0.21}, {"row": 213, "id": "AMI_EN2001a_H04_MEO069_0065590_0065678", "speaker": "MEO069", "... | 0.315 | 10.628 | <SPK_1>All the sound fights.<SPK_2>Hmm.<SPK_3>Yeah. Yeah. So what we have is we would have a word. Like we would have words with some priority levels. | 3 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0066569_0067627 | EN2001a | [{"row": 106, "id": "AMI_EN2001a_H04_MEO069_0066569_0067627", "speaker": "MEO069", "offset_s": 0.059, "dur_s": 10.58}] | 0.366 | 11.004 | <SPK_1>And that would basically be it already because even the selection would the summaries automatically feed from just how prioritized an individual word or how prioritized an individual utterance is. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0067627_0068143 | EN2001a | [{"row": 802, "id": "AMI_EN2001a_H04_MEO069_0067627_0068143", "speaker": "MEO069", "offset_s": 0.166, "dur_s": 5.16}, {"row": 1224, "id": "AMI_EN2001a_H04_MEO069_0068143_0068648", "speaker": "MEO069", "offset_s": 5.326, "dur_s": 5.05}, {"row": 1119, "id": "AMI_EN2001a_H03_MEE067_0068569_0068829", "speaker": "MEE067", "... | 0.397 | 12.584 | <SPK_1>Or are the summaries sort of refined from that and made by a machine to make sentences and stuff? Or are they just sort of taking out the words of the highest priority and then the words of the second highest priority?<SPK_2>Well, on the utterance level, I was thinking. | 2 | 3 |
EN2001a:AMI_EN2001a_H03_MEE067_0068829_0069254 | EN2001a | [{"row": 611, "id": "AMI_EN2001a_H03_MEE067_0068829_0069254", "speaker": "MEE067", "offset_s": 0.073, "dur_s": 4.25}, {"row": 1089, "id": "AMI_EN2001a_H04_MEO069_0068831_0068989", "speaker": "MEO069", "offset_s": 0.093, "dur_s": 1.58}, {"row": 1043, "id": "AMI_EN2001a_H04_MEO069_0069206_0069395", "speaker": "MEO069", "... | 0.05 | 21.563 | <SPK_1>So the utterances with the highest like mean information density.<SPK_2>And yeah. Okay. Are we doing it on the whole thing on the atonance level?<SPK_3>Um.<SPK_2>Are we doing it on word level, like the information density calculation?<SPK_1>Well, the trouble with doing it on the word level is if you want the aud... | 4 | 13 |
EN2001a:AMI_EN2001a_H04_MEO069_0070973_0071472 | EN2001a | [{"row": 404, "id": "AMI_EN2001a_H04_MEO069_0070973_0071472", "speaker": "MEO069", "offset_s": 0.287, "dur_s": 4.99}, {"row": 175, "id": "AMI_EN2001a_H03_MEE067_0071024_0071050", "speaker": "MEE067", "offset_s": 0.797, "dur_s": 0.26}, {"row": 1052, "id": "AMI_EN2001a_H03_MEE067_0071395_0071549", "speaker": "MEE067", "o... | 0.023 | 21.17 | <SPK_1>We should really maybe we can do that together at some point today. That we check out how the player works.<SPK_2>Yeah. I don't think the player will do it.<SPK_1>But there's maybe some merit in altogether doing it on an atonance level in the end.<SPK_2>We'll have to buffer it.<SPK_3>I I am getting quite lost, a... | 3 | 7 |
EN2001a:AMI_EN2001a_H01_FEO066_0073059_0073342 | EN2001a | [{"row": 592, "id": "AMI_EN2001a_H01_FEO066_0073059_0073342", "speaker": "FEO066", "offset_s": 0.34, "dur_s": 2.83}, {"row": 598, "id": "AMI_EN2001a_H03_MEE067_0073408_0073626", "speaker": "MEE067", "offset_s": 3.83, "dur_s": 2.18}, {"row": 1435, "id": "AMI_EN2001a_H04_MEO069_0073667_0073708", "speaker": "MEO069", "off... | 0.267 | 25.547 | <SPK_1>I mean, do we do both, or is it the same thing?<SPK_2>While the skimming is going to use the importance.<SPK_3>Yeah.<SPK_1>Okay.<SPK_2>But like at first it's just gonna be idea.<SPK_1>So, but when when we talk about summaries, we talk about this about skimming and not about.<SPK_2>Well, mostly skimming. Yeah.<SP... | 3 | 9 |
EN2001a:AMI_EN2001a_H04_MEO069_0075553_0076370 | EN2001a | [{"row": 1653, "id": "AMI_EN2001a_H04_MEO069_0075553_0076370", "speaker": "MEO069", "offset_s": 0.303, "dur_s": 8.17}, {"row": 815, "id": "AMI_EN2001a_H01_FEO066_0076251_0076279", "speaker": "FEO066", "offset_s": 7.283, "dur_s": 0.28}] | 0.317 | 8.791 | <SPK_1>But it's also displaying a summarized version, in you know like below the below the graph the part.<SPK_2>Yeah. | 2 | 2 |
EN2001a:AMI_EN2001a_H04_MEO069_0076370_0076471 | EN2001a | [{"row": 600, "id": "AMI_EN2001a_H04_MEO069_0076370_0076471", "speaker": "MEO069", "offset_s": 0.185, "dur_s": 1.01}, {"row": 968, "id": "AMI_EN2001a_H04_MEO069_0076649_0076694", "speaker": "MEO069", "offset_s": 2.695, "dur_s": 0.45}, {"row": 1380, "id": "AMI_EN2001a_H01_FEO066_0076680_0076899", "speaker": "FEO066", "o... | 0.196 | 15.221 | <SPK_1>Maybe. Yeah.<SPK_2>Yeah, it isn't like the skimming. Isn't bad to skimme.<SPK_1>Oh yeah, it's just like there's like audio skimming and there's display skimming.<SPK_2>Yeah, but it uses same data.<SPK_1>Yeah.<SPK_3>Yeah. | 3 | 8 |
EN2001a:AMI_EN2001a_H04_MEO069_0077882_0078504 | EN2001a | [{"row": 244, "id": "AMI_EN2001a_H04_MEO069_0077882_0078504", "speaker": "MEO069", "offset_s": 0.011, "dur_s": 6.22}, {"row": 1139, "id": "AMI_EN2001a_H04_MEO069_0078504_0078549", "speaker": "MEO069", "offset_s": 6.231, "dur_s": 0.45}, {"row": 830, "id": "AMI_EN2001a_H04_MEO069_0078549_0079271", "speaker": "MEO069", "o... | 0.018 | 28.439 | <SPK_1>Maybe there's some merit of going altogether for a trans level, not even bothered to calculate. I mean, if you have to do it internally. Then you can do. But maybe like not even store the importance levels for individual words and just sort of rank utterances as a whole.<SPK_2>The nice thing about that is it wil... | 2 | 9 |
