1 00:00:00,950 --> 00:00:06,890 Hello, my name is Matthew Graham and this is Module Two of two modules on Semantics. 2 00:00:06,890 --> 00:00:09,260 In the last module we discussed 3 00:00:10,740 --> 00:00:13,660 how to represent knowledge in a machine processable way. 4 00:00:14,850 --> 00:00:18,830 We talked about the various different representations of RDF, 5 00:00:18,830 --> 00:00:21,450 the technology that can be used to do this. 6 00:00:21,450 --> 00:00:24,500 How to embed this information into a web page, and 7 00:00:24,500 --> 00:00:29,960 also the linked data idea that makes use of this 8 00:00:32,200 --> 00:00:37,370 to create a, a, web of linked information that's accessible through browsers. 9 00:00:38,670 --> 00:00:43,820 In this module we're going to be considering how to 10 00:00:43,820 --> 00:00:48,850 actually represent domain knowledge in terms of machine processable 11 00:00:48,850 --> 00:00:53,190 form through concept schemes and ontologies. 12 00:00:55,140 --> 00:00:59,340 So we have a hierarchy of concept schemes that can be used 13 00:00:59,340 --> 00:01:01,560 to capture domain knowledge. 14 00:01:01,560 --> 00:01:07,420 The idea is in its simplest form we have a controlled vocabulary. 15 00:01:07,420 --> 00:01:11,310 So this can just be a closed list of terms that can be used for classification. 16 00:01:11,310 --> 00:01:16,400 So it can just be a certain list of words or phrases that we 17 00:01:16,400 --> 00:01:22,447 are using for a set of labels, and we will always make use of that closed list. 18 00:01:22,447 --> 00:01:26,180 But there's no notion of any sort of relationship between them. 19 00:01:26,180 --> 00:01:30,680 There's no notion of a hierarchy of definition within them. 20 00:01:30,680 --> 00:01:33,990 So the next level up from a controlled vocabulary, 21 00:01:33,990 --> 00:01:38,510 which can just really regard as, as a bag of terms, is what's called a taxonomy and 22 00:01:38,510 --> 00:01:42,990 then a taxonomy, you have a controlled vocabulary but 23 00:01:42,990 --> 00:01:47,890 you have imposed a hierarchy on it in terms of term and, and sub term. 24 00:01:49,470 --> 00:01:51,590 So for example, eh, 25 00:01:51,590 --> 00:01:55,980 when we were talking about anatomical structure in the last lecture. 26 00:01:57,560 --> 00:02:00,940 We had this notion of looking up the word Hindbrain and 27 00:02:00,940 --> 00:02:05,610 then their maybe in our, taxonomy that we're using, a, 28 00:02:05,610 --> 00:02:09,830 a, a super term of that would be central nervous system and 29 00:02:09,830 --> 00:02:15,870 a sub term of that would be an individual structure that makes up hindbrain. 30 00:02:15,870 --> 00:02:20,540 So these hierarchies of, meaning don't necessarily you need to be 31 00:02:20,540 --> 00:02:25,220 only in terms of obvious meaning, 32 00:02:25,220 --> 00:02:30,690 they can also be in terms of other hierarchical notions whether it's 33 00:02:30,690 --> 00:02:35,760 anatomical development or geographical location or things like that. 34 00:02:37,640 --> 00:02:43,450 About the taxonomy we then have this more advanced concept of a thesaurus 35 00:02:44,640 --> 00:02:49,090 and in a thesaurus we have a taxonomy but 36 00:02:49,090 --> 00:02:54,580 there are then notions of not only hierarchical relationships 37 00:02:54,580 --> 00:02:59,580 between the individual terms, there are notions that you have broader and 38 00:02:59,580 --> 00:03:03,440 narrower terms, that you can have synonymous terms. 39 00:03:03,440 --> 00:03:05,820 So these are terms that you can use at a, 40 00:03:05,820 --> 00:03:08,960 at an equal level in a hierarchy instead of each other. 41 00:03:08,960 --> 00:03:13,360 There is some notion of a top term a, a root term. 42 00:03:13,360 --> 00:03:17,820 There is a notion of a scope note, to, to determine that, you know, this is when 43 00:03:17,820 --> 00:03:22,110 this particular term is relevant, this is when this particular term is relevant and 44 00:03:22,110 --> 00:03:25,390 then there are also notions of related terms. 45 00:03:25,390 --> 00:03:29,700 So, that this term is related to this term over in this part of the hierarchy, or 46 00:03:29,700 --> 00:03:32,329 this term is related to this term in this other thesaurus. 