Polaris Import Specifications

Inside a import record, there can be described:

• Concepts (word meanings and word instances)
• Variants
• Internal links between concepts
• ILI equivalence links
• Properties and property values

The general structure of a word meaning (noun) import record is:

0 WORD_MEANING
  1 PART_OF_SPEECH "n"
  1 VARIANTS
     # further details on the variants
  1 INTERNAL_LINKS
     # further details on the internal links
  1 EQ_LINKS
     # further details on the equivalence links
  1 PROPERTIES
     # further details on the properties

The general structure of a word instance (proper noun) import record is:

0 WORD_INSTANCE
  1 PART_OF_SPEECH "pn"
  1 VARIANTS
     # further details on the variants
  1 INTERNAL_LINKS
     # further details on the internal links
  1 EQ_LINKS
     # further details on the equivalence links
  1 PROPERTY_VALUES
     # further details on the property values

Concepts

Level 0 of a concept record is WORD_MEANING or WORD_INSTANCE. The first child field is PART_OF_SPEECH field. For WORD_INSTANCE, the part of speech is proper noun ("pn").

Concept Variants

The VARIANTS field is the parent of one or more subtrees, that are headed by LITERAL fields. The value for a LITERAL field describes the literal string for the variant. The first required child field for all LITERAL fields is the SENSE field, which provides the sense number.

The general structure of a VARIANTS subtree:

1 VARIANTS
  2 LITERAL "test"
    3 SENSE 1
    # further literal details
  2 LITERAL "experiment"
    3 SENSE 1
    # further literal details

Internal Links

The INTERNAL_LINKS field heads up the subtree of fields that describes all the links this concept has with other concepts in the Language WordNet. Each individual internal link is headed by a RELATION field. The value for this field is the name of the internal link.

The RELATION field has a subtree headed by a TARGET_CONCEPT field, which identifies the link’s target concept. It can also have an optional subtree headed by a FEATURES field, which add feature information to the link.

The whole thing fits together like this:

1 INTERNAL_LINKS
  2 RELATION "synonym"
    3 TARGET_CONCEPT
      # fields to describe the target go here
    3 FEATURES
      # fields to describe features go here

A target concept is described by specifying three things: a part-of-speech, a literal, and a sense number. With this combination it is possible to uniquely address any concept.

Interlingual Equivalence Links

The EQ_LINKS field heads up the subtree of fields that describes all the links this concept has with ILI records in the Interlingua. Each equivalence relation is headed by a EQ_RELATION field, the value for which is the name of the relation. The TARGET_ILI field is a child under each EQ_RELATION field. It identifies which ILI record is the target of the equivalence link.

Identifying a target ILI record can be done in the following ways:

• the combination of a part-of-speech, literal and sense number. Note that the literal will be looked up in the ILI variants of the site’s language.
• a combination of part-of-speech and a WordNet 1.5 synset file offset value.
• by specifying the ILI record’s database number.

Properties

In a WORD_MEANING record, it is possible to specify a list of properties. The properties specified must already have been created as property types.

The example shows that specifying properties is just a matter of listing them after each other:

1 PROPERTIES
  2 NAME "population"
  2 NAME "capital"
  2 NAME "government"

Property Values

In a WORD_INSTANCE record, it is possible to assign values to the properties specified in one of the WORD_MEANING records that appears in the hyperonym hierarchy of the instance.

The example shows how this can be done:

1 PROPERTY_VALUES
  2 NAME "population"
    3 VALUE_AS_INTEGER 11500000
  2 NAME "capital"
    3 VALUE_AS_TEXT "Jenil'ag"
  2 NAME "government"
    3 VALUE_AS_WORD_MEANING
      4 PART_OF_SPEECH "n"
      4 LITERAL "mulberian"
        5 SENSE 1