EN2001a:AMI_EN2001a_H04_MEO069_0080723_0081057 | EN2001a | [{"row": 1645, "id": "AMI_EN2001a_H04_MEO069_0080723_0081057", "speaker": "MEO069", "offset_s": 0.062, "dur_s": 3.34}, {"row": 1208, "id": "AMI_EN2001a_H04_MEO069_0081057_0081113", "speaker": "MEO069", "offset_s": 3.402, "dur_s": 0.56}, {"row": 11, "id": "AMI_EN2001a_H02_FEO065_0081159_0081631", "speaker": "FEO065", "o... | 0.092 | 27.874 | <SPK_1>If you can't find anything, I'll just sort of make the mean of the words over the utterance. You know what I mean.<SPK_2>Not quite. So what did you wanted to do? You just wanted to assign.<SPK_1>It's. What's the smallest chunk at the moment you're thinking of assigning an importance measure to? Is it a word or i... | 2 | 9 |
EN2001a:AMI_EN2001a_H04_MEO069_0083571_0083927 | EN2001a | [{"row": 817, "id": "AMI_EN2001a_H04_MEO069_0083571_0083927", "speaker": "MEO069", "offset_s": 0.468, "dur_s": 3.56}, {"row": 1358, "id": "AMI_EN2001a_H04_MEO069_0083927_0084374", "speaker": "MEO069", "offset_s": 4.028, "dur_s": 4.47}, {"row": 1105, "id": "AMI_EN2001a_H01_FEO066_0084215_0084269", "speaker": "FEO066", "... | 0.4 | 8.899 | <SPK_1>And for in the display, it's probably better if you have whole utterances than, I don't. Like, what it, it's like if you just take single words out of utterances.<SPK_2>Hmm. | 2 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0084374_0084912 | EN2001a | [{"row": 282, "id": "AMI_EN2001a_H04_MEO069_0084374_0084912", "speaker": "MEO069", "offset_s": 0.051, "dur_s": 5.38}, {"row": 1303, "id": "AMI_EN2001a_H01_FEO066_0084433_0084485", "speaker": "FEO066", "offset_s": 0.641, "dur_s": 0.52}, {"row": 1668, "id": "AMI_EN2001a_H03_MEE067_0084650_0084684", "speaker": "MEE067", "... | 0.213 | 5.644 | <SPK_1>It probably doesn't make any sense at all. Whereas if you just show important utterances, but the utterances as a whole, it makes more sense.<SPK_2>.<SPK_3>Yeah. | 3 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0084912_0085308 | EN2001a | [{"row": 352, "id": "AMI_EN2001a_H04_MEO069_0084912_0085308", "speaker": "MEO069", "offset_s": 0.323, "dur_s": 3.96}, {"row": 658, "id": "AMI_EN2001a_H04_MEO069_0085308_0085600", "speaker": "MEO069", "offset_s": 4.283, "dur_s": 2.92}] | 0.11 | 7.314 | <SPK_1>So it doesn't actually make a difference for your algorithm because it just means that if you work on a word level. Let me just mean it over the utterance. | 1 | 2 |
EN2001a:AMI_EN2001a_H01_FEO066_0085529_0086549 | EN2001a | [{"row": 1498, "id": "AMI_EN2001a_H01_FEO066_0085529_0086549", "speaker": "FEO066", "offset_s": 0.252, "dur_s": 10.2}, {"row": 1292, "id": "AMI_EN2001a_H02_FEO065_0085553_0085612", "speaker": "FEO065", "offset_s": 0.492, "dur_s": 0.59}, {"row": 1272, "id": "AMI_EN2001a_H04_MEO069_0086519_0086578", "speaker": "MEO069", ... | 0.09 | 15.142 | <SPK_1>And yeah, I think we also thought about combining that measure with the measures I get from, um, hotspots and so on.<SPK_2>Okay.<SPK_3>They are on.<SPK_1>So that would also be on assurance level I think. | 3 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0086937_0087075 | EN2001a | [{"row": 220, "id": "AMI_EN2001a_H04_MEO069_0086937_0087075", "speaker": "MEO069", "offset_s": 0.494, "dur_s": 1.38}, {"row": 1316, "id": "AMI_EN2001a_H01_FEO066_0087009_0087067", "speaker": "FEO066", "offset_s": 1.214, "dur_s": 0.58}, {"row": 798, "id": "AMI_EN2001a_H03_MEE067_0087034_0087088", "speaker": "MEE067", "o... | 0.05 | 4.365 | <SPK_1>So that's good anyway. Then yeah.<SPK_2>I think.<SPK_3>Ask.<SPK_1>Because that makes it a lot easier than to put it on a trans level. | 3 | 4 |
EN2001a:AMI_EN2001a_H03_MEE067_0087250_0088042 | EN2001a | [{"row": 128, "id": "AMI_EN2001a_H03_MEE067_0087250_0088042", "speaker": "MEE067", "offset_s": 0.366, "dur_s": 7.92}, {"row": 1559, "id": "AMI_EN2001a_H04_MEO069_0087958_0087994", "speaker": "MEO069", "offset_s": 7.446, "dur_s": 0.36}, {"row": 61, "id": "AMI_EN2001a_H04_MEO069_0087994_0088294", "speaker": "MEO069", "of... | 0.179 | 10.985 | <SPK_1>They'll need to be calculated at word level though because otherwise there won't be enough occurrences of the terms, to make any meaningful sense.<SPK_2>Oh yeah. I mean, like how how just mean does it internally? I don't know. | 2 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0088294_0088580 | EN2001a | [{"row": 496, "id": "AMI_EN2001a_H04_MEO069_0088294_0088580", "speaker": "MEO069", "offset_s": 0.085, "dur_s": 2.86}, {"row": 852, "id": "AMI_EN2001a_H03_MEE067_0088415_0088444", "speaker": "MEE067", "offset_s": 1.295, "dur_s": 0.29}, {"row": 366, "id": "AMI_EN2001a_H04_MEO069_0088580_0089083", "speaker": "MEO069", "of... | 0.459 | 24.544 | <SPK_1>But it's probably yeah. You probably have to work on vert levels for importance.<SPK_2>Yeah.<SPK_1>But there should be ways of easily going from a world level to another level.<SPK_2>Yeah, I reckon you can just mean it over the sentence.<SPK_1>Okay.<SPK_3>Yeah, but how about those words which don't carry any mea... | 3 | 10 |
