Size: 908
Comment:
|
Size: 3042
Comment:
|
Deletions are marked like this. | Additions are marked like this. |
Line 1: | Line 1: |
= Install and run = | = Install and run (last tested 17-01-2018) = Dataset (version 22-12-2017): ''vulcano:/local/raid/ad/qleverUI/index'' QLeverUI code (version 22-12-2017): ''vulcano:/local/raid/ad/qleverUI/svn'' (also contains ''db.sqlite3'' from Django) QLever binary (version 10-01-2018): 'vulcano:/local/raid/ad/schnelle/QLever/build/ServerMain'' {{{ ssh vulcano cd /local/raid/ad/qleverUI/svn # SETUP (already done on vulcano) python manage.py migrate # Create initial db.sqlite3 for Django python manage.py createsuperuser # Create admin user, bast ist currently admin vim qlever/settings.py # Set DEBUG=False and set ALLOWED_HOSTS appropriately # START QLEVER UI python /local/raid/ad/qleverUI/manage.py runserver 0.0.0.0:8900 --insecure # START QLEVER /local/raid/ad/schnelle/QLever/build/ServerMain -i /local/raid/ad/qleverUI/index/wikipedia-freebase -p 9001 -t -l }}} This is what I last did on 17-01-2018 and it worked: {{{ |
Line 7: | Line 34: |
We can mimick the relation suggestions using standard SPARQL features. Here is an example: | We can mimick the suggestions using standard SPARQL features. Here are examples for each box (which can be tried via the [[http://metropolis.informatik.uni-freiburg.de:9000/sparql|Aqqu Virtuoso instance]]). == Instance suggestions == {{{ PREFIX fb: <http://rdf.freebase.com/ns/> SELECT ?person_name WHERE { ?person_id fb:people.person.profession ?profession_id . ?profession_id fb:type.object.name "Astronaut"@en . ?person_id fb:type.object.name ?person_name } GROUP BY ?person_id }}} With popularity scores (assuming an explicit predicate <popularity_score>): {{{ PREFIX fb: <http://rdf.freebase.com/ns/> SELECT DISTINCT ?person_name ?score WHERE { ?person_id fb:people.person.profession ?profession_id . ?profession_id fb:type.object.name "Astronaut"@en . ?person_id fb:type.object.name ?person_name . ?person_id <popularity_score> ?score } }}} == Class suggestions == They are redundant (fb:type.object.type + instance suggestions) == Relation suggestions == |
Line 22: | Line 82: |
== Word suggestions == There is (of course) nothing for word suggestions in SPARQL. But the QLever index can do just what the Broccoli index can do concerning word suggestions. One just needs to find a reasonable syntax to ask for it (and implement it). |
Install and run (last tested 17-01-2018)
Dataset (version 22-12-2017): vulcano:/local/raid/ad/qleverUI/index QLeverUI code (version 22-12-2017): vulcano:/local/raid/ad/qleverUI/svn (also contains db.sqlite3 from Django) QLever binary (version 10-01-2018): 'vulcano:/local/raid/ad/schnelle/QLever/build/ServerMain This is what I last did on 17-01-2018 and it worked: With popularity scores (assuming an explicit predicate <popularity_score>):
They are redundant (fb:type.object.type + instance suggestions)
Important note: for QLever, we will not be able to have queries with "?p ?o" in general. However, we could do something special for such queries (e.g. using a has-relations like Broccoli does) + we should use a syntax similar (or even identical) to that of SPARQL.
There is (of course) nothing for word suggestions in SPARQL. But the QLever index can do just what the Broccoli index can do concerning word suggestions. One just needs to find a reasonable syntax to ask for it (and implement it). ssh vulcano
cd /local/raid/ad/qleverUI/svn
# SETUP (already done on vulcano)
python manage.py migrate # Create initial db.sqlite3 for Django
python manage.py createsuperuser # Create admin user, bast ist currently admin
vim qlever/settings.py # Set DEBUG=False and set ALLOWED_HOSTS appropriately
# START QLEVER UI
python /local/raid/ad/qleverUI/manage.py runserver 0.0.0.0:8900 --insecure
# START QLEVER
/local/raid/ad/schnelle/QLever/build/ServerMain -i /local/raid/ad/qleverUI/index/wikipedia-freebase -p 9001 -t -l
TODO: describe how to start the service on nkaba, for now, see: https://docs.google.com/document/d/1gy3z3UkyicYaF75ZXsxTEzS8c28hPTFa6dQU6Y0dj0o/edit#
= Broccoli-like features =
We can mimick the suggestions using standard SPARQL features. Here are examples for each box (which can be tried via the [[http://metropolis.informatik.uni-freiburg.de:9000/sparql|Aqqu Virtuoso instance]]).
== Instance suggestions ==
{{{
PREFIX fb: <http://rdf.freebase.com/ns/>
SELECT ?person_name WHERE {
?person_id fb:people.person.profession ?profession_id .
?profession_id fb:type.object.name "Astronaut"@en .
?person_id fb:type.object.name ?person_name
}
GROUP BY ?person_id
PREFIX fb: <http://rdf.freebase.com/ns/>
SELECT DISTINCT ?person_name ?score WHERE {
?person_id fb:people.person.profession ?profession_id .
?profession_id fb:type.object.name "Astronaut"@en .
?person_id fb:type.object.name ?person_name .
?person_id <popularity_score> ?score
}
Class suggestions
Relation suggestions
PREFIX fb: <http://rdf.freebase.com/ns/>
SELECT ?p COUNT(DISTINCT ?person_name) WHERE {
?person_id fb:people.person.profession ?profession_id .
?profession_id fb:type.object.name "Astronaut"@en .
?person_id fb:type.object.name ?person_name .
?person_id ?p ?o
}
GROUP BY ?p
ORDER BY DESC(COUNT(DISTINCT ?person_name))
Word suggestions