<?xml version="1.0" standalone="yes"?>
<Paper uid="C90-3083">
  <Title>Local Cohesive Knowledge for A I)ialogue-machine Translation System</Title>
  <Section position="3" start_page="0" end_page="0" type="metho">
    <SectionTitle>
2. Contextual Robustness in l)ialogues
</SectionTitle>
    <Paragraph position="0"> Context-depcndent problems such as ellipses, anaphoras, pro-forms and referent-transfers, present complications as shown in Figure 1.</Paragraph>
    <Paragraph position="1"> (1) Anaphora: the previous utterance is the same, however, &amp;quot;it&amp;quot; points to the different terms, &amp;quot;the registration fee&amp;quot; in Example (1) and &amp;quot;the conference&amp;quot; in Example (2). Therefore context is complicated. In Example 1, the sequences of questions are disordered. In Example 2, the answer is a negation for the sentence, &amp;quot;I would like ...&amp;quot;.</Paragraph>
    <Paragraph position="2"> (2) Ellipsis'. in Example 3, there is an ellipsis in a Japanese sentence, (2) &amp;quot;motte inai no desu ga.&amp;quot; The F, xample 1 : sent, ence disorder  (2) I'm sorry. ~'s closed.</Paragraph>
    <Paragraph position="3"> l~xample 3 : (in Japanese) (1) k gr@tto kaado (credit card,) no (of) namae (name) o (OBJ) oshicte kudasai (Could you tell).</Paragraph>
    <Paragraph position="4"> \[ = Could yoti tell me the name of your credit card ?1 (2) sumimasen (I'm sorry). ~nottc (have) ina~_\[not) no desu ga ~. \[= l'msorry. Idon'thaveacreditcard.l Example 4: (in Japanese) (l) totlro\[~.uryou. (registration fee) wa (topic) en (yen) de (by) shiharatte yoroshii dcshou ka (can I pay).</Paragraph>
    <Paragraph position="5"> \[ = Can I pay the registration fee in yen?l (2) doru (U.S. dollars) de (in) ~',qishi C/na:s{~ (prg-\['orm = Wc \[ = We would like you to pay in US. doIlars,\] Figure 1 Examples of contextual phenomena, ellipsis depends on the context and means 'credit card'; it is both a focus and an object (OBJ).</Paragraph>
    <Paragraph position="6"> (3) Pro-form: in Example 4 (2), &amp;quot;onegaishimasu&amp;quot; is a pro- null form and means 'We would like you to pay' in Japanese. The meaning is dependent on the context.</Paragraph>
    <Paragraph position="7"> We call processing the disarranged phenomena &amp;quot;contextual robustness!~&amp;quot;. In order to process such phenomena, it is necessary to understand cohesion in a context correctly.</Paragraph>
    <Paragraph position="8"> 3. l,oeal cohesive knowledge We define cohesion in the view of computational linguistics. Here cohesion regulates whether two sentences are connected or not. However it does not regulate a relationship between two sentences. That is, cohesion is a constraint for two sentences.</Paragraph>
    <Paragraph position="9"> \[The definition of &amp;quot;local cohesive knowledge&amp;quot;\] In our approach, &amp;quot;cohesion&amp;quot; is grasped in a context with &amp;quot;local cohesive knowledge&amp;quot;. It includes not only the constraints tbr &amp;quot;local cohesioW~&amp;quot; but also its results such as interpretations of ellipses, anaphoras, pro-forms and referent-transfers. Therefm'e if constraints are satisfied, the interpretations are obtained. Therefore &amp;quot;local cohesive knowledge&amp;quot; has two parts, &amp;quot;constraints for cohesion&amp;quot; and &amp;quot;inter-pretations&amp;quot;, as follows. (Constraints for local cohesion) = &gt; (interpretation) tl. Ordinarily, robustness means an ungrammatically sentence. I lowevm' &amp;quot;contextual robustness&amp;quot; is used for the discourse level. t2. We treat the contextual phenomena which occur locally, thus we use the term, &amp;quot;local cehcsion&amp;quot;.</Paragraph>
    <Paragraph position="11"> The constraints are described as follows.</Paragraph>
    <Paragraph position="12"> verbl &lt; X1,Y1,Z1 &gt; ,verb2 &lt; X2,Y2,Z2 &gt;.</Paragraph>
