refs/heads/escape-regex - KorAP/Koral

commit	87b2ff515606bf6fcf21454024d1f2c440b00a8c	[log] [tgz]
author	Akron <nils@diewald-online.de>	Tue Mar 15 21:29:25 2016 +0100
committer	Akron <nils@diewald-online.de>	Tue Mar 15 21:29:25 2016 +0100
tree	e5a248aca2c862d60624337e5e8588b43b4be924
parent	c41303d4a0a51f4c9803f5a7b771d23619f0fe97 [diff]

tree: e5a248aca2c862d60624337e5e8588b43b4be924

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.