A framework for lexical selection in natural language generation

Abstract

A~M~'g~ef~ This t~i~ dc:~cdbcs a proeedmc Ru' lexical se.. le=,:![~ia~ of OlW, n-claSS k:xicN items in a natmal langp.age gca~:~atkm sysRm. A,l optimum R;xical seleetiol; ua}t~nig ~mlst [i~.: able tO Ira|nice. rt~.!lizati,'m decisions mldc; vary-. i~g rcugextna~ cit'cmastances. First, it must be N)le to iy;;,?-,i~ ~, withont the illlhlCUCe of coiitext~ based ou |llCT, ll. lug cm~e~poadcuces I~elwcen elemeuls of couccptual ill" imt a,,d ~hu k:xic.~l i~e~'emo~y of thc talget language. S(m-O~ld, it must t~.~ ~,bh~ to use contextl~d constlaims, as sup° F;~f~ by coll~,.:adoaM iaf(~nuatio~ ii~ tfie gcneralion lexb c(m. '~ hi~d, ihcru lilllSt ])!) illt oplion O~ realizing input tepresenh liolts laonomiaally or th~ongh detinite dc.scripliuns. t,'il!;iii3, the~e nntst alsu b(3 au option of tLsing elliplical co~tstruciions. The ~tttum of 1)ackgtouud kuowledgc and th{; all o~ithu~ v/c sugge.~t lbr this task are. dcscrilav& The icxical xetcctiou prt~dttre iS a part of a couq)fehgnsive j~eltcra{ton Systeub D[OGI,2qE%

A new model of lexical choice for nouns1

“…Halliday (1966), Mel'Cuk and Polguere (1987), Nirenburg and Nirenburg (1988), Smadja and McKeown (1991), and many others have looked at collocation phenomena, such as the fact that the same wavelength of light is. lexicalized as "blond" when applied to hair but "yellow" when applied to tables.…”

Lexical Semantics and Knowledge Representation in Multilingual Text Generation

“…The scheme of word-concept linking that truly differs from d-nets is nearestneighbor classification [Nirenburg and Nirenburg 1988]. Clusters offeaturejvalue pairs are matched against each other to find the word that best expresses the desired content.…”

“…As discussed in Section 2.3, the DIOGENES system [Nirenburg and Nirenburg 1988] employs a 'meaning-matching metric' to determine the best possible lexeme and thereby differs from the standard discrimination net approach. Similar to MOOSE, the lexicon entries in DIOGENES, expressed in a frame language, pair a conceptual pattern with linguistic information-but again, the latter is of syntactic nature, and there is no distinction between a conceptual and a semantic level of representation.…”

“…(Section 2.4) Quite often, researchers have lamented that the problem of word choice has not received sufficient attention, e.g., [Marcus 1987;Nirenburg and Nirenburg 1988;McDonald 1991]-many language generators assume that for every concept in the input expression there is exactly one associated word. Yet, when lexicalization is indeed seen as a matter of choice, factors determining the differences between lexical items need to be found, and taking at least some of them into consideration can enhance the expressiveness of a generator considerably.…”

Resolving paraphrases to support modeling language perception in an intelligent agent

“…If one chooses the lower-level, narrower ontological concept to start with, such constraints may be inherent in its definition (as is the case with AERIAL-MOTION-EVENT). This preference is the inverse of the lexical choice in text generation off of text meaning representations (for details see Nirenburg and Nirenburg, 1988). OntoSem can yield either of the above basic text meaning representations.…”