File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/intro/86/c86-1120_intro.xml
Size: 24,677 bytes
Last Modified: 2025-10-06 14:04:31
<?xml version="1.0" standalone="yes"?> <Paper uid="C86-1120"> <Title>Japanese Honorifics and Situation Semantics</Title> <Section position="3" start_page="0" end_page="509" type="intro"> <SectionTitle> 1. Introduction </SectionTitle> <Paragraph position="0"> Tile Japanese language, like Korean and many nonoEm'opean languages, contains a grmmnaticalized system of Ironer)tic forms. It is well known that the use of honor)ties is closely connected to context, inehating aspects like relatiw, social standing. No effective mechanisms have heen developed to deal with this problem. Situation Semantics (SS) IBarwise 1982,1984a,1984b,1985a, 1985b,1985c,1985d,1985ell Barwise & Perry 1983\] \[l,experaneeJ\[Pollard 1985HCreary & Pollard\] is a ttmory of context used here to construct a nmdel of honorific sentences to analyze the relationship tmtween sentence and context.</Paragraph> <Paragraph position="1"> About Japanese, we call make use of Mizutani's theory of honor)lies \[Mizutani 1983a,1983b\]. This theory dees tackle tire relation between context and sentence, but it seems that SS can describe context move usefldly than Mizutani's theory. In this palter, SS is used to reconstruct the context given by Mizutani's theory. Ilonorifie forms arc analyzed and basic rules flu' context switching are established.</Paragraph> <Paragraph position="2"> Table 1 gives the relation between Mizutani's theory of honorifles and the model. This model can be divided into two parts. The first part describes basic context features and the second describes lexical rules based on Mizutani's Jal)anese grammar, along with some basic mechanisms of &quot;context switching.&quot; Table 1 Mizutani's Timory Our Model Social status Binary reiatioia Context of honorifics as vector element as event type in SS Discourse formalism none SS Gramnmr Japanese grammar Japanese grammar in CFG Form in IICG form Context switch in none In complex sentence honorific sentence</Paragraph> <Paragraph position="4"> I~ is very easy to reln'esent some context features in discourse in SS, but context involves some very difficult problems like &quot;focusing.&quot; We want to analyze this in future research. In this paper only enough elements for context required in Mizutani's theory are set up.</Paragraph> <Paragraph position="5"> The main concern here is the second part of the model which deals with the relationship between contextual elements and lexical elements, attd especially the mechanism of &quot;context switching&quot; on honor)tics.</Paragraph> <Paragraph position="6"> Mizutani's theory of honorific exltression in Japanese is described first.</Paragraph> <Paragraph position="7"> Then the model in ~{S is presented. The context feature of relative social status in Mizutani'n theory is realized in the first part of the model. This is followd by explanal:ion ef some basic features designed to handle the mechanism of context switching in the second part. Finally, au experimental system based on this model is given. Tiffs system was implemented in CfL (Complex Indeterminate Imnguage)\[Mukai 1985a, 1985b1. Results from some experiments processing Japanese honorific expressions are given at the end of the paper.</Paragraph> <Paragraph position="8"> 2. Mizutani's theory of Japanese honorific expression</Paragraph> <Section position="1" start_page="0" end_page="508" type="sub_section"> <SectionTitle> 2.1. Honorific Status </SectionTitle> <Paragraph position="0"> tlonorific relations are represented as vectors in an abstract twodimensional henorifie space. In the honorific space the speaker in tim discourse is set as tile origin. Other individuals like hearer and agents wire are presented in the sentence of discourse are represented by vectors as in the following example.