47 00:03:36,340 --> 00:03:39,770 Other ideas or concept schemes that you, you could use when you're, 48 00:03:39,770 --> 00:03:44,970 you're programatically manipulating knowledge and information, also things 49 00:03:44,970 --> 00:03:49,610 like subject heading list, or terminology, glossaries, faceted classifications. 50 00:03:52,410 --> 00:03:58,910 So the idea is that you would employ one of these when you are working with 51 00:04:00,520 --> 00:04:06,100 a knowledge base and, and trying to 52 00:04:06,100 --> 00:04:11,120 encapture a domain knowledge into a, a programmatic fashion. 53 00:04:13,790 --> 00:04:18,480 Now the way you would represent these sort of simple concept 54 00:04:18,480 --> 00:04:22,830 schemes is through yet another W3C standard. 55 00:04:24,274 --> 00:04:28,320 This one is known as SKOS which stands for the simple knowledge organization system. 56 00:04:31,878 --> 00:04:36,480 It's used for expressing knowledge concept schemes in a machine-understandable way, 57 00:04:36,480 --> 00:04:42,400 and it's, is itself expressed in terms of RDF 58 00:04:42,400 --> 00:04:48,050 statements and a thing called RDFS, which is a, a simple schema language for 59 00:04:48,050 --> 00:04:51,810 RDF, which can be used for defining RDF data structures. 60 00:04:51,810 --> 00:04:56,610 And the idea there is that you express your entire concept scheme 61 00:04:56,610 --> 00:04:59,130 as an RDF graph with content and structure. 62 00:04:59,130 --> 00:05:04,050 And there are various words there you can see which identify 63 00:05:04,050 --> 00:05:08,010 specific terms within, within the SKOS system 64 00:05:08,010 --> 00:05:12,300 relating to things that we've identified in the particular type of data structure, 65 00:05:12,300 --> 00:05:15,030 particular type of knowledge structure we may be interested in. 66 00:05:15,030 --> 00:05:18,840 So we have scopeNotes and narrower, broader, related, hiddenLabels. 67 00:05:18,840 --> 00:05:20,480 Whether something's an OrderedCollection, 68 00:05:20,480 --> 00:05:23,570 it's a ConceptScheme that sort of thing preferred labels. 69 00:05:23,570 --> 00:05:28,480 So the idea is that I could take one of my vocabularies or 70 00:05:28,480 --> 00:05:34,650 taxonomy that I"m using and I can then write it up formally in the form of a, 71 00:05:34,650 --> 00:05:37,130 a SKOS vocabulary, and 72 00:05:37,130 --> 00:05:40,870 then that there are programmatic tools which I can then use that to ingest. 73 00:05:40,870 --> 00:05:47,300 And I could then start programmatically manipulating RDF triples, 74 00:05:47,300 --> 00:05:52,070 which are expressed in terms of that SKOS vocabulary, 75 00:05:52,070 --> 00:05:56,240 such that they may be the subject or mainly the predicate 76 00:05:57,290 --> 00:06:02,590 are terms which are relevant or that are parts of these SKOS vocabularies. 77 00:06:04,250 --> 00:06:11,150 And here's an example of, of SKOS, of of of a skos vocabulary. 78 00:06:11,150 --> 00:06:15,100 This is actually defining a temperature scale, 79 00:06:17,770 --> 00:06:22,800 and it's saying that I'm listing a whole load of different temperature scales here. 80 00:06:23,980 --> 00:06:27,420 But the one I'm particularly describing in this example is the absolute 81 00:06:27,420 --> 00:06:28,740 temperature scale. 82 00:06:28,740 --> 00:06:31,760 I'm saying that this has a preferred label, which is abso, 83 00:06:31,760 --> 00:06:34,510 absolute temperature scale in lower case. 84 00:06:34,510 --> 00:06:38,010 The definition in my provider, even readable definition here. 85 00:06:38,010 --> 00:06:43,010 There's a broader term of it in terms of this concept which is temperature scales, 86 00:06:43,010 --> 00:06:46,710 there is a narrower term which is Kelvin, and 87 00:06:46,710 --> 00:06:52,330 then we have related, concepts in our scheme to this temperature, 88 00:06:52,330 --> 00:06:54,940 just because the absolute temperature scale which is the Celsius temperature 89 00:06:54,940 --> 00:06:59,100 scale, and another related concept is, is temperature itself. 90 00:06:59,100 --> 00:07:03,650 So we could define an entire skos vocabulary 91 00:07:03,650 --> 00:07:07,710 devoted to the physical properties of temperature and measuring temperature. 92 00:07:07,710 --> 00:07:10,590 Temperature scales, how that information is encoded, 93 00:07:10,590 --> 00:07:13,910 how those different systems are related to each other. 