Lexical Realisation Unit

Revision as of 14:03, 13 January 2010

In the UNLarium framework, a Lexical Realisation Unit (or simply LRU) is any discrete, recurring and standardized unit of meaning of a given natural language. It can be a subword (a root, an affix), a simple word or a multiword expression (compounds, collocations, idioms).

LRUs are lexical realisations for concepts

The UNLarium is first and foremost a generation-driven framework, which has been developed mainly to provide resources for generating natural language texts out of UNL graphs. In that sense, UNLarium entries should correspond to the most likely realisations, in a given language, of a given concept. The expression “realisation" stands here for a mixture of wording and phrasing, i.e., the manner in which a concept is articulated in a given language. For instance, the concept “the natural satellite of the Earth” is realised, in English, by the word “moon”; in French, by “lune”; in German, by “Mond”; in Russian, by “луна”; in Spanish, by “luna”; in Chinese, by 月; and so on. Your first task in the UNLarium is exactly to find out linguistic realisations for concepts, which will be always presented by their corresponding definition in English.

LRUs, however, are not simply linguistic realisations; they are lexical realisations. This means that LRUs should correspond to the units of the vocabulary of a language, i.e., to a "lexical item". Let’s come back to our previous example. Apart from “moon”, the concept “the natural satellite of the Earth” can be realised, in English, by the very expression “the natural satellite of the Earth”, which is indeed very frequent (2.130.000 occurrences in Google). This expression, however, is a “definition” rather than a “lexical realisation” for the concept, and should therefore not correspond to a LRU.

The differences between definitions and lexical items, or between “defining” and “naming” a concept, are fairly subjective, and are normally ascribed to the compositionality (or analyticity) of the candidate term: if the meaning of the compound can be reduced to the combination of the meaning of its components, it is said to be simply a definition; otherwise, i.e., if there is a sort of semantic surplus, a supplementary (or even complementary) sense added to the simple combination, the term is considered a lexical item. The above-mentioned expression "the natural satellite of the Earth", for instance, does not bring any semantic content other than the ones conveyed by its components. This is not the case of "geostationary communications satellite", which includes the idea of "orbit" which is not explicitly present in the compound. Accordingly, "geostationary communications satellite" (208.000 occurrences in Google) should be treated as a LRU, whereas "the natural satellite of the Earth", in spite of its higher frequency, should not.

Lexicalisation divergences

As languages have different lexicalisation processes, a single definition may correspond to several different LRUs, which are said to be synonyms. The definition “pass from physical life and lose all bodily attributes and functions necessary to sustain life”, for instance, may be realised in English by several different LRUs: “die”, “croak”, “decease”, “drop dead”, “buy the farm”, “cash in one's chips”, “give-up the ghost”, “kick the bucket”, “pass away”, “perish”, “snuff it”, “pop off”, “expire”, “conk”, “exit”, “choke”, “go”, “pass”, etc. In such cases, all realisations should be informed in the UNLarium.

There are cases, however, in which the definition cannot be lexically realised in the target language. This happens in two situations:

• When the concept is underspecified, i.e., too broad (or vague) to be realised. The concept of “red entity”, for instance, may be coextensive with several different English LRUs (“blood”, “cherry”, “ruby”, “ketchup”, “Spiderman”, etc), but these are rather subordinate terms (or hyponyms), in the sense they only include and partly match the intended sense. As the expression “red entity” itself is too compositional and too occasional to be considered already lexicalized (Google brings only 8,040 occurrences for this bigram), there will no LRU in this case.
• When the concept is overspecified, i.e., too narrow (or specific) to be realised. Consider, for instance, the definition “a person who is ready to forgive any transgression a first time and then to tolerate it for a second time, but never for a third time”. This definition does not lead to any LRU in English, French or Russian, even though it corresponds to a single word (“ilunga”) in Tshiluba, a language spoken in the Republic of Congo. We may obviously express the concept in any language, but we have to do it through a periphrasis (as we have done for English) or through a superordinate term (or hypernym), such as “forgiver”, “excuser”, “pardoner”, which are again fairly accurate.

In both cases, there will be no realisation to be informed, and it is important to indicate, in the UNLarium, that the concept has not been lexicalized yet, which means that it can be expressed in the target language only by means of definitions (periphrases) and other semantically related (and inaccurate) LRUs (such as hyponyms or hypernyms).

Types of LRU

In the UNLarium framework, there can be three different types of LRUs, depending on their internal structure:

Subwords (SBW) are LRUs that do not have independent existence in the language and only appear together with other morphemes to form a lexeme. They are common in synthetic (inflected) languages, such as the Indo-European ones, where some concepts (such as "contrary of" and "plural") are normally realised by bound morphemes ("un-" and "-s", for instance). Subword LRUs include affixes (such as "un-", "re-", "-ful", "-ness") and roots that do not occur alone (such as "rupt" in "interrupt", "disrupt", "corrupt", "rupture", etc).

Simple Words (WRD) are the ordinary (indivisible) lexemes in the semantic system of a language. They may consist of:

• one single free morpheme (such as "happy", "break");
• one single free morpheme and bound morphemes ("unhappy", "happiness", "happily", "unbreakable", "unbreakableness"); and
• compounds of bound morphemes (such as "interrupt", "disrupt", "corrupt").

Multiword Expressions (MTW) are lexical structures made up of a sequence of two or more lexemes. They can be concatenated ("darkroom", "skinhead") or isolated by hyphens ("blue-green", "Afro-American") or blank spaces ("round table", "part of speech"). Multiword expressions can be continuous ("get over") or discontinuous ("get <something> together"). They correspond to compounds ("fireman", "hardware"), phrases ("in spite of", "take into account"), idioms ("kick the bucket", "play cat and mouse"), fragments of sentences ("and so on", "whatever the case") or sentences ("Every evil is followed by some good", "No flies enter a mouth that is shut").

Classical compounds ("agriculture", "photograph") and their derivations ("agricultural", "photographically") are to be treated as simple words if they do not include more than one free morpheme. Phrasal verbs ("give in", "come across") are treated as multiword expressions.

How to express a LRU

In the UNLarium, LRUs are represented by a canonical form, to be obtained as follows:

• Subwords are to be expressed by their corresponding lemma without any special sign. Allomorphs (such as "es" for plural) will be generated by special rules and should not be represented:
  • s (and not "-s")
  • un (and not "un-")
  • chrono (and not "chrono-")
• Simple words are to be expressed by the lemma of the corresponding lexeme
  • happy
  • unhappy
  • unhappiness
  • table
  • foot
  • clothes
  • love
  • be
• Multiword expressions are to be expressed as continuous sequences of required lexemes, which should be represented by their corresponding lemma, if variable, or by their word form, otherwise:
  • get over
  • take into account (and not "take something into account", or "take <something> into account", or "take ... into account", etc)
  • throw to the lions (and not "throw to the lion")

Examples