Input data

The Grew command line interface can be used with two different ways to describe the input linguistic data:

• the Mono mode where one corpus is considered
• the Multi mode where a set of corpora is considered

The table below shows which are the modes compatible with each subcommand.

	transform	grep	count	compile	clean
Mono	✅	✅	✅ (🆕 in 1.10)	❌	❌
Multi	❌	✅ (🆕 in 1.10)	✅	✅	✅

The Multi mode is also use in Grew-match to describe the set of corpora on which one can request.

The Mono mode

The Mono mode corresponds the following arguments on the command line:

• a sequence of arguments -i <file>.<ext> with extension .conll, .conllu, .cupt or .orfeo → all sentences in the different files are loaded, following the CoNNL-U format or CoNLL-U Plus format.
• a sequence of arguments -i <file>.json with JSON files following Graph JSON encoding → all sentences in the different JSON files
• a sequence of arguments -i <file>.amr or -i <file>.txt → all graphs in the different files are loaded, following the PENMAN notation
• one argument -i directory → load all files in the directory like in one of the3 items above
• no -i argument → CoNNL-u data is read on stdin

The Multi mode

The Multi mode is used when the command line argument contains a sequence of arguments -i <file>.json with JSON files following the JSON description of a set of corpora below.

JSON description of a set of corpora

Set of corpora are used both for the Grew-match online tool and for Grew.

A JSON file encodes a list of corpus. Each corpus is described by:

• a unique identifier id
• a directory where the files of the corpus are stored (use absolute paths)
• a files field with a list of file names. This field is optional, by default all files with extension conll, conllu, cupt or orfeo are loaded.

For instance, the file en_fr_zh.json 🔗 describes 3 corpora from UD 2.15 (of course, directories should be modified to match your local installation).

[
  { 
    "id": "UD_English-EWT@2.15",
    "directory": "/Users/guillaum/resources/ud-treebanks-v2.15/UD_English-EWT/",
    "files": ["en_ewt-ud-dev.conllu", "en_ewt-ud-test.conllu", "en_ewt-ud-train.conllu"]
  },
  { 
    "id": "UD_French-Sequoia@2.15",
    "directory": "/Users/guillaum/resources/ud-treebanks-v2.15/UD_French-Sequoia/"
  },
  { 
    "id": "UD_Chinese-GSD@2.15",
    "directory": "/Users/guillaum/resources/ud-treebanks-v2.15/UD_Chinese-GSD/"
  }
]

NB: A few other fields are used for the description of corpora used in the Grew-match. See here for examples of the JSON files used in different instances of Grew-match.