<?xml version="1.0" standalone="yes"?>
<Paper uid="C92-4176">
  <Title>RIDDLES: ACCESSIBI1JTY AND KNOWRID(H'I I{I':t)IC/I~SI'INTA'I'ION</Title>
  <Section position="2" start_page="0" end_page="0" type="metho">
    <SectionTitle>
2. ACC\]U',SSIBIIJTY IIII'~RAR(Jlt\[I:,S
</SectionTitle>
    <Paragraph position="0"> Cognitive psychologists have long recognized that people form taxonomic eaLegories (l{oseb 1978) wilt) smnc members heiul; more typical instances of those categories than others. This graded structure is not limited to taxonomic categories btR seems to include such .tmlikely possibilities as formal categories like the category of odd numbers a\]ld that of sq!Jares (Armstrong, Gleitman, and GMtman 19831 and linguistic categories for phones, phonemes and syntactic structures (Lakoff 1986). In recent years, researchers have shown that categories arc not structurally invarianL but are, in fact, demonstrably unstable (l/arsalou 1987). Their graded structure varies with &gt;leh factors as linguistic context and point of view and even differs for the same individual over time.</Paragraph>
    <Paragraph position="1"> The formation of ad hoe categories to accomplish specific /*oals (Harsalou 19fl3) is another area of instability irl human category formation, l&amp;quot;or example, the category &amp;quot;things to-take-out-of-a burning /louse&amp;quot; might include subordinate categories like 1The order of the names of the two authors is arbitrary. This paper represents tile joint work of both authors.</Paragraph>
    <Paragraph position="2"> AClEs DE COLING-92, NAIgllKS, 23-28 AOUI' 1992 1 1 2 1 Pl~oc. OF COLING-92, NANTES, AUO. 23-28, 1992 &amp;quot;children,&amp;quot; &amp;quot;jewels,&amp;quot; &amp;quot;paintings.&amp;quot; and &amp;quot;portable tv's&amp;quot; (Murphy and Medin 1985) and is formed only when one's house is burning or during a discussion like this one. Ad hoe categories, once formed, function similarly to more traditional categories, As we show later in this paper, ad hoe category formation is an important component in the generation or solution of riddles, A model that is to account for the human tendency to form categories must account for both the stable and the unstable aspects. Barsalou's approach to the instability in categories is to recognize the existence of both context-independent and context-dependent information m long-term memory, where it is arranged as interrelated, continuous knowledge. It is the context-independent information that is most likely to be shared by a large number of individuals within a speeeh community. Its activation is obligatory, When one thinks of r_cbilla, for example, &amp;quot;red-breasted&amp;quot; springs to mind whereas &amp;quot;poisonous&amp;quot; is triggered by rattlesnakes. Context dependent information, by contrast, is aeeessed only within a relevant context So &amp;quot;hard&amp;quot; may be triggered by ice while diseussing/all. The instability of categories is accounted for by different information for a given category being compiled in working memory at different times depending on the situation. Some information, e.g., eontext-independent information, is more aceessible than other information, We have extended this model (Weiner and De Palma 1993) to explain the tendency of people to think of the mouth of a person before mouth of a river in (1) above, Given the presumed universality of certain principles governing eategorization, it seems likely that, in context neutral situations sueh as (1), ambiguous words form ad hoe category-like structures of their multiple meanings onto which an aecessibility hierarchy is imposed. For example, in (1), there is a category-like structure corresponding to the phonemic realization of the word mo_ulk to whieh the different meanings belong; ill (1), one thinks of the mouth of a person before tile mouth of a river.</Paragraph>
  </Section>
  <Section position="3" start_page="0" end_page="0" type="metho">
    <SectionTitle>
3. THE KNOW~DGE BASE
</SectionTitle>
    <Paragraph position="0"> We thus offer our exposition of the structure that underlies the kind of lexieal ambiguity found in riddles as linguistic evidence for the episteinologieal requirements of a knowledge representation system whieh can support both normal discourse and riddles. Riddles will use the knowledge in one way: normal discourse will use it ill another. The representation will remain the same; only the algorithms will differ.</Paragraph>
    <Paragraph position="1"> Consider Figure 1, a knowledge-base fragment in the style of KL-ONE that contains the information necessary to generate or solve riddle (1). The KL-ONE entities most relevant to this discussion are Concepts (diagrammatieally represented by ellipses) and RoleSets (represented by encircled squares). The Concept is the primary representational entity. For us, it represents the category of objects indicated by the Concept name, Thus, in Figure 1, Concepts stand for the category RIVER-MOUTH, the category ANIMATE_MOUTlt, and so on. Concepts are connected to one another by superC links, represented in tile figures by double arrows, A superC link indicates that tile subordinate Concept (subConeept) stands in an inheritance and subsumption relationship with the superordinate Concept (superConeept). (The higher Concept subsumes the lower ojae; the lower one inherits from the higher one). Thus, PERSON_MOUTH is an ANIMATLMOUTH and a MOUTH, In our knowledge base, RoleSets represent, predicates of a Concept, the fi!\]ers of which, known as Value Restrictions (v/r's}, are themselves Concepts. So PERSON-MOUTH has a RoleSet &amp;quot;funetion&amp;quot; with the filler EAT, meaning ill our representation that a function of a person's mouth is to eat, {Of course there are others not shown here).</Paragraph>
