Training:Advanced Ontology Engineering at FOI - 2011/Reengineering - Revision history

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T20:14:52Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T20:14:21Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T20:09:29Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Using D2R */

2011-10-04T20:02:35Z

‎Using D2R

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T20:01:35Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T19:59:19Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T19:55:55Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T18:57:39Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T09:22:14Z

‎Reengineering and Refactoring using Semion

EvaBlomqvist: /* Reengineering and Refactoring using Semion */

2011-10-04T09:10:18Z

‎Reengineering and Refactoring using Semion

← Older revision		Revision as of 20:14, 4 October 2011
Line 22:		Line 22:
	'''The Semion rule language syntax:'''		'''The Semion rule language syntax:'''

−	The complete Semion rule syntax can be found [http://www.ontologydesignpatterns.org/ont/training/foi/~~musicindustry~~.~~owl~~ here], however you don't need it for the first part of this exercise.	+	The complete Semion rule syntax can be found [http://www.ontologydesignpatterns.org/ont/training/foi/SemionRules.pdf here], however you don't need it for the first part of this exercise.

← Older revision		Revision as of 20:14, 4 October 2011
Line 21:		Line 21:

	'''The Semion rule language syntax:'''		'''The Semion rule language syntax:'''
−	The complete Semion rule syntax can be found [~~[Training~~:~~Advanced_Ontology_Engineering_at_FOI_-_2011~~/~~Reengineering~~/~~Syntax \|~~ here]].	+
		+	The complete Semion rule syntax can be found [http://www.ontologydesignpatterns.org/ont/training/foi/musicindustry.owl here], however you don't need it for the first part of this exercise.

← Older revision		Revision as of 20:09, 4 October 2011
Line 19:		Line 19:

	Content: The database contains information about a past ISWC conference, and in one table called "persons" there is information about the people that organized or attended the conference.		Content: The database contains information about a past ISWC conference, and in one table called "persons" there is information about the people that organized or attended the conference.
		+
		+	'''The Semion rule language syntax:'''
		+	The complete Semion rule syntax can be found [[Training:Advanced_Ontology_Engineering_at_FOI_-_2011/Reengineering/Syntax \| here]].

← Older revision		Revision as of 20:02, 4 October 2011
Line 82:		Line 82:

	The example data is already loaded into the database, but you have to edit the example mapping to reflect the details of the database above. Then you can run the D2R server (with the mapping file in the same directory) and try accessing the DB through SPARQL in your browser.		The example data is already loaded into the database, but you have to edit the example mapping to reflect the details of the database above. Then you can run the D2R server (with the mapping file in the same directory) and try accessing the DB through SPARQL in your browser.
		+
		+	How do the results compare to the ones we got with Semion?

@@ Line 66: / Line 66: @@
 * Wait until the processing has finished and then go to the SPARQL tab of the alignment. Run the default query that gets all the triples generated by the alignment.
 * Now you should see a set of triples representing a set of instances of foaf:Person, each with a foaf:firstName and foaf:lastName - this is our refactored dataset!
 === Using D2R ===

@@ Line 55: / Line 55: @@
 * Give a name to the rule, e.g. foaf_person. Then copy the following text as the content of the rule and click Finish:
 ** has(dbs:isRecordOf, ?person, dbSchema:persons_table) . has(dbs:isDatumOf, ?y, ?person) . has(dbs:bindsToColumn, ?y, dbSchema:persons_FirstName-column) . values(dbs:hasContent, ?y, ?fname) . has(dbs:isDatumOf, ?z, ?person) . has(dbs:bindsToColumn, ?z, dbSchema:persons_LastName-column) . values(dbs:hasContent, ?z, ?lname) -> is(foaf:Person, ?person) . values(foaf:firstName, ?person, ?fname) . values(foaf:lastName, ?person, ?lname)
-** This may seems like a very complex rule, but lets break it down to see what it means. The condition consists of three parts:
+** This may seems like a very complex rule, but lets break it down to see what it means. "has" is the keyword for object property assertions, "is" indicates a rdf:type relation, and "values" indicates a data property assertion. The condition then consists of three parts:
 *** has(dbs:isRecordOf, ?person, dbSchema:persons_table) - this matches triples that says that something (the variable ?person) is a record of the table person of the reengineered database, i.e. ?person is a row in the persons table.
 *** has(dbs:isDatumOf, ?y, ?person) . has(dbs:bindsToColumn, ?y, dbSchema:persons_FirstName-column) . values(dbs:hasContent, ?y, ?fname) - this gets the the entry in that row that is the first name, and extracts the value into the variable ?fname

@@ Line 53: / Line 53: @@
 * Go back to the tab of the alignment, and select the Alignment Rules tab.
 * At the top you see the list of rules currently inserted in this alignment (at the moment it is empty). Click on the Add-button below the table.
-* Give a name to the rule, e.g. foaf_person. Then copy the following text as the content of the rule:
+* Give a name to the rule, e.g. foaf_person. Then copy the following text as the content of the rule and click Finish:
-**
+** has(dbs:isRecordOf, ?person, dbSchema:persons_table) . has(dbs:isDatumOf, ?y, ?person) . has(dbs:bindsToColumn, ?y, dbSchema:persons_FirstName-column) . values(dbs:hasContent, ?y, ?fname) . has(dbs:isDatumOf, ?z, ?person) . has(dbs:bindsToColumn, ?z, dbSchema:persons_LastName-column) . values(dbs:hasContent, ?z, ?lname) -> is(foaf:Person, ?person) . values(foaf:firstName, ?person, ?fname) . values(foaf:lastName, ?person, ?lname)
 === Using D2R ===

