Longind

Longind is a term used in corpus linguistics and natural language processing to denote marked long-distance dependencies between non-adjacent elements within a sentence. In annotation schemes and parse representations, a longind marker identifies a syntactic link whose elements are separated by one or more intervening words or clauses, such as wh-movement or topicalization phenomena.

Origin and usage: The term is a contraction of long-distance indicator and is not universally standardized; it appears in a few niche corpora and parsing tools as an optional annotation. It is intended to facilitate algorithms that must preserve or exploit nonlocal relations during parsing or coreference resolution.

Representation: In practice, longind can be encoded as a property on a dependency edge or as a separate attribute attached to the dependent, with values indicating distance or the type of nonlocal relation. Some schemes use a binary flag (longind present/absent) while others encode the number of intervening nodes or the specific relation type.

Examples and relevance: In English, a sentence like “Who did the editor praise the author?” contains a long-distance dependency between “Who” and the main verb. Marking longind helps parsers recover the correct antecedent and relation. Longind is related to, but distinct from, general notions of nonlocal dependencies, wh-movement, and coreference.

Limitations: The label is not common outside specialized annotation projects, and inconsistent use can hinder cross-corpus comparison. Researchers emphasize clear documentation of what constitutes a longind relation in a given dataset.

See also: long-distance dependency, syntactic parsing, dependency grammar, corpus annotation, wh-movement, coreference.