</Paragraph> <Paragraph position="1"> An honorific attitude is defined as tile vector between these points. For example, the honorific attitude from tim speaker (1) to the hearer (Y) is defined as a vector fi'om I{0,0) to Y(yl ,y2). The honorific attitude fi'mn the hearer Y to agent2 is defined as the vector fi'om (yl,y2) to (a21, a22), i.e., the vector (a2l-yl, a22-y2).</Paragraph> <Paragraph position="2"> Next, we define the honorific wlluc and the dh'ection in the following way, Definition 1 Honorific value</Paragraph> <Paragraph position="4"> We represent the &quot;flat&quot; honor,fie relation between agents explicitly, but the &quot;across&quot; relation is represented as in which there is no need to express the honor,fie relation explicitly. For conventional use, we define the following direetlons.</Paragraph> <Paragraph position="5"> 5) anyl up or down or flat. 6) any2 up or down or flat or across Te analyze ttte sentence uttered, definitions of tile following honorific relations are necessary, Definition.3 First order honorific relation The honer,fie attitude of the speaker to tim hearer.</Paragraph> <Paragraph position="6"> IY Definition4 Second order honorific relation The honorific attitude of tl,e ,;peaker or tile hearer toward the agent in the sentence of discourse. In this case the original point of tile vector should be tile point of the hearer or sl)eaker, whichever is higher. If I > 1\[ the original point of the vector will be l, att(t if \] < I1 tile origin will H We introduce the notation J which stands for&quot; tile higher individual. JA Definition 5 Third order honorific relation The tmnorific relation between agents ofs discourse.</Paragraph> <Paragraph position="7"> A1 A2 2.2. dapanese Gramnmr Now we can represent tim relation between these three honorific relations and the sentences of discourse. Before looking at this honorific relation, we will show the small oumber of Japanese grammar rules used fo define the structure of sample Japanese seutenees. This grammar is taken fi'om the &quot;Sketeh of Japanese Gramlnar&quot; \[Mizutani 1983al, a part of which is shown in Figure l. This grammar is presented in CFG, but we give the rules in I)CG.</Paragraph> <Paragraph position="8"> Figure. 1 > sn0,1toLsn0. *2) sn ~-> sm,s0,em,ltel,s0.</Paragraph> <Paragraph position="9"> --> ph,ps, ph -.-> np,cm,ph.</Paragraph> <Paragraph position="10"> > vp. vp --> subl,v,sub2, honm'ificform --> v. vp --> v,sub4, honorific form > n. nit --> n,sub3, honorific fbrm - > \[(tarol;\[hanako I. (nanm of individual) --> \[gal. em - > \[nil. agent/objectcase ma,'ker --> \[anl. (meet) v --> \[iu\]. (say) > \[tal. ps ~-> \[mashi,tal. honorific form > \[ol, sub2 - > \[nasarul.</Paragraph> <Paragraph position="11"> --> \[mousul. sub3- > \[samaJ. (like &quot;Mr.&quot;) tu his theory, Mizutani attaches Japanese terms to each terminal and non terminal node. Table 2 gives the correspondence between Mizutani's Japanese terms and standard English grammatical terms. Rules *1) and &quot;2) above are not represented in his grammar. These rules were included specifically to represent direct and indirect speech.</Paragraph> <Paragraph position="12"> 'Fable. 2 sn(sentence nuclear) = juttaiku sn0(senteuce nuclear 0) = juttaiku0 era(case marker) = kakuhyoji, ps (post sentence marker) = juttaiji. ph (phrase element) = jutsuso, vp (verb phrase) = yorengo. v (verb) = doshi np (noun phrase} = tairengo, n (noun) = taigen. 2.3.Relation between honorific context and Japanese grammar. The following relation holds between honorific relations and the sentence of discourse. The notation &quot;= = = >&quot; means the left element has some effect on the right element.</Paragraph> <Paragraph position="13"> l ) Changes in the the first order honorific relation arc shown. 2) Changes in the second order honorific relation are shown by description of individual A1 and the honorific elements subl and sub2 of vp in the agent case, and by description of individual A2 as the np in the object case.