94 00:07:13,910 --> 00:07:17,760 And then we could have, a piece of code, 95 00:07:17,760 --> 00:07:21,770 which could use that to programatically manipulate, 96 00:07:21,770 --> 00:07:26,410 temperature readings for climate data models for example. 97 00:07:27,423 --> 00:07:30,930 There is, 98 00:07:30,930 --> 00:07:35,910 a query language for RDF called sparql. 99 00:07:35,910 --> 00:07:38,140 Again, it's a W3 standard. 100 00:07:38,140 --> 00:07:42,510 I'm not going to go into it in any great detail here, just to show you though that 101 00:07:42,510 --> 00:07:47,460 there are particular keywords in there which will look familiar to the sorts of 102 00:07:47,460 --> 00:07:52,340 keywords that we covered when, in the two modules we did on SQL in the databases. 103 00:07:52,340 --> 00:07:57,130 So you will see select and you will see aware and order by, 104 00:07:57,130 --> 00:08:00,870 distinct, limit, offset, that sort of thing. 105 00:08:00,870 --> 00:08:03,260 This is an example of a Sparql query. 106 00:08:06,510 --> 00:08:12,000 In sparql variables are identified by question marks at the start of it. 107 00:08:12,000 --> 00:08:18,190 What we're doing here is here is we are, querying a database of elements, 108 00:08:18,190 --> 00:08:22,190 and we're asking for the, the name, the chemical symbol, the weight, 109 00:08:22,190 --> 00:08:26,850 the atomic weight, the atomic number, and the color from our 110 00:08:26,850 --> 00:08:31,630 data collection of linked data or whatever expressed in a machine-readable way. 111 00:08:31,630 --> 00:08:36,470 And then we're saying where we have to think we're looking specifically for 112 00:08:36,470 --> 00:08:37,380 uranium. 113 00:08:37,380 --> 00:08:42,820 And this is the way we would identify the things which are, 114 00:08:42,820 --> 00:08:47,920 are transuranic and heavier than uranium and then we order by, them by weight. 115 00:08:47,920 --> 00:08:50,650 So this is an example of this sort of thing. 116 00:08:50,650 --> 00:08:53,750 We could write this as a similar example if this data was expressed in 117 00:08:55,590 --> 00:08:58,540 in a relational database form with the SQL statement. 118 00:08:58,540 --> 00:09:02,620 The advantage of sparql however is that the idea is that the data doesn't 119 00:09:02,620 --> 00:09:06,940 necessarily reside in a single database. 120 00:09:06,940 --> 00:09:09,970 It already has this notion that the data is all linked together out 121 00:09:09,970 --> 00:09:12,450 there somewhere, so there's already a notion when you're, 122 00:09:12,450 --> 00:09:16,300 you're constructing a sparkle query that it's going to be a distributed query 123 00:09:16,300 --> 00:09:21,210 hitting multiple data sources where the data is identifiable and, and resided. 124 00:09:21,210 --> 00:09:24,810 And that's given by those URIs in the RDF statements that you're working with. 125 00:09:27,350 --> 00:09:31,340 Now, the most elaborate, or 126 00:09:31,340 --> 00:09:36,900 the most powerful concept scheme that you can use, is a thing called an ontology. 127 00:09:38,590 --> 00:09:42,390 And the formal definition of ontology is, 128 00:09:42,390 --> 00:09:45,660 is that a formal specification of a shared conceptualization. 129 00:09:45,660 --> 00:09:52,540 What that means is that you have an agreed 130 00:09:52,540 --> 00:09:58,040 set of ideas of what the domain knowledge base for a particular de, subdomain or 131 00:09:58,040 --> 00:10:02,340 domain is, the particular knowledge base for that is, and that's how you 132 00:10:02,340 --> 00:10:06,530 are expressing it in a machine processable way and that's called an ontology. 133 00:10:06,530 --> 00:10:10,180 So it is essentially a data model that represents a set of concepts within 134 00:10:10,180 --> 00:10:13,610 the domain and all the relationships between those concepts. 135 00:10:14,712 --> 00:10:18,824 An ontology allows you to define arbitrary relationships, 136 00:10:18,824 --> 00:10:23,580 arbitrary properties, arbitrary concepts so you're not limited to 137 00:10:23,580 --> 00:10:29,220 the types of the relationships you can have to just hierarchical or whatever. 138 00:10:29,220 --> 00:10:31,330 You can have any type of relationship and 139 00:10:31,330 --> 00:10:34,940 you can identify those and, and encode those. 140 00:10:34,940 --> 00:10:39,840 So ontology is generally describe individuals which are the ground 141 00:10:39,840 --> 00:10:43,930 level objects, the, the individual facts that you may be trying to represent. 