Loading...
Please wait, while we are loading the content...
Similar Documents
Modeling the Interpretation and Interpretation Ease of Noun-Noun Compounds Using a Relation Space Approach to Compound Meaning
| Content Provider | Semantic Scholar |
|---|---|
| Author | Costello, Fintan Devereux, B. J. |
| Copyright Year | 2006 |
| Abstract | Modelling the Interpretation and Interpretation Ease of Noun-Noun Compounds Using a Relation Space Approach to Compound Meaning Barry Devereux (Barry.Devereux@ucd.ie) Fintan Costello (Fintan.Costello@ucd.ie) School of Computer Science and Informatics, University College Dublin Dublin, Ireland Abstract process, such as how the correct relationship between the two constituent concepts is found or constructed. In this paper, our aim is to present evidence for a more compre- hensive approach to conceptual combination, allowing us to model both the interpretation and interpretation ease aspects of noun-noun comprehension. Conceptual combination can be regarded as a process which instantiates the most plausible or most appropri- ate relationship between the two constituent words in a compound (termed the modifier word and the head word, respectively). An important issue therefore for any model of conceptual combination is the manner in which the relationship between the modifier and head of a compound is represented; indeed, previous models of conceptual combination can be classified as belonging to two types, distinguished by how they represent relations. The first type, the concept specialization approach, as- sumes that instantiating a relation for a compound in- volves modifying a slot in the representation of the head word concept (for example, see Smith, Osherson, Rips & Keane, 1988). In the second type, the relationship be- tween the two nouns is specified by means of a taxonomy of general relation categories. For example Levi (1978) describes a set of recoverably deletable predicates such as CAUSE, HAVE & FROM which are used to spec- ify the meaning of compounds. The idea that the re- lationship between the constituents in a compound can be specified by a taxonomy of semantic primitives forms the basis for representing compound meaning in an im- portant cognitive theory about conceptual combination, namely the Competition Among Relations In Nominals (CARIN) model (Gagn´e & Shoben, 1997). The concept specialization approach and the taxo- nomic approach both assume that the meaning of a com- pound can be adequately captured by a simple label (ei- ther as a slot in the head concept or as a stand-alone relation category). One of the primary theses of this pa- per is that such a simple representation of compounds is inadequate; the relations instantiated during concep- tual combination are semantically detailed entities and as such require a more complex mode of representation. Our approach assumes that relations are as complex and as semantically non-trivial as the constituent con- cepts that they link are. We therefore represent relations in a way similar to how concepts have been represented in the classification literature (e.g. Nosofsky, 1984; Kr- uschke, 1992), using exemplars which are defined as sets of values on a set of dimensions. We generate relation exemplars using a corpus study where a large, represen- In this paper, we present a computational model of conceptual combination that introduces a new repre- sentation for the meaning of compounds: the relations used to interpret compounds are represented as points or vectors in a high-dimensional relation space. Such a representational framework has many advantages over other approaches. Firstly, the high-dimensionality of the space provides a detailed description of the compound's meaning; each of the space's dimensions represents a semantically distinct way in which com- pound meanings can differ from each other. Secondly, the spatial representation allows for a distance metric to measure how similar of different pairs of compound meanings are to each other. We conducted a corpus study, generating vectors in this relation space rep- resenting the meanings of a large, representative set of familiar compounds. A computational model of compound interpretation that uses these vectors as a database from which to derive new relation vectors for new compounds is presented. Also presented is a model of interpretation ease: that is, the ease or rapidity with which people can comprehend compounds. Our model uses ideas from the CARIN theory of conceptual combination about the modifier noun's role in the comprehension process; the model correlates as well as the traditional CARIN model with people's reaction times. Keywords: Conceptual combination; noun-noun compounds; mathematical modelling; CARIN. Introduction Conceptual combination, the process that people employ when interpreting novel noun-noun compounds such as volcano science, gas crisis or penguin movie, is a non- trivial cognitive task, often requiring people to access complex knowledge about the two constituent concepts and about the world in general. For example, people can quickly and efficiently determine that the compound penguin movie refers to a movie about penguins, and not a movie by penguins (which would be the correct way to interpret the compound penguin journey), nor a movie for penguins (the correct way to interpret the compound penguin enclosure), nor any of the infinitely many other possible but implausible ways of interpret- ing that compound. Perhaps because of the complex- ity of the phenomenon, previous theories of conceptual combination have tended to focus on only some aspects of conceptual combination. For example, in Gagn´e and Shoben's (1997) CARIN model, the focus is on mod- elling the ease and rapidity with which people interpret noun-noun compounds (as measured by reaction time), but not other features of the conceptual combination |
| File Format | PDF HTM / HTML |
| Volume Number | 28 |
| Alternate Webpage(s) | https://cloudfront.escholarship.org/dist/prd/content/qt0v59v30x/qt0v59v30x.pdf?t=op32ic |
| Alternate Webpage(s) | http://www.blutner.de/NeuralNets/Texts/NounNounCompounds.pdf |
| Alternate Webpage(s) | http://csjarchive.cogsci.rpi.edu/Proceedings/2006/docs/p184.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
