Deployments
The OKBQA Hackathon 2014 has produced a set of prototype implementations for each of the modules envisioned in the architecture. Each module is accessible through RESTful services that are available for a limited time on the demonstration server.
Repository
(old) http://121.254.173.77:30000
(old) http://110.45.246.131:30000
Question Formulation
Template Generation
http://121.254.173.77:1555/templategeneration/templator/
curl -i -H "Content-Type: application/json" -X POST -d @test.json http://121.254.173.77:1555/templategeneration/templator/
test.json: { "string": "Which rivers flow through Busan?", "language":"en" }
test.json: { "string": "한라산은 얼마나 높은가?", "language":"ko" }
Korean Dependency Parser: Input String(Question) -> output String(Stanford Form)
http://143.248.135.187:10117/controller/service/etri_parser
Disambiguation (English)
1. http://14.63.169.59:9345/rest/disambiguate
2. http://121.254.173.77:2357/agdistis/disambiguate
Example call (see JSON content for test.json below):
curl -i -H "Content-Type: application/json" -X POST -d @test.json http://121.254.173.77:2357/agdistis/disambiguate
3. http://121.254.173.77:2357/agdistis/run?data={"question":"Who did that and wo?","slots" : [ {"s" : "?x", "p" : "verbalization", "o" : "shopping center"}, {"s" : "?x", "p" : "is", "o" : "rdf:Class_rdf:Resource"}, {"s" : "?y", "p" : "verbalization", "o" : "University of Kiel"} ] }
{
"question": "Who did that and wo?",
"ned": [
{
"score": 1,
"entities": [
{
"score": 1,
"var": "?x",
"value": "http://dbpedia.org/resource/Strip_mall"
},
{
"score": 1,
"var": "?y",
"value": "http://dbpedia.org/resource/University_of_Kiel"
}
],
"classes": [
{
"score": 1.2173913043478262,
"var": "?x",
"value": "http://dbpedia.org/ontology/ShoppingMall"
}
],
"literals": [],
"properties": []
}
]
}
Disambiguation(Korean)
http://121.254.173.77:12345/request
AGDISTIS for korean: URL: http://121.254.173.77:33211/AGDISTIS Example call: curl --data-urlencode "text='<entity>덕혜옹주</entity>는 <entity>조선</entity>의 제26대 <entity>왕</entity>이자 <entity>대한제국</entity>의 초대 <entity> 황제</entity> <entity>고종</entity>과 <entity>귀인 양씨</entity>의 <entity>황녀</entity>이다.'" -d type=agdistis http://121.254.173.77:33211/AGDISTIS
Query Generation
Sparqlator (GraphFinder)
URL: http://121.254.173.77:38401/queries
Example call: curl -H "content-type:application/json" -d @test.json http://121.254.173.77:38401/queries
Answer Generation
URL: http://121.254.173.77:10009/test
Example input JSON : {"sentence":"Which rivers flow through Gunsan?", "lang":"en"}
Rendering
http://14.63.169.59:2020/rest/render
More info: IO Specification.
Test Service
http://121.254.173.77:2358/test/
The test service understands both POST and GET requests. The query parameters for GET requests are the URL of the controller (parameter name: url) and optionally the language in which the tests are to be carried out (parameter name: language, default = en). A corresponding call would be
http://121.254.173.77:2358/test/get?url=http://test.org/controller
for a controller located at http://test.org/controller.
The POST version understands JSON objects of the form
{"url":"http:\/\/test.org\/controller", "language":"korean"}.
and is available at
http://121.254.173.77:2358/test/runtest
Test Service (for OKBQA 2015)
URL: http://121.254.173.77:56789/controller Example call: curl -H "Content-Type: application/json" -X POST -d @input.json http://121.254.173.77:56789/controller Input JSON: {"string":"question", "language":"en"/"ko"}
Distributional Semantics Services (English is online and Korean is on its way)
You can experiment with five distributional models: ESA, LSA, RI, Word2Vec and Glove.
The services can be invoked as below.
For returning the semantic relatedness measure of two terms (e.g. `power' and `electric'):
http://vmdgsit04.deri.ie/relatedness/esa/en?term=power&targetSet=electric
As a response you will get the semantic relatedness measure:
{"electric":0.051038838852184375}
We can also calculate the semantic relatedness value of one term (e.g. `power') and a list of terms () :
http://vmdgsit04.deri.ie/relatedness/esa/en?term=power&targetSet=electric;cable;station;car;table
In this case we will receive as an answer the ranked list of semantic relatedness values:
{"electric":0.051038838852184375,"station":0.019850553560107974,
"cable":0.00895209238610573,"table":0.0,"car":0.0}
The parameters of the URL are the distributional semantic model (esa, lsa, ri, w2v,
glove) and language (e.g. en).
You can also get the distributional vector of a word (e.g. love), which consists of the
id of the vector component and of the weight associated with that component:
http://vmdgsit04.deri.ie/vector/esa/en?term=love
which will return the answer:
{"27002780":0.15384358167648315,"7874937":0.18914856016635895,
"26342547":0.15788961946964264,"24738938":0.15290413796901703,...
}
Named Entity Recognition
Path Finder
Full-Text Index
Korean Entity Linking Service
Example call: curl -X POST -H "Content-Type:application/x-www-form-urlencoded" -d "sentence=대한민국의 지방법원의 수는 총 18개이다." http://143.248.135.187:22334/controller/service/entitylinking_okbqa
Korean Predicate Linking Service - 1 (considering predicate argument structure)
output: S postposition [tab] O postposition [tab] S ner [tab] O ner [tab] Property [tab] score
Example call: curl -X POST -H "Content-Type:application/x-www-form-urlencoded" -d "sentence=태어나" http://143.248.135.187:22334/controller/service/predlinking1_okbqa
Korean Predicate Linking Service - 2 (without considering predicate argument structure)
output: Property [tab] score
Example call: curl -X POST -H "Content-Type:application/x-www-form-urlencoded" -d "sentence=태어나" http://143.248.135.187:22334/controller/service/predlinking2_okbqa
Korean Templete Generation Service (to be updated)
Example call: curl -X POST -H "Content-Type:application/x-www-form-urlencoded" -d "sentence=대한민국의 대통령은 누구인가? " http://143.248.135.187:22334/controller/service/templete_okbqa
Korean Query Generation
Example call: curl -X POST -H "Content-Type:application/x-www-form-urlencoded" -d "sentence=대한민국의 대통령은 누구인가?" http://143.248.135.187:22334/controller/service/query_okbqa
Korean Word Embedding Similarity Service
Example call: curl -X POST -H "Content-Type:application/x-www-form-urlencoded" -d "sentence=대한민국" http://143.248.135.187:22334/controller/service/similar_words_okbqa
Korean DBpedia 2014 Endpoint
http://143.248.135.60:45002/sparql