EN2001a:AMI_EN2001a_H02_FEO065_0090775_0091688 | EN2001a | [{"row": 1596, "id": "AMI_EN2001a_H02_FEO065_0090775_0091688", "speaker": "FEO065", "offset_s": 0.029, "dur_s": 9.13}, {"row": 530, "id": "AMI_EN2001a_H03_MEE067_0091714_0091999", "speaker": "MEE067", "offset_s": 9.419, "dur_s": 2.85}, {"row": 652, "id": "AMI_EN2001a_H03_MEE067_0091999_0092527", "speaker": "MEE067", "o... | 0.483 | 18.032 | <SPK_1>And there are quite unimportant words in there, but quite important words as well. I think we should just disregard the.<SPK_2>Maybe we should have like a cut off. So a word only gets a value if it's above a certain threshold. | 2 | 3 |
EN2001a:AMI_EN2001a_H03_MEE067_0092527_0093049 | EN2001a | [{"row": 1421, "id": "AMI_EN2001a_H03_MEE067_0092527_0093049", "speaker": "MEE067", "offset_s": 0.171, "dur_s": 5.22}, {"row": 1365, "id": "AMI_EN2001a_H02_FEO065_0092528_0092595", "speaker": "FEO065", "offset_s": 0.181, "dur_s": 0.67}, {"row": 317, "id": "AMI_EN2001a_H02_FEO065_0093105_0093128", "speaker": "FEO065", "... | 0.423 | 22.424 | <SPK_1>So anything that has less than say 0.5 importance gets assigned to zero.<SPK_2>Okay. Alright.<SPK_3>Or we do a pre-filtering of sort of the whole thing.<SPK_1>Yeah, that's the other.<SPK_3>Sort of like, but that like the problem with that is it's easy to do in the text. But that would mean it would still play th... | 3 | 9 |
EN2001a:AMI_EN2001a_H04_MEO069_0094993_0095433 | EN2001a | [{"row": 1196, "id": "AMI_EN2001a_H04_MEO069_0094993_0095433", "speaker": "MEO069", "offset_s": 0.221, "dur_s": 4.4}, {"row": 1112, "id": "AMI_EN2001a_H04_MEO069_0095433_0095803", "speaker": "MEO069", "offset_s": 4.621, "dur_s": 3.7}, {"row": 133, "id": "AMI_EN2001a_H03_MEE067_0095821_0095998", "speaker": "MEE067", "of... | 0.298 | 12.26 | <SPK_1>I think, before we can like answer that specific question, how we deal with that. It's probably good for us to look at what the audio player is capable of doing.<SPK_2>I think we have to buffer the audio. But I don't think it'll be very hard.<SPK_1>Yes. | 2 | 5 |
EN2001a:AMI_EN2001a_H03_MEE067_0096167_0096348 | EN2001a | [{"row": 903, "id": "AMI_EN2001a_H03_MEE067_0096167_0096348", "speaker": "MEE067", "offset_s": 0.205, "dur_s": 1.81}, {"row": 599, "id": "AMI_EN2001a_H04_MEO069_0096447_0096554", "speaker": "MEO069", "offset_s": 3.005, "dur_s": 1.07}, {"row": 215, "id": "AMI_EN2001a_H04_MEO069_0096554_0096730", "speaker": "MEO069", "of... | 0.236 | 11.141 | <SPK_1>I think it'll be like an hour or two's work.<SPK_2>So, what do you mean by buffering? Like you think directly feeding.<SPK_1>Like just build, and another, wave file essentially.<SPK_3>Yeah, you just concatenate them together. | 3 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0097129_0097681 | EN2001a | [{"row": 999, "id": "AMI_EN2001a_H04_MEO069_0097129_0097681", "speaker": "MEO069", "offset_s": 0.156, "dur_s": 5.52}, {"row": 1167, "id": "AMI_EN2001a_H03_MEE067_0097179_0097414", "speaker": "MEE067", "offset_s": 0.656, "dur_s": 2.35}, {"row": 40, "id": "AMI_EN2001a_H03_MEE067_0097639_0097766", "speaker": "MEE067", "of... | 0.254 | 18.97 | <SPK_1>But yeah, but not but not stored on the hard disk and then loaded in, but loaded indirectly from memory.<SPK_2>Yeah, I mean I bet there will be packages. In memory, yeah. So just like there's bound to be like a media wave object or something like that. And just build one in memory.<SPK_1>But it's probably a stre... | 2 | 9 |
EN2001a:AMI_EN2001a_H03_MEE067_0098944_0099297 | EN2001a | [{"row": 1001, "id": "AMI_EN2001a_H03_MEE067_0098944_0099297", "speaker": "MEE067", "offset_s": 0.125, "dur_s": 3.53}, {"row": 1153, "id": "AMI_EN2001a_H03_MEE067_0099297_0099721", "speaker": "MEE067", "offset_s": 3.655, "dur_s": 4.24}, {"row": 43, "id": "AMI_EN2001a_H04_MEO069_0099586_0099696", "speaker": "MEO069", "o... | 0.419 | 12.604 | <SPK_1>But it must have, like, classes for dealing with files. Nivit has classes for concatenating files. He can do it in memory.<SPK_2>Okay. Okay, so I mean, so that means that there's probably even if you go on a per occurrence level. | 2 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0100150_0100741 | EN2001a | [{"row": 180, "id": "AMI_EN2001a_H04_MEO069_0100150_0100741", "speaker": "MEO069", "offset_s": 0.371, "dur_s": 5.91}, {"row": 1575, "id": "AMI_EN2001a_H04_MEO069_0100741_0101536", "speaker": "MEO069", "offset_s": 6.281, "dur_s": 7.95}, {"row": 1474, "id": "AMI_EN2001a_H04_MEO069_0101536_0101650", "speaker": "MEO069", "... | 0.006 | 15.377 | <SPK_1>There's still some merit on within utterances cutting out stuff which clearly isn't relevant at all, and that maybe also for the audio we'd have to do. So, let's say we play the whole phrase, but then in addition to that, we have some information that says minus that part or something. That's okay that we can do... | 1 | 3 |
