1 00:00:00,005 --> 00:00:10,975 >> And >> [INAUDIBLE] 2 00:00:10,975 --> 00:00:13,040 >> Yeah. >> [INAUDIBLE] 3 00:00:13,040 --> 00:00:16,288 >> It cannot replace. You, this is one way of evaluating 4 00:00:16,288 --> 00:00:20,900 pattern queries, right? >> You can do that. 5 00:00:20,900 --> 00:00:22,774 >> [INAUDIBLE] >> Clearly. 6 00:00:22,774 --> 00:00:26,538 Yeah, yeah. I mean, rarely is the case that you have 7 00:00:26,538 --> 00:00:30,960 a free lucnh, right? so, you cannot have the cake and eat it 8 00:00:30,960 --> 00:00:35,445 too, is exemplified from the fact that in many settings, your click constructions 9 00:00:35,445 --> 00:00:41,794 are not good enough. Linked open data for example the answer 10 00:00:41,794 --> 00:00:44,396 can be one. I can, I mean we can discuss this later 11 00:00:44,396 --> 00:00:47,336 on. And then the kind of queries you can get, 12 00:00:47,336 --> 00:00:51,484 are not making use of the clique structure, or the community structure as 13 00:00:51,484 --> 00:00:59,165 you can see. We, we can work this out, I mean, I'll 14 00:00:59,165 --> 00:01:07,752 show you some example >> [INAUDIBLE] Yeah. 15 00:01:07,752 --> 00:01:07,770 [COUGH] >> [INAUDIBLE] 16 00:01:07,770 --> 00:01:10,510 >> No, no, here you are looking for some cliques, all right? 17 00:01:10,510 --> 00:01:16,741 It does not mean that, every community need not be a clique. 18 00:01:16,741 --> 00:01:24,398 >> [INAUDIBLE] >> One of them, all right? 19 00:01:24,398 --> 00:01:26,201 >> [INAUDIBLE] >> Sure. 20 00:01:26,201 --> 00:01:32,505 >> [INAUDIBLE] >> Perfectly right. 21 00:01:32,505 --> 00:01:35,690 >> [INAUDIBLE] >> Exactly. 22 00:01:35,690 --> 00:01:39,956 But if your query workload >> Consists not of click, but of some 23 00:01:39,956 --> 00:01:44,802 of the triangles. Then your clicks are no longer really 24 00:01:44,802 --> 00:01:45,539 useful. >> [INAUDIBLE] 25 00:01:45,539 --> 00:01:47,312 >> Yeah. >> [INAUDIBLE] 26 00:01:47,312 --> 00:01:51,776 >> So, since you don't know what kind of query workload you're expecting, you 27 00:01:51,776 --> 00:01:56,633 need to come up with something that works in general. 28 00:01:56,633 --> 00:02:01,897 That's really the challenge. Okay? 29 00:02:01,897 --> 00:02:06,382 So now, these graph problems as we had studied in our high school and undergrad 30 00:02:06,382 --> 00:02:11,883 was largely without addition of labels on these edges and notes. 31 00:02:11,883 --> 00:02:16,707 So clearly SPARQL has the additional constraint of groundings and constraints 32 00:02:16,707 --> 00:02:22,602 that you may have on the limits. >> When we are talking about the LOB, 33 00:02:22,602 --> 00:02:30,522 how do you [UNKNOWN] but how do you realistically capture the whole thing? 34 00:02:30,522 --> 00:02:31,662 >> You cannot. >> You cannot? 35 00:02:31,662 --> 00:02:33,480 >> By I mean. >> Keep it updated? 36 00:02:33,480 --> 00:02:37,278 >> Exactly, by definition you cannot. So that's why, you have to go back to the 37 00:02:37,278 --> 00:02:40,995 good old federated query processing solution, which again, database guys 38 00:02:40,995 --> 00:02:44,790 worked on earlier. But now with added twists that. 39 00:02:44,790 --> 00:02:49,160 So, in fact, there is a lot going on, on that front as well. 40 00:02:49,160 --> 00:02:53,940 So you have this big LOD, nobody wants to give the data to you. 41 00:02:53,940 --> 00:02:58,060 Everybody wants to maintain their data, to later keep the quality. 