</Paragraph> <Paragraph position="14"> Rule2 relation J,A1 ===> np (in agent case) vp (subl and sub2) J, A2 = = = > np (in object case) 3) The third order honorific relation is shown by v in vp Rule 3 relation A1,A2 ===> v 3. Model of the honorific sentence In this section, we present our model in SS. Readers who are familiar with SS can skip Section 3.1.</Paragraph> <Paragraph position="15"> 3.1. First Part on Context Features 3.1.1. Model of discourse In the theory of situation senmntics, discourses are represented within situations and constraints. In formal representation the sentence and context are expressed in the form &quot;d,c \[sentence\[s, e&quot;. d stands for discourse situation (l)U), c is the speaker's connection (CS), s is abe setting and e is the described situation. There is another situation called a resource situation in this theory. In our model we define a resource situation for each individual.</Paragraph> <Paragraph position="16"> (I) Dhcourse Situation (I)U) lIere is an example of the representation of a sentence in discourse. Ex 2. John said &quot;Tom met Jane&quot; to Jim.</Paragraph> <Paragraph position="17"> el : = loc, 10 ; yes saying John ; yes addressing John,Jim ; yes uttering &quot;Tom met Jane&quot; ; yes l0 < ld (ld means discourse location) Discourse situations are represented in the following form as series of terms expressing relations between speaker A, hearer B and uttered sentence \]\[alpha\]\]. in SS, spatio4emporal location is defined in discourse situations, but we have no effective way to formalize it in our model, so spatio-temporal location is not represented.</Paragraph> <Paragraph position="18"> DU : = speaking A;yes addressing,A,B;yes saying,A,alpha;yes (11) Speaker's Connection (CS) The speaker's connection is a series of the following types. A is the actual object and \[AI is the word that represents A.</Paragraph> <Paragraph position="19"> CS : = speaking A;yes refers A \[A\] ; yes (111) Resource Situation(RS) A resource situation is defined for each individual in a discourse; it contains many events and constraints.</Paragraph> <Paragraph position="21"> 3.1.2. Honorific context in the sentence (I) Honorific Attitude Event type We introduce an &quot;honorific attitude event type (Eh)&quot; which stands for an honorific attitude in Mizutani's theory.</Paragraph> <Paragraph position="22"> Formula 1) represents the mind status (HE) of the speaker towards Eh. A denotes the individual and &quot;Itr&quot; indicates that in the nfind of the individual A, the honorific event type Eh is &quot;represented.&quot; Eh shows the external honorific event including social honorifics.</Paragraph> <Paragraph position="23"> Formulas 2) through 5) represent the basic honorific attitude event types (Eh). In Mizutani's theory, this event type is represented as a vector, but we represent it in the form of a binary relation. For one example, in 2) &quot;honor-up&quot; denotes the honorific relation between individuals t and t'. &quot;ind t __ ; yes&quot; denotes that t is an individual which can be represented in&quot; &quot;as its name.</Paragraph> <Paragraph position="24"> Frame of Mind 1)ilE:= Itr, A, Eh ;yes oft,_;yes oft',.r_;yes 2) Eh : = honor up t,t';yes ind t ,_;yes ind t ,__ ;yes 3) Eh : = honor down t,t';tyes ind t, __ ;yes ind t',__ ;yes 4) Eh honor~across t,t ;yes ind t, __ ;yes ind t' ;yes 5) Eh : = honor eq t,t';yes ind t, _ ;yes ind t' _ ;yes (11) Conditional Constraint on Word Selection In Section 2.3. we described the relation between honorific context and grammatical rules. This section corresponds to Section 2.3. In Japanese there are many honorific words, and some interesting phenomena are to be found in daily life.</Paragraph> <Paragraph position="25"> For example, a secretary in the company president's office should have many honorific words in his resource situation (RS) because he always has to be careful to use the appropriate honorific expression in his wm'k. On the other hand university students will have a poor stock of honorific words for there is no need to express honorific status, except to teachers. A constraint is required which determines the relation between the honorific event type (Eh) and word representation. This is the &quot;conditional constraint on word selection (Cw).