EN2001a:AMI_EN2001a_H03_MEE067_0101660_0102614 | EN2001a | [{"row": 368, "id": "AMI_EN2001a_H03_MEE067_0101660_0102614", "speaker": "MEE067", "offset_s": 0.14, "dur_s": 9.54}] | 0.023 | 9.703 | <SPK_1>Well, what I think I might try and build is basically a class that you just feed it a linked list of different waveforms. | 1 | 1 |
EN2001a:AMI_EN2001a_H03_MEE067_0102614_0103261 | EN2001a | [{"row": 140, "id": "AMI_EN2001a_H03_MEE067_0102614_0103261", "speaker": "MEE067", "offset_s": 0.176, "dur_s": 6.47}, {"row": 1143, "id": "AMI_EN2001a_H04_MEO069_0103214_0103633", "speaker": "MEO069", "offset_s": 6.176, "dur_s": 4.19}] | 0.477 | 10.843 | <SPK_1>It will just string them all together with like maybe, I don't know tenth of a second silence in between each one or something like that.<SPK_2>Yeah, maybe even. I mean, that sort of that depends on how how advanced we get. | 2 | 2 |
EN2001a:AMI_EN2001a_H04_MEO069_0103633_0104019 | EN2001a | [{"row": 1326, "id": "AMI_EN2001a_H04_MEO069_0103633_0104019", "speaker": "MEO069", "offset_s": 0.473, "dur_s": 3.86}, {"row": 1340, "id": "AMI_EN2001a_H04_MEO069_0104019_0104613", "speaker": "MEO069", "offset_s": 4.333, "dur_s": 5.94}, {"row": 881, "id": "AMI_EN2001a_H03_MEE067_0104116_0104228", "speaker": "MEE067", "... | 0.18 | 13.903 | <SPK_1>If maybe if you realize that there's massive differences in in gain or in something. You can probably just make some simple, simple normalization, but that really depends on how much time you have and in how much it's necessary.<SPK_2>Normalize it. Yeah.<SPK_1>Yeah, if like I I don't know anything about Audion, ... | 2 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0104958_0105645 | EN2001a | [{"row": 1216, "id": "AMI_EN2001a_H04_MEO069_0104958_0105645", "speaker": "MEO069", "offset_s": 0.358, "dur_s": 6.87}, {"row": 1552, "id": "AMI_EN2001a_H04_MEO069_0105645_0105848", "speaker": "MEO069", "offset_s": 7.228, "dur_s": 2.03}, {"row": 889, "id": "AMI_EN2001a_H04_MEO069_0105848_0106916", "speaker": "MEO069", "... | 0.311 | 20.248 | <SPK_1>So if you find that the player accepts some input from memory, if it's easy to do, then I guess that's that's fairly doable. So, but that means in the general structure, we're actually quite lucky. So we we have, we load into memory for the whole series of meetings, just the utterances and rankings for the utter... | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0106916_0107170 | EN2001a | [{"row": 621, "id": "AMI_EN2001a_H04_MEO069_0106916_0107170", "speaker": "MEO069", "offset_s": 0.325, "dur_s": 2.54}, {"row": 518, "id": "AMI_EN2001a_H03_MEE067_0107283_0107415", "speaker": "MEE067", "offset_s": 3.995, "dur_s": 1.32}, {"row": 218, "id": "AMI_EN2001a_H01_FEO066_0107332_0107405", "speaker": "FEO066", "of... | 0.207 | 12.003 | <SPK_1>Well, I guess that goes with the atrans. Who's speaking?<SPK_2>Oh yeah, yeah we need that.<SPK_3>Yes, sure.<SPK_1>Because then we can also do the display about who's speaking.<SPK_2>We also really want to be able to search by who's speaking as well.<SPK_1>Yeah. | 3 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0108365_0109244 | EN2001a | [{"row": 85, "id": "AMI_EN2001a_H04_MEO069_0108365_0109244", "speaker": "MEO069", "offset_s": 0.167, "dur_s": 8.79}] | 0.096 | 9.053 | <SPK_1>But I'm I'm still confused because I thought like that's just what Jonathan said that we that we can't do like load a massive document of that size. | 1 | 1 |
EN2001a:AMI_EN2001a_H03_MEE067_0109242_0109529 | EN2001a | [{"row": 1339, "id": "AMI_EN2001a_H03_MEE067_0109242_0109529", "speaker": "MEE067", "offset_s": 0.189, "dur_s": 2.87}, {"row": 1264, "id": "AMI_EN2001a_H04_MEO069_0109244_0109319", "speaker": "MEO069", "offset_s": 0.209, "dur_s": 0.75}, {"row": 148, "id": "AMI_EN2001a_H04_MEO069_0110014_0110547", "speaker": "MEO069", "... | 0.319 | 12.378 | <SPK_1>It doesn't matter because all the calculation is done offline.<SPK_2>The other end. The other end, I mean, it shouldn't be like should be like fifty megabyte in RAM or something. It shouldn't be massive. Actually, fifty hundred megabyte is quite big in the room.<SPK_3>Hmm. | 3 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0111599_0111834 | EN2001a | [{"row": 116, "id": "AMI_EN2001a_H04_MEO069_0111599_0111834", "speaker": "MEO069", "offset_s": 0.207, "dur_s": 2.35}, {"row": 1449, "id": "AMI_EN2001a_H04_MEO069_0112011_0112650", "speaker": "MEO069", "offset_s": 4.057, "dur_s": 6.39}, {"row": 1202, "id": "AMI_EN2001a_H04_MEO069_0112650_0112818", "speaker": "MEO069", "... | 0.284 | 26.451 | <SPK_1>Just thinking. What's this sim? So. We do get an error message with the project if we load everything into the project with all the data they load. So we know that doesn't work. So our hope is essentially that we load less into it.<SPK_2>Yes.<SPK_3>Yeah.<SPK_1>What's this lazy loading thing? Somebody explain laz... | 3 | 10 |
