refs/heads/dupColl - KorAP/Koral

commit	eeea33520d7a11345d2457baef1a80794733a18b	[log] [tgz]
author	Michael Hanl <micha.hanl@gmail.com>	Thu Jun 02 17:26:19 2016 +0200
committer	Michael Hanl <micha.hanl@gmail.com>	Thu Jun 02 17:26:19 2016 +0200
tree	7b475dacaff8c231b30276ecb7eeb74dc793c489
parent	15311e0827e613dee28bba8b6c7d365a5ed891cd [diff]

tree: 7b475dacaff8c231b30276ecb7eeb74dc793c489

README.md

Koral

Koral is a library designed for the translation of different corpus query languages to KoralQuery, a JSON-LD-based protocol for the common representation of linguistic queries. This work has been carried out within the KorAP project (see below) and forms the major part of a Master thesis that is due to appear. The detailed specifications of KoralQuery will be covered in that thesis.

As of v0.1, the following corpus query languages (QLs) are supported:

Cosmas-II QL
ANNIS QL
Poliqarp QL (extended by numerous operators to "PoliqarpPlus" QL)
CQL

You can use the main class QuerySerializer to translate and serialize queries for you. The usage example below illustrates this. Valid QL identifiers are cosmas2, annis, poliqarp, poliqarpplus and cql.

Usage Example

import de.ids_mannheim.korap.query.serialize.QuerySerialzer;
QuerySerializer qs = new QuerySerializer();
String query = "contains(<s>,[orth=zu][pos=ADJA])";
qs.setQuery(query, "poliqarpplus");
System.out.println(qs.toJSON());

This will print out the following JSON-LD string for the Koralized query. The query asks for a sentence element (<s>) that is contained in a sequence of the surface form zu and a token with the part-of-speech tag ADJA. In the KoralQuery string, a containment relation is defined over two operands, an s span and a sequence of two tokens.

{
  "@context": "http://korap.ids-mannheim.de/ns/KoralQuery/v0.2/context.jsonld",
  "query": {
    "@type": "koral:group",
    "operation": "operation:position",
    "frames": [
      "frames:isAround"
    ],
    "operands": [
      {
        "@type": "koral:span",
        "key": "s"
      },
      {
        "@type": "koral:group",
        "operation": "operation:sequence",
        "operands": [
          {
            "@type": "koral:token",
            "wrap": {
              "@type": "koral:term",
              "layer": "orth",
              "key": "zu",
              "match": "match:eq"
            }
          },
          {
            "@type": "koral:token",
            "wrap": {
              "@type": "koral:term",
              "layer": "pos",
              "key": "ADJA",
              "match": "match:eq"
            }
          }
        ]
      }
    ]
  }
}

Motivation

Koral enables the design and implementation of corpus query systems independently of any specific query languages. All the system needs to do on the query processing side is have the query translated to KoralQuery (see usage) and feed the translated query to its search engine. In particular, several query languages can be supported without further adjustments to the search engine.

Koral and KoralQuery have been designed and developed within the KorAP Project, and are used in KorAP to translate queries to a common format before sending them to the backend.

Installation

Installation is straightforward (Maven3 required):

git clone https://github.com/korap/Koral [install-dir]
cd [install-dir]
mvn test
mvn install

There is also a command line version. After installation, simply run

java -jar target/Koral-0.2.jar [query] [queryLanguage]

Authorship

Koral and KoralQuery were developed by Joachim Bingel, Nils Diewald, Michael Hanl and Eliza Margaretha at IDS Mannheim.

The ANTLR grammars for parsing ANNIS QL and COSMAS II QL were developed by Thomas Krause (HU Berlin) and Franck Bodmer (IDS Mannheim), respectively. Minor adaptations of those grammars were implemented by the Koral authors.

The authors wish to thank Piotr Bański, Franck Bodmer, Elena Frick and Carsten Schnober for their valuable input.

License

Koral is published under the BSD-2 License. See also the attached LICENSE.

The ANNIS grammar is licensed under the Apache License 2.0.