Elas4RDF-search

Performs keyword search over RDF data, with classic IR techniques, upon triple-based documents using Elasticsearch (ES).

This project initializes a REST API for exploiting the indexes created here. System's response contains a ranked list of both a) triples and b) entities.

Requires Java 8 (or later) and a running Elasticsearch (elasticsearch-6.8) instance.

Install

Build project and package with mvn package. The generated .war file inside the Elas4RDF-search/target folder can be used for deploying the service in any server (e.g. apache-tomcat).

When deployed, application expects an application.properties file with options about the ES running instance: elastic.address=[string] & elastic.port=[int]. If file is not found, options default to localhost:9200.

Search service

1) Initializing an index

Application can serve different indexes (collections) by preserving a state (non-persistent) that corresponds to a given configuration.

Initialize an index through a POST /datasets method with request body a .json file with the following syntax:

  {
     "id": <dataset_identifier>,
     "index.name": <ES_index_name>,
     "index.fields": {
       <field_1> : <field_boost_1>,
       <field_2> : <field_boost_2>,
        ...
     }
   }

On success, response contains a confirmation .JSON message.

Note, this .json file should be automatically created after completing the index process from here. An example can be found in src/resources/examples/ folder.

2) Performing a Query

Queries are expressed through the GET method while requests accept either a high-level or a low-level syntax.

• High-Level syntax -> GET /high-level

  required id=[string] query=[string]

  optional size=[int] offset=[int] type=[string]

  id is the dataset identifier created in (1) and query can include any free-text keywords.

  size corresponds to the number of the returned triples, offset supports pagination (only for triples) and type corresponds to the answer return type: "triples", "entities" or "both" (default).

• Low-Level syntax ->GET /low-level

  required body=[json] index[string]

  optional size=[int] type[string]

  body is used for expressing more complicated queries through the use of ES Query DSL. Parameter index correponds to the ES index name.

  e.g. a multi-match type query

  body = 
  {
   "query": {
     "multi_match" : {
       "query":    "beatles abbey road", 
       "fields": ["subjectKeywords", "objectKeywords^2", "rdfs_comment_sub"],
       "type" : "cross_fields"
     }
   }
  }
  

example: using the curl command a (high-level syntax) request can be expressed as:

`curl --header "Content-Type: application/json" --request GET '<host>:<port>/elas4rdf_rest/high-level/?id=dataset_id&query=the%20beatles'`