EN2001a:AMI_EN2001a_H04_MEO069_0114200_0114333 | EN2001a | [{"row": 532, "id": "AMI_EN2001a_H04_MEO069_0114200_0114333", "speaker": "MEO069", "offset_s": 0.209, "dur_s": 1.33}, {"row": 589, "id": "AMI_EN2001a_H04_MEO069_0114333_0114581", "speaker": "MEO069", "offset_s": 1.539, "dur_s": 2.48}, {"row": 811, "id": "AMI_EN2001a_H04_MEO069_0114581_0115056", "speaker": "MEO069", "of... | 0.351 | 14.89 | <SPK_1>Okay. So that is that only by type of file. Like if if if the same thing is in different files, would it then maybe, like you know. If if utterances are split over three or ten or one hundred different files. | 1 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0115633_0116009 | EN2001a | [{"row": 348, "id": "AMI_EN2001a_H04_MEO069_0115633_0116009", "speaker": "MEO069", "offset_s": 0.223, "dur_s": 3.76}, {"row": 906, "id": "AMI_EN2001a_H00_MEE068_0116076_0116436", "speaker": "MEE068", "offset_s": 4.653, "dur_s": 3.6}, {"row": 1298, "id": "AMI_EN2001a_H04_MEO069_0116340_0116507", "speaker": "MEO069", "of... | 0.023 | 21.016 | <SPK_1>Is there a chance maybe that it doesn't try to load them all into memory at the same time, but just<SPK_2>Yeah I think that's idea just loads the particular ones that need.<SPK_1>So why does it fail then in the first place? Then it shouldn't ever fail because then it should never.<SPK_2>But if you were doing a s... | 3 | 7 |
EN2001a:AMI_EN2001a_H00_MEE068_0117919_0117947 | EN2001a | [{"row": 1008, "id": "AMI_EN2001a_H00_MEE068_0117919_0117947", "speaker": "MEE068", "offset_s": 0.264, "dur_s": 0.28}, {"row": 1489, "id": "AMI_EN2001a_H01_FEO066_0118082_0118130", "speaker": "FEO066", "offset_s": 1.894, "dur_s": 0.48}, {"row": 967, "id": "AMI_EN2001a_H01_FEO066_0118130_0118441", "speaker": "FEO066", "... | 0.079 | 5.563 | <SPK_1>Hmm.<SPK_2>Yes, it does. It's a defined baseline on what it loads. | 2 | 3 |
EN2001a:AMI_EN2001a_H01_FEO066_0118441_0119776 | EN2001a | [{"row": 1622, "id": "AMI_EN2001a_H01_FEO066_0118441_0119776", "speaker": "FEO066", "offset_s": 0.378, "dur_s": 13.35}, {"row": 26, "id": "AMI_EN2001a_H04_MEO069_0119247_0119290", "speaker": "MEO069", "offset_s": 8.438, "dur_s": 0.43}, {"row": 103, "id": "AMI_EN2001a_H01_FEO066_0119776_0119995", "speaker": "FEO066", "o... | 0.281 | 17.968 | <SPK_1>For example, it loads all the utterances and so on, but it doesn't load um the discourse acts and. For example, not the. What is what is what there.<SPK_2>Hmm.<SPK_1>Not the summaries. On your notes, those on your mind. | 2 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0120444_0120583 | EN2001a | [{"row": 37, "id": "AMI_EN2001a_H04_MEO069_0120444_0120583", "speaker": "MEO069", "offset_s": 0.156, "dur_s": 1.39}, {"row": 263, "id": "AMI_EN2001a_H04_MEO069_0120583_0120926", "speaker": "MEO069", "offset_s": 1.546, "dur_s": 3.43}, {"row": 1309, "id": "AMI_EN2001a_H04_MEO069_0120926_0121370", "speaker": "MEO069", "of... | 0.247 | 21.533 | <SPK_1>Let's check that out. Um, I'll I'll probably ask Jonathan about it. So alternatively, if we realize that we can't do the whole thing in one go. We could probably just process some sort of metadata. You know what I mean? Like sort of, sort of for the whole series. Chunks representing the individual meetings. | 1 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0122764_0123754 | EN2001a | [{"row": 7, "id": "AMI_EN2001a_H04_MEO069_0122764_0123754", "speaker": "MEO069", "offset_s": 0.438, "dur_s": 9.9}, {"row": 977, "id": "AMI_EN2001a_H04_MEO069_0123754_0124222", "speaker": "MEO069", "offset_s": 10.338, "dur_s": 4.68}] | 0.405 | 15.422 | <SPK_1>Like something that represents the whole series. In in a in a structure very similar to the structure in which we represent individual um meetings. But with data sort of always, combined from the whole, series. | 1 | 2 |
EN2001a:AMI_EN2001a_H04_MEO069_0124222_0125153 | EN2001a | [{"row": 1491, "id": "AMI_EN2001a_H04_MEO069_0124222_0125153", "speaker": "MEO069", "offset_s": 0.5, "dur_s": 9.31}, {"row": 991, "id": "AMI_EN2001a_H04_MEO069_0125330_0125531", "speaker": "MEO069", "offset_s": 11.31, "dur_s": 2.01}, {"row": 1149, "id": "AMI_EN2001a_H01_FEO066_0125501_0126258", "speaker": "FEO066", "of... | 0.094 | 20.684 | <SPK_1>So instead of having a single utterance that we display, it probably be like that would be representing a whole topic, a segment in a meeting. And so, so that using the same data.<SPK_2>You mean that you um basically split up the big thing into um different summaries. | 2 | 3 |
EN2001a:AMI_EN2001a_H01_FEO066_0126258_0127308 | EN2001a | [{"row": 1171, "id": "AMI_EN2001a_H01_FEO066_0126258_0127308", "speaker": "FEO066", "offset_s": 0.363, "dur_s": 10.5}, {"row": 1389, "id": "AMI_EN2001a_H04_MEO069_0126264_0126333", "speaker": "MEO069", "offset_s": 0.423, "dur_s": 0.69}, {"row": 1499, "id": "AMI_EN2001a_H04_MEO069_0127218_0127804", "speaker": "MEO069", ... | 0.042 | 20.315 | <SPK_1>For example, that you have a very, um, top level, um, summary and a separate file, for for each level.<SPK_2>Well, in a sense. I'm thinking of in the sense of like creating a virtual a virtual meeting out of the whole meeting series sort of.