142 00:10:43,930 --> 00:10:48,940 They will describe classes and if you're familiar with object oriented 143 00:10:48,940 --> 00:10:52,250 programming then, then that should make some degree of sense for you to have. 144 00:10:52,250 --> 00:10:54,830 You have classes and instances in object oriented programming and 145 00:10:54,830 --> 00:10:57,210 in ontology you have individuals in classes. 146 00:10:57,210 --> 00:11:00,780 You have attributes that the classes and the individuals have, properties, 147 00:11:00,780 --> 00:11:04,030 features, characteristics, those sort of things and then you have relationships 148 00:11:04,030 --> 00:11:09,020 which you you use to associate all of those together. 149 00:11:10,120 --> 00:11:14,350 And then there can be events which define how those attributes or 150 00:11:14,350 --> 00:11:17,120 relationships can change in pre-described ways. 151 00:11:20,470 --> 00:11:24,790 Yet again, there is a W3C standard for authoring ontologists for 152 00:11:24,790 --> 00:11:26,030 representing ontologists. 153 00:11:26,030 --> 00:11:29,470 This is the web ontology language or OWL, as its called. 154 00:11:31,140 --> 00:11:34,878 It's again, based on RDF so 155 00:11:34,878 --> 00:11:39,720 an ontology is expressed in terms of RDF structures. 156 00:11:39,720 --> 00:11:43,380 And this is largely regarded as one of the fundamental technologies underpinning 157 00:11:43,380 --> 00:11:48,313 the semantic web because it gives you the richness of expression to offer 158 00:11:48,313 --> 00:11:54,380 arbitrarily encode any set of concepts, 159 00:11:54,380 --> 00:11:58,060 any piece of domain knowledge into a machine processible fashion. 160 00:11:59,650 --> 00:12:04,090 As I said OWL allows descriptions of relations between classes 161 00:12:04,090 --> 00:12:06,750 particularly disjointedness and that sort of thing. 162 00:12:06,750 --> 00:12:10,170 It allows you to express cardinality constraints on particular things, so 163 00:12:10,170 --> 00:12:11,590 you have exactly one. 164 00:12:11,590 --> 00:12:16,078 Characteristic of properties and numerator classes, and all sorts of, of, 165 00:12:16,078 --> 00:12:17,440 very powerful things. 166 00:12:18,740 --> 00:12:24,489 So, owl regards data as being interpreted as a set of individuals which it describes 167 00:12:25,490 --> 00:12:30,000 a set of property assertions relating the individuals to each other and 168 00:12:30,000 --> 00:12:34,270 a set of axioms placing constraints on, on sets of individuals, 169 00:12:34,270 --> 00:12:38,660 on, on classes and the types of relationships allowed between them. 170 00:12:38,660 --> 00:12:44,730 So, for example, we may have a family ontology which is defining this notion of 171 00:12:45,840 --> 00:12:50,810 what a family is and the concepts that that contains that may have for, example, 172 00:12:50,810 --> 00:12:56,320 a, a hasMother relation which is only 173 00:12:56,320 --> 00:13:01,580 present between two individuals when one of them, has a parent, is also present. 174 00:13:03,810 --> 00:13:05,840 And in our notion of family ontology, 175 00:13:05,840 --> 00:13:10,170 we may also have this relationship called HasTypeOBlood. 176 00:13:12,190 --> 00:13:16,790 And members of the class of that relationship, 177 00:13:16,790 --> 00:13:21,510 HasTypeOBlood, are never related via the hasParent 178 00:13:21,510 --> 00:13:26,640 constraint two members of HasTypeABBlood. 179 00:13:26,640 --> 00:13:31,810 So in that particular example, we can put biological constraints on our notion 180 00:13:31,810 --> 00:13:36,210 of family into this knowledge base and have that as a machine processable form. 181 00:13:37,310 --> 00:13:41,250 So it's showing that these are not just simple relationships, there are, 182 00:13:41,250 --> 00:13:43,680 there are ways we can encode any 183 00:13:43,680 --> 00:13:46,100 sort of information we want into something like that. 184 00:13:48,030 --> 00:13:53,650 We can also, make inferences, or say that if Ada has mother Ann, 185 00:13:53,650 --> 00:13:58,640 and Ann is, has type O blood, then Ann is not a member of 186 00:13:58,640 --> 00:14:03,370 the HasTypeABlood relationship. 