    <Paragraph position="2"> f'urther, each RoleSet filler has a number restriction represented by two numbers within parentheses. These represent the lower and upper bounds on the number of fillers for a  given RoleSet, In Figure 1, we have arbitrarily estimated that people's mouths have a minimum of 6 and a maximum of 5 functions, Notiee that every Concept has a diamond-shaped symbol associated with it. This symbol is not part of the KL-ONE language. We are introdueing it here as a new primitive, l.edfical, which contains lexieal information about a Concept, For our purposes, Lexieal contains the phonemic representation of a Concept (although, for simplieity in this figure, only certain phonemic representations are aetually provided). This arrangement allows us to acknowledge the relationship between a Concept and the word used to name the Concept without asserting that they are the same thing, separating meanings of polysemous words from their phonemic representation, As discussed above, ambiguous (polysemous, homophonous) words can form ad hoe eategory-like structures of their multiple meanings. Thus, we can have a superConeept MOUTH, a category of polysemous words, with subConeepts ANIMATE_MOUTH and INANIMATE_MOUTH. We reeognize the probability that in the ease of ambiguous forms with a choice of animate vs, inanimate meaning, the animate one is thought of before the inanimate one (Weiner and De Palma 1993), So the ideas eneoded in Figure 1, although not explicitly spelled out with respect to aeeessibility, are based on the assumption that, in context-independent situations, people tend to think of animate things before they think of inanimate ones.</Paragraph>
    <Paragraph position="3"> In riddle (2),  (2) RQ: What has four legs and only one foot? RA: A bed.</Paragraph>
    <Paragraph position="4">  we model the riddling process by assuming that the phrase Lo~ ke4~ causes the formation of an ad hoe eategory &amp;quot;four legged thing.&amp;quot; A representation of a portion of the knowledge needed to generate or solve riddle (2) will be given in a future paper.</Paragraph>
  </Section>
  <Section position="4" start_page="0" end_page="0" type="metho">
    <SectionTitle>
4. THE ALGORITHM
</SectionTitle>
    <Paragraph position="0"> The following algorithm refers to Vqgure 1 and will generate riddle (1). The algorithm requires three functions: 1, I&amp;quot;indI-loms(HC1,HC2,C1,C2) - searches the knowledge base for two homophonous Coneepts, HC1 and HC2 where HCl and HC2 are the value restrictions of two Coneepts' RoleSets. Call these Concepts Cl and C2. CI must eontain the more accessible (i.e. in these examples, eontext-independent, animate) eoneept. For example, after an application of FindHoms(llCl,HC2,C1,C2), on the KB fragment contained in Figure 1, the variables would look like this:  RIVER_MOUTH, a value restrietion of C2 (RIVER) and the Concept in C1 (PERSON) is a more accessible Coneept than tile one in C2 (RIVER). 2. Lex(A,B) -. returns in B the word by which a  Concept, A, is known. Remember that the phonemic representation of this word is contained in &amp;quot;Lexieal&amp;quot; (represented in the figure by tile diamond shape) for each eoneept. For example, Lex(RIVER_-MOUTH,H) returns/mawO'/in 13, :3, MisMateh(CI,C2,HC1,HC2,Type,RSVR) - examines the knowledge base (KB) for a mismateh of the following type: HCI has a RoleSet value restrietion (RSVR) that He2 does have. hi Figure 1, this RSVR for HC1 would be EAT. Mismatch returns this in RSVR. Thus, using Figure 1, Mismatch would return EAT in RSVR. Note that HC1 is more aeeessible than HC2 by virtue of being animate. The algorithm, then, looks like this:  It should be noted that, in the interest of simplicity, we have eonflated tile issues involved ill generating or solving riddles. Once you know the heuristic with which riddles of tile type considered in this paper are constructed and have created a KB of Concepts, generation is a simple matte,'. Solution, of course, is the inverse of this algorithm</Paragraph>
  </Section>
  <Section position="5" start_page="0" end_page="0" type="metho">
    <SectionTitle>
5. CONCLUSIONS
</SectionTitle>
    <Paragraph position="0"> Our examples in this paper use KL-ONE as a convenient model of a knowledge representation system. We propose the addition of accessibility as all important epistemological primitive to the KL-ONE system since it appears critical to build this factor into any knowledge base which can both support a system for natural language processing and be used for certain kinds of humor. Our work also highlights other requirements for knowledge representation systems capable of supporting natural language:  1. Links between the phonemic representation of linguistic entities and their associated concepts (Lexical) 2. The necessity of representing homophonoua categories 3. The ability to form ad hoe categories such as  those based on homophonous phrases</Paragraph>
  </Section>
class="xml-element"></Paper>