<SPK_3>Ah, easy. We just like create a new XML document in memory. | 3 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0128172_0128897 | EN2001a | [{"row": 627, "id": "AMI_EN2001a_H04_MEO069_0128172_0128897", "speaker": "MEO069", "offset_s": 0.339, "dur_s": 7.25}, {"row": 1584, "id": "AMI_EN2001a_H04_MEO069_0128897_0130026", "speaker": "MEO069", "offset_s": 7.589, "dur_s": 11.29}, {"row": 1422, "id": "AMI_EN2001a_H04_MEO069_0130026_0130325", "speaker": "MEO069", ... | 0.201 | 27.64 | <SPK_1>Yeah, sort of like offline create a virtual meeting, which which basically treats the meeting serious as if it was a meeting. And treats the individual meetings within the series as if they were segments, and treats the individual segments within meetings as if they were utterances. You know, so we just we shift... | 2 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0130882_0131346 | EN2001a | [{"row": 875, "id": "AMI_EN2001a_H04_MEO069_0130882_0131346", "speaker": "MEO069", "offset_s": 0.359, "dur_s": 4.64}, {"row": 1125, "id": "AMI_EN2001a_H04_MEO069_0131346_0131976", "speaker": "MEO069", "offset_s": 4.999, "dur_s": 6.3}, {"row": 862, "id": "AMI_EN2001a_H04_MEO069_0131976_0132222", "speaker": "MEO069", "of... | 0.107 | 13.865 | <SPK_1>If you are on whole document, whole series level, and that was a double click. Then, don't just go into that segment, but load a new file or something like it. But in general, use the same algorithm. That would be an alternative if we can't actually load the whole thing and. | 1 | 3 |
EN2001a:AMI_EN2001a_H04_MEO069_0132480_0133236 | EN2001a | [{"row": 1074, "id": "AMI_EN2001a_H04_MEO069_0132480_0133236", "speaker": "MEO069", "offset_s": 0.264, "dur_s": 7.56}] | 0.411 | 8.235 | <SPK_1>Because also like even if we. Maybe this whole like maybe I'm worrying too much about the whole series in one thing this. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0133236_0133618 | EN2001a | [{"row": 285, "id": "AMI_EN2001a_H04_MEO069_0133236_0133618", "speaker": "MEO069", "offset_s": 0.325, "dur_s": 3.82}] | 0.059 | 4.204 | <SPK_1>Because actually, I mean, probably users wouldn't view that one too often. | 1 | 1 |
EN2001a:AMI_EN2001a_H03_MEE067_0133618_0134351 | EN2001a | [{"row": 1308, "id": "AMI_EN2001a_H03_MEE067_0133618_0134351", "speaker": "MEE067", "offset_s": 0.489, "dur_s": 7.33}, {"row": 1473, "id": "AMI_EN2001a_H03_MEE067_0134351_0134673", "speaker": "MEE067", "offset_s": 7.819, "dur_s": 3.22}, {"row": 1219, "id": "AMI_EN2001a_H03_MEE067_0134673_0135532", "speaker": "MEE067", ... | 0.14 | 25.439 | <SPK_1>I don't think it's really that much of a problem because if it's too big, what we can do is just, well all the offline stuff doesn't really matter. And all we can do is just process a bit at a time. Like for summarizations, say we wanted a hundred utterances in the summary. Just look at the meeting, take the top... | 1 | 4 |
EN2001a:AMI_EN2001a_H03_MEE067_0136099_0136309 | EN2001a | [{"row": 1180, "id": "AMI_EN2001a_H03_MEE067_0136099_0136309", "speaker": "MEE067", "offset_s": 0.198, "dur_s": 2.1}, {"row": 304, "id": "AMI_EN2001a_H04_MEO069_0136453_0136599", "speaker": "MEO069", "offset_s": 3.738, "dur_s": 1.46}] | 0.427 | 5.625 | <SPK_1>And then you only have to process one meeting at a time.<SPK_2>Yeah, but I'm I'm still bored. | 2 | 2 |
EN2001a:AMI_EN2001a_H04_MEO069_0136599_0137220 | EN2001a | [{"row": 1345, "id": "AMI_EN2001a_H04_MEO069_0136599_0137220", "speaker": "MEO069", "offset_s": 0.226, "dur_s": 6.21}] | 0.137 | 6.574 | <SPK_1>Like for example, for the display, if you actually if you want to display like for the whole series. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0137220_0137825 | EN2001a | [{"row": 1283, "id": "AMI_EN2001a_H04_MEO069_0137220_0137825", "speaker": "MEO069", "offset_s": 0.067, "dur_s": 6.05}, {"row": 179, "id": "AMI_EN2001a_H04_MEO069_0137825_0138410", "speaker": "MEO069", "offset_s": 6.117, "dur_s": 5.85}, {"row": 82, "id": "AMI_EN2001a_H03_MEE067_0138344_0139072", "speaker": "MEE067", "of... | 0.431 | 24.447 | <SPK_1>The information density levels, based on and the only granularity you have is individual utterance. That means you have to go through every single utterance in a series of seventy hours of meetings.<SPK_2>Okay, so maybe we should build a store a mean measure for the segments and meetings as well.<SPK_1>Yeah. Yea... | 2 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0139615_0139834 | EN2001a | [{"row": 1150, "id": "AMI_EN2001a_H04_MEO069_0139615_0139834", "speaker": "MEO069", "offset_s": 0.174, "dur_s": 2.19}, {"row": 83, "id": "AMI_EN2001a_H04_MEO069_0139834_0140209", "speaker": "MEO069", "offset_s": 2.363, "dur_s": 3.75}, {"row": 1460, "id": "AMI_EN2001a_H04_MEO069_0140544_0140993", "speaker": "MEO069", "o... | 0.325 | 15.379 | <SPK_1>Like the way how we store individual latrines and stuff. Then maybe we can more or less use the same code and just make a few ifs and stuff. So but still, so in in general, we having we having utterances. And they have a score. | 1 | 4 |