@@ Line 48: / Line 48: @@
 * In the Project Explorer, under your datasource, and the Alignment subcategory, you alignment should now appear. Double click on it to open its related tab.
 * In the overview you can see any annotations related to the ontology you have chosen to align to. And in the SPARQL tab you can query your alignment, however, currently it is empty. Instead go to the Alignment Rules tab. This is where you can now create your alignment, i.e. the transformation rules that brings the  data from the RDF model of your database into something expressed using the FOAF model instead.
-* We now want to identify all the persons in the database and set them as instances of FOAF person. If you go back to the tab showing the database, and execute the example SPARQL query in the SPARQL tab, you can see all the RDF triples that were generated. If you scroll down the list you will find the actual data of the persons table, expressed as record instances. For instance, for each entry in the persons table there are two triples looking like this: dbSchema:persons_record_x rdf:type dbs:Record and dbSchema:persons_record_x dbs:isRecordOf dbSchema:persons_table. You should write rules that fetch each such persons_record_x and takes the data it has associated, i.e. through the dbSchema:persons_record_x_datum_x dbs:isDatumOf dbSchema:persons_record_x, and extracts its values, e.g. the first and last name of each person as properties of the new FOAF person you are creating.
+* We now want to identify all the persons in the database and set them as instances of FOAF person.
-* The following rule will
+** If you go back to the tab showing the database, and execute the example SPARQL query in the SPARQL tab, you can see all the RDF triples that were generated. If you scroll down the list you will find the actual data of the persons table, expressed as record instances. For instance, for each entry in the persons table there are two triples looking like this: dbSchema:persons_record_x rdf:type dbs:Record and dbSchema:persons_record_x dbs:isRecordOf dbSchema:persons_table.
 === Using D2R ===

@@ Line 44: / Line 44: @@
 * However, so far we have only transformed the source syntactically, so now we want to align the data to the FOAF ontology. Start by right-clicking on the data source in the Project Explorer, and select New, Alignment.
 * In the dialogue that appears, you have to first give the alignment a name. Then you can either select to simply use the OWL vocabulary itself as the basis for your alignment, or you can choose to add your own ontology. We want to align our data to the FOAF vocabulary, hence click on the "new ontology" button.
-* First you give the ontology and ID and a name (both can be anything you choose, but the ID needs to be unique for the workspace). Then provide the online URI of the ontology, i.e for FOAF you input http://xmlns.com/foaf/0.1/
+* First you give the ontology and ID and a name (both can be anything you choose, but the ID needs to be unique for the workspace). Then provide the online URI of the ontology, i.e for FOAF you input http://xmlns.com/foaf/0.1/ Then click Finish.
 Then click Finish.
 * Now the ontology appears in the drop-down list under the name you gave it. Select it and then click on Finish.
 * In the Project Explorer, under your datasource, and the Alignment subcategory, you alignment should now appear. Double click on it to open its related tab.
 * In the overview you can see any annotations related to the ontology you have chosen to align to. And in the SPARQL tab you can query your alignment, however, currently it is empty. Instead go to the Alignment Rules tab. This is where you can now create your alignment, i.e. the transformation rules that brings the  data from the RDF model of your database into something expressed using the FOAF model instead.
-* First we want to identify all the persons in the database and set them as instances of FOAF person.
+* We now want to identify all the persons in the database and set them as instances of FOAF person. If you go back to the tab showing the database, and execute the example SPARQL query in the SPARQL tab, you can see all the RDF triples that were generated. If you scroll down the list you will find the actual data of the persons table, expressed as record instances. For instance, for each entry in the persons table there are two triples looking like this: dbSchema:persons_record_x rdf:type dbs:Record and dbSchema:persons_record_x dbs:isRecordOf dbSchema:persons_table. You should write rules that fetch each such persons_record_x and takes the data it has associated, i.e. through the dbSchema:persons_record_x_datum_x dbs:isDatumOf dbSchema:persons_record_x, and extracts its values, e.g. the first and last name of each person as properties of the new FOAF person you are creating.
 === Using D2R ===

@@ Line 28: / Line 28: @@
 Start Semion from the commandline as follows (from the directory where you unpacked Semion):
 java -jar -Xmx512m -XstartOnFirstThread it.cnr.istc.semion.tool-0.6-SNAPSHOT.one-jar.jar
 When the tool starts, the following procedure lets you do the first syntactical transformation (reengineering):
@@ Line 44: / Line 49: @@
 * In the Project Explorer, under your datasource, and the Alignment subcategory, you alignment should now appear. Double click on it to open its related tab.
 * In the overview you can see any annotations related to the ontology you have chosen to align to. And in the SPARQL tab you can query your alignment, however, currently it is empty. Instead go to the Alignment Rules tab. This is where you can now create your alignment, i.e. the transformation rules that brings the  data from the RDF model of your database into something expressed using the FOAF model instead.
 * First we want to identify all the persons in the database and set them as instances of FOAF person.
 === Using D2R ===