&quot; Mizutani's rules for first order through third order honorific relations given in 2.3. correspond to this constraint Cw. Word representation should be an event type called &quot;honorific word selection event (HW)&quot;, and if a person does not have this event type in his resource situation, he or she will have a poor range of honorifics. We represent these Cw and HW in formula 6).</Paragraph> <Paragraph position="26"> In formula 6), &quot;Cwl(Eh,~q&quot; denotes the conditional constraint of word selection Cw which has Eh and anchor &quot;f&quot; as its conditional schema. Anchor &quot;f&quot; determines the relation between indeterminates in Eh and objects like taro and hanako. See Barwise's work for details. 6) is read : ifEb is factual and Cw is satisfied, then ItW is actual.</Paragraph> <Paragraph position="28"> Conditional constraint on word selection has an honorific event type and its anchor as its scheme. In Japanese honorific expressions, if Eh and f are given, the word representation will follow very easily. Some instances of thls are given in formulas 7) through 15). For example, formula 7) can be read : if Eh is actual and the anchor f anchors t to individual J (J is the speaker or the hearer) or t' to A1 in the agent case, then hw is actual so that when refering to t' honorific form tairen-go (np) is added to the description Before going into the representation of lexical rules which give the correspondence between honorific word expressious and sentences, and context switching on honorific sentences, we describe CIL used in the description of lexical rules. Accurate accounts of C1L can be found in \[Mukai 1985a,1985b\]. llere only the part needed to understand the lexical rules in section 3.2.2 is describe(\[.</Paragraph> <Paragraph position="29"> C1L has the unique data structure called &quot;complex indeterminate.&quot; This data structure can be regarded as a frame and represented as in the following example.</Paragraph> <Paragraph position="30"> (1) term((Xwitha:= X,b:= YwhereX = Y)).</Paragraph> <Paragraph position="31"> (2) > term(Z),a!Z = abc,b!Z = abc.</Paragraph> <Paragraph position="32"> yes.</Paragraph> <Paragraph position="33"> Formula (1) is the specification of the data type and this can be used in formula (2). &quot;a := X&quot; means assign value X to slot a. &quot;X=Y&quot; denotes the condition part. If this condition is not satisfied the unification &quot;term(X)&quot; in (2) will fail. Formula (2) utilizes the complex indeterminate and unifies &quot;abc&quot; to each slot value.</Paragraph> <Paragraph position="34"> 3.2.2. DCG Rules for honorlfies (I) First order bonorifics As Rule 1 in Section 2.3. shows, first order honorifics affect &quot;ps.&quot; This is illustrated by the following examples. Ex.3 This corresponds to Eh in 11) in section 3.1.2.</Paragraph> <Paragraph position="36"> eq(speaker\](ds!Context),object!(lex!Context))).</Paragraph> <Paragraph position="37"> ,,< 2, denotes the operator &quot;:-&quot; in Prolog \[Bowen 1982\]. &quot;\[ta\]X\],X&quot; is the DCG parsing mechanism. &quot;Context&quot; is a complex indeterminate variable for the context for this parsing part. &quot;dsolve&quot; is a Prolog predicate with the following mechanism.</Paragraph> <Paragraph position="38"> dsolve(X,Y). X is a list like \[a,b,c,dl which contains Prolog atmns or terms. Y is a Prolog atom or term.</Paragraph> <Paragraph position="39"> 1) Search list X for Y.</Paragraph> <Paragraph position="40"> 2) If there is a term in X with the same arguments but different term name, fail and return.</Paragraph> <Paragraph position="41"> 3) if Y is not in X, then add Y to list X,succeed and return. 4) lfY is in X, succeed and return.</Paragraph> <Paragraph position="42"> &quot;hono!((agent!(ds!Context))!