EN2001a:AMI_EN2001a_H04_MEO069_0141541_0141928 | EN2001a | [{"row": 1209, "id": "AMI_EN2001a_H04_MEO069_0141541_0141928", "speaker": "MEO069", "offset_s": 0.009, "dur_s": 3.87}, {"row": 385, "id": "AMI_EN2001a_H04_MEO069_0141928_0142419", "speaker": "MEO069", "offset_s": 3.879, "dur_s": 4.91}, {"row": 150, "id": "AMI_EN2001a_H03_MEE067_0142298_0142500", "speaker": "MEE067", "o... | 0.139 | 25.309 | <SPK_1>And that's as much as we really need. And and they also have a time, a time information of course.<SPK_2>And speaker.<SPK_1>Hmm. And a speaker information. Yeah.<SPK_2>Speaker and topic segmenting will need as well.<SPK_1>Yeah, so in information which topic they are in. Yeah.<SPK_2>Yeah.<SPK_1>And and probably s... | 2 | 11 |
EN2001a:AMI_EN2001a_H04_MEO069_0144239_0144900 | EN2001a | [{"row": 1077, "id": "AMI_EN2001a_H04_MEO069_0144239_0144900", "speaker": "MEO069", "offset_s": 0.279, "dur_s": 6.61}] | 0.34 | 7.23 | <SPK_1>So the skimming can work on that because the skimming just sort of sorts the utterances, and puts as many in as it needs. | 1 | 1 |
EN2001a:AMI_EN2001a_H03_MEE067_0144840_0145325 | EN2001a | [{"row": 594, "id": "AMI_EN2001a_H03_MEE067_0144840_0145325", "speaker": "MEE067", "offset_s": 0.455, "dur_s": 4.85}, {"row": 323, "id": "AMI_EN2001a_H04_MEO069_0145423_0145515", "speaker": "MEO069", "offset_s": 6.285, "dur_s": 0.92}, {"row": 3, "id": "AMI_EN2001a_H04_MEO069_0145515_0146152", "speaker": "MEO069", "offs... | 0.469 | 23.554 | <SPK_1>Well yeah and then, it'll preserve the order when it's displayed though.<SPK_2>Yeah. Yeah, it'll play them in some order in which they were set because otherwise it's gonna be more entertaining.<SPK_1>Yeah.<SPK_2>Um. But that that's enough data for the skimming, and the the searching. So what the searching does ... | 2 | 7 |
EN2001a:AMI_EN2001a_H04_MEO069_0147114_0148063 | EN2001a | [{"row": 986, "id": "AMI_EN2001a_H04_MEO069_0147114_0148063", "speaker": "MEO069", "offset_s": 0.351, "dur_s": 9.49}] | 0.375 | 10.215 | <SPK_1>Goes to the SQL database, and gets like basically in the end gets just a time marker for, whether it is like that utterance that we're concerned with. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0148063_0148671 | EN2001a | [{"row": 686, "id": "AMI_EN2001a_H04_MEO069_0148063_0148671", "speaker": "MEO069", "offset_s": 0.451, "dur_s": 6.08}, {"row": 779, "id": "AMI_EN2001a_H04_MEO069_0148671_0149139", "speaker": "MEO069", "offset_s": 6.531, "dur_s": 4.68}, {"row": 77, "id": "AMI_EN2001a_H00_MEE068_0149028_0149049", "speaker": "MEE068", "off... | 0.356 | 20.358 | <SPK_1>And then we have to find. I'm sure there's some way in in Night Owl to just say set position to that time mark. And then it shifts the whole frame, and it alerts every single element of the display, and the display updates.<SPK_2>Hmm. Yeah, do you not want it to do the little tree display as well for multiple re... | 2 | 7 |
EN2001a:AMI_EN2001a_H04_MEO069_0150018_0150263 | EN2001a | [{"row": 853, "id": "AMI_EN2001a_H04_MEO069_0150018_0150263", "speaker": "MEO069", "offset_s": 0.269, "dur_s": 2.45}, {"row": 426, "id": "AMI_EN2001a_H00_MEE068_0150352_0150490", "speaker": "MEE068", "offset_s": 3.609, "dur_s": 1.38}, {"row": 981, "id": "AMI_EN2001a_H04_MEO069_0150519_0150946", "speaker": "MEO069", "of... | 0.186 | 16.515 | <SPK_1>So, if in that tree display somebody clicks on something.<SPK_2>You just need to find the timestamp.<SPK_1>Yeah, and then you sort of feed the timestamp to an identity XML central manager. And that central manager alerts everything that's there, like alerts the skim, like the the audio display.<SPK_2>Yeah.<SPK_1... | 2 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0151624_0152326 | EN2001a | [{"row": 890, "id": "AMI_EN2001a_H04_MEO069_0151624_0152326", "speaker": "MEO069", "offset_s": 0.004, "dur_s": 7.02}] | 0.294 | 7.318 | <SPK_1>Alerts the visual display and says we have a new time frame, and then they all sort of do their update routines with respect to the current level of zoom. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0152326_0152938 | EN2001a | [{"row": 112, "id": "AMI_EN2001a_H04_MEO069_0152326_0152938", "speaker": "MEO069", "offset_s": 0.395, "dur_s": 6.12}] | 0.167 | 6.682 | <SPK_1>So how much to display, and starting position at where the, or maybe the mid position of it. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0152938_0153515 | EN2001a | [{"row": 1450, "id": "AMI_EN2001a_H04_MEO069_0152938_0153515", "speaker": "MEO069", "offset_s": 0.31, "dur_s": 5.77}, {"row": 115, "id": "AMI_EN2001a_H04_MEO069_0153515_0153569", "speaker": "MEO069", "offset_s": 6.08, "dur_s": 0.54}, {"row": 837, "id": "AMI_EN2001a_H04_MEO069_0153569_0153775", "speaker": "MEO069", "off... | 0.169 | 19.819 | <SPK_1>I don't know. Like, if you start where the thing was found or if the thing was found is in the middle of the part that we display. That I don't know. What that we can decide about, but in general sort of. It's the same thing if, like, whether you play and it moves forward or whether you jump to a position throug... | 2 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0154957_0155654 | EN2001a | [{"row": 1451, "id": "AMI_EN2001a_H04_MEO069_0154957_0155654", "speaker": "MEO069", "offset_s": 0.491, "dur_s": 6.97}] | 0.082 | 7.543 | <SPK_1>It's only that the event gets triggered by the search routine, which sort of pushes it into NightXMed and says, "Please go there now." | 1 | 1 |