187 00:14:03,370 --> 00:14:07,530 And one of the powers of things that are represented through ontology is 188 00:14:07,530 --> 00:14:13,580 particularly is that we can use logical inferencing to identify statements 189 00:14:13,580 --> 00:14:17,466 that are not consistent with the knowledge base that we're, we're using. 190 00:14:17,466 --> 00:14:23,330 This is done but using reasoning surfaces, services. 191 00:14:23,330 --> 00:14:29,767 This is done using reasoning services. 192 00:14:31,010 --> 00:14:34,480 And, what's called first order predicate logic and, 193 00:14:34,480 --> 00:14:39,920 the idea is because we've expressed our, our, our knowledge base, in a for most, 194 00:14:39,920 --> 00:14:46,010 and rigorous way then we can use mathematical logic to either make 195 00:14:47,730 --> 00:14:52,070 further inferences or look for statements that are in congruence with each other. 196 00:14:52,070 --> 00:14:54,730 So we can say that sheep only eat grass. 197 00:14:54,730 --> 00:14:57,060 We can say that grass is a plant. 198 00:14:57,060 --> 00:15:02,050 We can identify plants and parts of plants as disjoint from animals and 199 00:15:02,050 --> 00:15:03,670 parts of animals. 200 00:15:03,670 --> 00:15:07,480 We can say vegetarians only eat things which are not animals or 201 00:15:07,480 --> 00:15:13,230 parts of animals and then having encoded those statements up we can ask 202 00:15:13,230 --> 00:15:18,590 our logical inferencing action to make some inferences and 203 00:15:18,590 --> 00:15:22,120 it can come up with that statement that sheep are vegetarians. 204 00:15:22,120 --> 00:15:27,600 Which is a logically, consistent statement based on the, the set of 205 00:15:27,600 --> 00:15:30,970 statements we've already given it, a set of instances we've already given it, 206 00:15:30,970 --> 00:15:36,140 plus the relationship rules, and the properties we've defined between those. 207 00:15:36,140 --> 00:15:40,540 So that gives a little toy example of the kind of inferencing that you can do. 208 00:15:44,920 --> 00:15:48,750 Some of different things examples on one of our smart applications might use. 209 00:15:48,750 --> 00:15:52,877 If you remember the, the set of zebra fish, meta data, 210 00:15:52,877 --> 00:15:57,700 multimedias data that we talked about as an example in, in the last module. 211 00:15:59,120 --> 00:16:02,960 Then we may have some inferences that we're going to run, 212 00:16:02,960 --> 00:16:07,620 we have an ontology behind it, we have our data marked up in terms of that ontology, 213 00:16:07,620 --> 00:16:10,608 and so we may want to look for and 214 00:16:10,608 --> 00:16:15,120 atomical terms which are relevent for a particular stage of development. 215 00:16:15,120 --> 00:16:20,170 Or we may want to search for terms which are related to a particular term 216 00:16:20,170 --> 00:16:22,560 in some way, from some class whatever. 217 00:16:22,560 --> 00:16:27,620 Or we may want to find, two classes, and we're looking for 218 00:16:27,620 --> 00:16:32,070 the lowest common ancestor in the hierarchy, because we 219 00:16:32,070 --> 00:16:36,130 want to use that as a single search term which would then encompass both terms, in, 220 00:16:36,130 --> 00:16:38,770 in a hierarchical fashion that we're interested in doing. 221 00:16:38,770 --> 00:16:42,430 And we can use logical inferencing for all finding all those sort of problems. 222 00:16:45,430 --> 00:16:51,250 Finally, software that you may want to look at for working with ontologies 223 00:16:51,250 --> 00:16:57,090 analysis, there's Jena and Protege- I would highly recommend looking at Protege. 224 00:16:57,090 --> 00:17:01,570 It allows you to explore ontologies, it allows you construct your own ontologies, 225 00:17:01,570 --> 00:17:04,060 it allows you to reason over ontologies. 226 00:17:04,060 --> 00:17:06,230 It's, it's a very powerful tool and 227 00:17:06,230 --> 00:17:10,548 then there's a, a, an instance of an infrontention here which is come at, 228 00:17:10,548 --> 00:17:15,960 at the bottom, CWM, which you may also want to have a look at. 229 00:17:15,960 --> 00:17:21,820 There are other like pellet and hermit which have I think 230 00:17:21,820 --> 00:17:26,460 both Java and Python wrappers that you can use in, in your code base as well 231 00:17:26,460 --> 00:17:29,950 if you are working with RDF and ontological information. 232 00:17:29,950 --> 00:17:35,361 And that is the end of the module. 233 00:17:35,361 --> 00:17:35,861 Thank you.