(rs!Cantext))&quot; represents a list of honorific event types in the resom'ce situation of the agent of discourse. &quot;object! (lex!Context))&quot; represents the lexical object in this parsing stage. This notation for lexical items has its origin in Lexical Functional Gramrnar tKaplan & Bresnaul, so this expression can be represented like (~object) in the LFG manner.</Paragraph> <Paragraph position="43"> This example states that if&quot;ps&quot; = |taL then there should he honorific information in the resource situation of the individual who is the speaker. If the speaker's RS contains two or nmre different terms expressing the honorific relationship between the same agents, fail. Thus, the mechanism of 2) in dsolve is very important because it shows that in the honorific information of one individual there should not be different information about the binary honorific relation between two individuals. (11) Second order honorifics Lexieal rules for second order lmnorifies can be represented as in following example program.</Paragraph> <Paragraph position="44"> Ex.4 This corresponds to Eh in 7) and 8) in Section 3.1.2.</Paragraph> <Paragraph position="46"> up(agenl,!(ds!X),obj!(lex!X))).</Paragraph> <Paragraph position="47"> (Ill) Third order honorifics Lexical rules for third order honorifics can be represenl;ed as in the following example program.</Paragraph> <Paragraph position="48"> Ex.5 This corresponds to Eh in 10) in section 3.1.2.</Paragraph> <Paragraph position="50"> up(agenti(lex!X),obj!(lex!X))).</Paragraph> <Paragraph position="51"> 3.2.3 Context switching in honorific sentences When we utilize the contextual elements like I)S and RS in discourse it is very difficult to decide the context for each sentence. A sentence in discourse can be represented by the expression &quot;I)S,CS,I\[alptm\]\]S,E&quot;, but then how do we nmp contexts like DS and CS to complex sentences ? Mizutani's theory of honorific forms does not go into context switching in a complex sentence. So we have expanded his grammar and propose a basic mechanism for contcx~ switching.</Paragraph> <Paragraph position="52"> Consider sentence l) below uttered by individual S to R which means &quot;individual T said that individual U said that Taro met llanako.&quot; In this example, we establish relations a) through j) among S, T, U, Taro and ltanako. The operater > denotes the situation in which tim left hand side honm's the right hand side, < denotes the situation in which the right had side hmmrs the left hand side, and = denotes the situation in which there is no need to use honerifics between left hand side and right hand side.</Paragraph> <Paragraph position="53"> The main point in utterances of this fm'm is that honorifics in these sentence change according to the form of speech, such as direct or indirect speech. BuL in the Japanese discourse there are no markers like &quot;'&quot; and &quot;'&quot;, so in order to process these sentences correctly, we need the mechanism of&quot;context switching.&quot; Without this mechanism, all sentences wonld be parsed with one context, but this cannot explain the reason why honorifics change in complex sentences.</Paragraph> <Paragraph position="54"> 1) In S's utterance he said to R ( &quot;T said, U said, Taro met tlanako&quot;) &quot;taro sama ga hanako sama ni ai nasa( ta' to U ga iware ta' to T ga iu ta.</Paragraph> <Paragraph position="56"> These are the parsing rules used to analyze utterances.</Paragraph> <Paragraph position="57"> IA) sentence--> an(X), ds!X = Y, agent!Y = s, ohilY = r,at!Y = L,</Paragraph> <Paragraph position="59"> LI) and L2) are fornml rules to start the process, while L3),L4} and LS) are basic rules for determining context switch in sentences of discourse.</Paragraph> <Paragraph position="60"> LI) specifies the initial stage for parsing. In 1) above S tells R something so this context is set in the slot denoted by ds!X.</Paragraph> <Paragraph position="61"> L2) states that all of the features of sn0 are transfered to sa to meet a requirement of Mizu(ani's grammar. This is done easily by unification.</Paragraph> <Paragraph position="62"> 1,3) means that all of the featnres in sn0 are transfered to sn. This mechanism corresponds to indirect speech.