EN2001a:AMI_EN2001a_H03_MEE067_0155756_0156186 | EN2001a | [{"row": 495, "id": "AMI_EN2001a_H03_MEE067_0155756_0156186", "speaker": "MEE067", "offset_s": 0.274, "dur_s": 4.3}, {"row": 156, "id": "AMI_EN2001a_H03_MEE067_0156392_0156585", "speaker": "MEE067", "offset_s": 6.074, "dur_s": 1.93}, {"row": 655, "id": "AMI_EN2001a_H03_MEE067_0156730_0157162", "speaker": "MEE067", "off... | 0.197 | 13.972 | <SPK_1>So we should basically make our own XML document in memory. That everyone's Module changes that, rather than the underlying data. | 1 | 3 |
EN2001a:AMI_EN2001a_H03_MEE067_0157162_0157650 | EN2001a | [{"row": 442, "id": "AMI_EN2001a_H03_MEE067_0157162_0157650", "speaker": "MEE067", "offset_s": 0.12, "dur_s": 4.88}, {"row": 437, "id": "AMI_EN2001a_H04_MEO069_0157742_0157937", "speaker": "MEO069", "offset_s": 5.92, "dur_s": 1.95}, {"row": 369, "id": "AMI_EN2001a_H04_MEO069_0157937_0158012", "speaker": "MEO069", "offs... | 0.239 | 13.759 | <SPK_1>And then have that XML document tied to the interface.<SPK_2>Why do we have to do memory? But that stuff.<SPK_1>Well, you can make it in a file if you want.<SPK_2>I mean, like the information is coming from offline. | 2 | 5 |
EN2001a:AMI_EN2001a_H04_MEO069_0158831_0159339 | EN2001a | [{"row": 665, "id": "AMI_EN2001a_H04_MEO069_0158831_0159339", "speaker": "MEO069", "offset_s": 0.262, "dur_s": 5.08}] | 0.09 | 5.431 | <SPK_1>So we probably we don't even have to change the utterance document. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0159339_0159765 | EN2001a | [{"row": 548, "id": "AMI_EN2001a_H04_MEO069_0159339_0159765", "speaker": "MEO069", "offset_s": 0.164, "dur_s": 4.26}] | 0.202 | 4.626 | <SPK_1>Right, because the whole way like the whole beauty of the NADex model is that it ties together lots of different files. | 1 | 1 |
EN2001a:AMI_EN2001a_H04_MEO069_0159765_0160373 | EN2001a | [{"row": 452, "id": "AMI_EN2001a_H04_MEO069_0159765_0160373", "speaker": "MEO069", "offset_s": 0.454, "dur_s": 6.08}, {"row": 297, "id": "AMI_EN2001a_H03_MEE067_0160340_0160372", "speaker": "MEE067", "offset_s": 6.204, "dur_s": 0.32}, {"row": 1334, "id": "AMI_EN2001a_H04_MEO069_0160373_0160722", "speaker": "MEO069", "o... | 0.05 | 16.574 | <SPK_1>So we can just create an additional XML file, which for every utterance, like the utterances have IDs, I presume, some references.<SPK_2>嗯。<SPK_1>So we just we tie just a very short XML file.<SPK_3>Yeah.<SPK_4>Yes.<SPK_1>Which it's the only information it has. It has, whatever a number for. For the weight. | 4 | 6 |
EN2001a:AMI_EN2001a_H04_MEO069_0161372_0161949 | EN2001a | [{"row": 117, "id": "AMI_EN2001a_H04_MEO069_0161372_0161949", "speaker": "MEO069", "offset_s": 0.488, "dur_s": 5.77}, {"row": 1411, "id": "AMI_EN2001a_H02_FEO065_0162071_0162292", "speaker": "FEO065", "offset_s": 7.478, "dur_s": 2.21}, {"row": 377, "id": "AMI_EN2001a_H02_FEO065_0162292_0162541", "speaker": "FEO065", "o... | 0.42 | 18.109 | <SPK_1>For the information density, and we just tie that to the existing utterances and tie them to the existing speaker changes.<SPK_2>But there is no idea for an attorney. I think. It's just for individual words. So how do we do that then?<SPK_3>No, no, it's for.<SPK_2>We, for our chances as well. | 3 | 6 |
EN2001a:AMI_EN2001a_H01_FEO066_0163256_0163329 | EN2001a | [{"row": 586, "id": "AMI_EN2001a_H01_FEO066_0163256_0163329", "speaker": "FEO066", "offset_s": 0.35, "dur_s": 0.73}, {"row": 198, "id": "AMI_EN2001a_H01_FEO066_0163329_0163411", "speaker": "FEO066", "offset_s": 1.08, "dur_s": 0.82}, {"row": 744, "id": "AMI_EN2001a_H02_FEO065_0163384_0163615", "speaker": "FEO065", "offs... | 0.24 | 11.539 | <SPK_1>No, you're right. Yeah.<SPK_2>I think it's just for one.<SPK_1>It's for.<SPK_2>Word.<SPK_3>Yeah, I think I think those segments for each utterance, they split up.<SPK_2>So we have to. Yeah.<SPK_3>Think so. | 3 | 9 |
EN2001a:AMI_EN2001a_H04_MEO069_0164681_0165696 | EN2001a | [{"row": 1639, "id": "AMI_EN2001a_H04_MEO069_0164681_0165696", "speaker": "MEO069", "offset_s": 0.367, "dur_s": 10.15}] | 0.151 | 10.668 | <SPK_1>Well, otherwise we'd probably have to go over it and like add some integer that we just increment from top to bottom, sort of to every other run as an as an ID. | 1 | 1 |
End of preview. Expand in Data Studio
Speaker-ASR window pool (AMI)
Window manifests for tiny-audio's speaker-attributed ASR recipe
(ta speaker-asr build-pool, configs/speaker_asr). Each row of
{split}.parquet lists the AMI utterances of one window (row index into
edinburghcstr/ami ihm/{split}-*.parquet, utterance id, speaker, offset, duration) and the
<SPK_n> target; transcripts-{split}.jsonl holds the base model's
transcript of each utterance; signatures/{split}.json the settings that
built the split. No audio is stored here.
Derived from the AMI Meeting Corpus (Carletta et al., 2005), CC BY 4.0: https://groups.inf.ed.ac.uk/ami/corpus/
- Downloads last month
- 53