42 00:02:58,060 --> 00:03:01,756 They will only expose a sparke interface to you. 43 00:03:01,756 --> 00:03:04,790 >> [INAUDIBLE] >> Exactly, they can have anything. 44 00:03:04,790 --> 00:03:08,328 Now given a query how do you decompose in such a way that each one can run 45 00:03:08,328 --> 00:03:12,654 separately and then put them together? >> Maybe you can [INAUDIBLE] 46 00:03:12,654 --> 00:03:14,996 [INAUDIBLE] >> Yeah. 47 00:03:14,996 --> 00:03:21,757 >> [INAUDIBLE] But the any database [INAUDIBLE] 48 00:03:21,757 --> 00:03:25,080 >> Yes. I mean, for example, music prints. 49 00:03:25,080 --> 00:03:29,160 Doesn't allow the database to be, because it's a very valuable data source. 50 00:03:29,160 --> 00:03:31,862 Why would they want to part with it? They make money on it. 51 00:03:31,862 --> 00:03:32,366 >> Okay. so. 52 00:03:32,366 --> 00:03:37,314 >> [INAUDIBLE] open this. >> Open this, because they are low 53 00:03:37,314 --> 00:03:43,628 SPARQL endpoints. You can access the data without paying 54 00:03:43,628 --> 00:03:46,440 any extra fees. That's not an issue. 55 00:03:46,440 --> 00:03:48,214 They want. >> it's probably. 56 00:03:48,214 --> 00:03:52,018 >> Yeah. But they have restrictions on what kind 57 00:03:52,018 --> 00:03:54,600 of SPARQL cannot. So in order to put. 58 00:03:54,600 --> 00:03:54,998 >> Yeah. [INAUDIBLE]. 59 00:03:54,998 --> 00:03:56,281 >> [INAUDIBLE] [CROSSTALK]. >> Exactly, you cannot just say, 60 00:03:56,281 --> 00:03:58,950 question mark, question mark, question mark. 61 00:03:58,950 --> 00:04:02,070 Just give everything. >> I try my fields, work, fields, 62 00:04:02,070 --> 00:04:06,610 texts, plurals. >> Yeah, yeah. 63 00:04:06,610 --> 00:04:10,390 I mean, so there are some conditions under which you can be part of [UNKNOWN] 64 00:04:10,390 --> 00:04:14,480 Consortium, but these conditions are relatively loose. 65 00:04:14,480 --> 00:04:19,380 They allow for maintaining your IT, all right, I mean these are. 66 00:04:19,380 --> 00:04:22,894 >> It's not so open anymore. >> This is the closest to open you can 67 00:04:22,894 --> 00:04:24,218 get. >> Oh. 68 00:04:24,218 --> 00:04:26,612 >> Okay [LAUGH]. So as open as you can get. 69 00:04:26,612 --> 00:04:28,090 >> As open as you can get. >> Yeah. 70 00:04:28,090 --> 00:04:30,782 You have to balance money versus freedom, okay? 71 00:04:30,782 --> 00:04:34,286 >> Of course [INAUDIBLE]. >> [INAUDIBLE]. 72 00:04:34,286 --> 00:04:37,289 >> But, but we know, 73 00:04:37,289 --> 00:04:39,550 [INAUDIBLE]. 74 00:04:39,550 --> 00:04:40,906 >> And more is [INAUDIBLE]? >> Yes. 75 00:04:40,906 --> 00:04:44,522 >> And that is not small? >> That's not small. 76 00:04:44,522 --> 00:04:47,121 >> That's not small? >> Yeah, yeah. 77 00:04:47,121 --> 00:04:53,786 At least for research it's not small. For Google it's trivial but for research 78 00:04:53,786 --> 00:04:59,811 it's a failry huge data set, right? >> [INAUDIBLE] 79 00:04:59,811 --> 00:05:03,770 Okay, thanks a lot. [SOUND]