</Paragraph> <Paragraph position="63"> L4) means that there should be context switch. As the discourse situation for sn0, set agent of discourse of sn0 to agent of Z who utters sn0 and set object of diseom'se of sn0 to object of Z who hears this utterance.</Paragraph> <Paragraph position="64"> L5) means that in a sentence with no marker, there cat) be context switch, so if a parsing failed because of the context ofhonorifics, use this rule.</Paragraph> <Paragraph position="65"> Sentence 1) is analyzed using rule 1,5) and the mechanism of context switch is dcerived fi'om phase (I) to phase 010.</Paragraph> <Paragraph position="66"> (I) Parsing really starts with the rule 3) estimating that there is no context switching. But at point *1, a conflict between S's resource situation and honorific expression occurs. In S's resource situation, the honorific relation between S and Taro is down(S, taro), but \[taro,samal requests the honorific relation up(S,taro), so context switch occurs at *2. Rule 5) switches the agent of discourse fi'om S to U. Context switch does not occur at this point again. We use the notation I)S(S) to state that the agent of discourse is S. The symbol --> means the context of left hand side is changed to the context of the right hand side as the result of context switch. (T said U said Taro met llanake) taro sama ga hanako sama ni ai nasa( ta to U ga iware ta to T ga iu ta.</Paragraph> <Paragraph position="67"> in RS of S s</Paragraph> <Paragraph position="69"> taro sanla ga banako sanla ni o all nasat I,a to g ga iwaro ta ph ~ps nlFeI-lih (t0 Next, the parsing mechanism finds a conflict at point '3). lU,ga, iware,tal requests honorific relation up(S,U), hut the resource situation of S contains the honorific relation down(S,U). Then context switch given in 5) is al)plied at p(fint *4).</Paragraph> <Paragraph position="70"> in RS of S s n sub3 su 3 su ub2 n v I I I taro sama ga hanako sama nio au nasat ta to U ga iware ta to T ga it ta</Paragraph> </Section> <Section position="2" start_page="508" end_page="509" type="sub_section"> <SectionTitle> 3.3. Ch and Cw </SectionTitle> <Paragraph position="0"> Now, we have come to the main point of our model, but there remains an interesting feature of Cw. This constraint is not verified so with some trepidation we touch on it briefly here.</Paragraph> <Paragraph position="1"> For example, when a worker &quot;Suzuki&quot; refers to his friend &quot;Tanaka&quot; with contempt, he will intentionally use a polite word to refer to him such as &quot;Tanaka sensei&quot; (Mr.Tanaka). When the hearer (Y)bears this polite expression,he decides on honorific event types but finds conflicts between these types and the normal social event types in his resource situation.</Paragraph> </Section> <Section position="3" start_page="509" end_page="509" type="sub_section"> <SectionTitle> Ex.6 </SectionTitle> <Paragraph position="0"> expression \[Tanaka sensei\] where Eh : = at h honor up,Suzuki,Tanaka expression \[Tanaka sensei\] where Eh : = at l: honor down,Suzuki,Tmml~a in RS of Y</Paragraph> <Paragraph position="2"> --> hearer find conflict and the hearer wonders why \[! Then the hearer (Y) wonders why he broke the universal honorific event. Finally, he comes to the conclusion that Suzuki intends either to praise Tanaka or berate him. We can go no further on this problem here.</Paragraph> <Paragraph position="3"> There are other aspects to Eh. If the sentence is given first, Eh will be calculated for each word and there remains a possibility of conflict between honorific event types Eh in a simple sentence. In a complex sentence the mechanism of context switch will be used, but in a simple sentence this mechanism is not effective. When the hearer tries to deal with this conflict, he or she will assume that the speaker has some illegal honorific constraint Ch'. We have implemented this mechanism in our model system.</Paragraph> </Section> </Section> class="xml-element"></Paper>