    <Paragraph position="13"> In the &amp;quot;verbl&lt;Xl.,Y1,Zl&gt;&amp;quot;, &amp;quot;XI&amp;quot;, &amp;quot;YI&amp;quot; and &amp;quot;ZI&amp;quot; means the case elements of &amp;quot;verb1&amp;quot;; subjective (SUBJ), objective (OSJ) and second objective (oBJg) cases. If two sentences are satisfied with these constraints, they are called &amp;quot;local cohesion&amp;quot; here. As shown in Figure 2, there are 18 types, determined by three constraints for verbs and six constraints for nouns.</Paragraph>
    <Paragraph position="14"> Type h the same verbs and the same nouns.</Paragraph>
    <Paragraph position="15"> For example, &amp;quot;Could you send me a paper?&amp;quot; &amp;quot; \[ sent you the paper yesterday.&amp;quot; Both of the verbs in the question sentence and the answer sentence are the same words, &amp;quot;send&amp;quot;. Also, its object is the same word, &amp;quot;paper&amp;quot;. This constraint is described as follows.</Paragraph>
    <Paragraph position="16"> send &lt; Xl,paper, Z1 &gt;, send &lt; X2,paper,Z2 &gt;.</Paragraph>
    <Paragraph position="17"> This constraint means that if two sentences include &amp;quot;send&amp;quot; and its object, &amp;quot;paper&amp;quot;, the sentences are cohesive. Therefore the following sentences are cohesive because they satisfy the same constraint.</Paragraph>
    <Paragraph position="18"> For example, &amp;quot;May I send you a paper to your office?&amp;quot; &amp;quot;Please send me the paper to my home address.&amp;quot;  problems such as anaphoras, ellipses, pro-forms and referent-transfers. Local cohesive knowledge has interpretation. If the constraints are satisfied, its interpretation is obtained. Examples are shown in Figure3 (b) and (c).</Paragraph>
    <Paragraph position="19"> (b) Interpretation of an anaphora: for example, &amp;quot;Could you send me a paper?&amp;quot; &amp;quot;I will send it to you. &amp;quot; (c) Interpretation of an ellipsis: for example, &amp;quot;Could you send me a paper?&amp;quot; &amp;quot;I will send ~ to you.&amp;quot; ; ~ means an ellipsis. ( In Japanese dialogues, such an ellipsis is found often.) Ca) send&lt;Xl.,paper, Zl&gt;,seud&lt;X2,paper,Z2&gt;.</Paragraph>
    <Paragraph position="20"> (b) send &lt; Xl,paper, Z1 &gt;, send &lt; X2,it,Z2&gt;,</Paragraph>
    <Paragraph position="22"> 4. Context proeessing with local eohesive knowledge  I will now explain the mechanism which is useful for &amp;quot;contextual robustness&amp;quot;, and interpret contextual phenomena such as anaphoras, ellipses and pro-forms. A flow of the system is shown in Figure 4. Inputted sentences are analyzed with grammar rules and lexicons, based on LexicaLfunctional Grammar (LFG) (1), and then intermediate representations ( F-structures of LFG ) are obtained. An intermediate representation is converted into its skeleton, because it has too much information to process for a context, in Figure 5. It is used to unify with &amp;quot;local cohesive knowledge&amp;quot; in the context processing. The algorithm of the context processing mechanism is as follows.</Paragraph>
    <Paragraph position="23">  (1) Make a pair of skeletons: to check the local cohesion, bring the skeletons of the previous utterance and make a pair of skeletons.</Paragraph>
    <Paragraph position="24"> (2) Check the local cohesion: look up the table of &amp;quot;local  cohesive knowledge&amp;quot; as a key of the pair of skeletons. If the pair satisfies the constraints of &amp;quot;local cohesive knowledge&amp;quot;, the pair is cohesive and then the  interpretations of ellipses, anaphoras, pr0-forms and ref'erent transfers are obtained with &amp;quot;local cohesive knowledge&amp;quot;.</Paragraph>
  </Section>
  <Section position="4" start_page="0" end_page="0" type="metho">
    <SectionTitle>
5. The experiment
</SectionTitle>
    <Paragraph position="0"> When we built the system, one of the most important problems was how to produce the knowledge. We defined the local cohesive knowledge and used its definition to extract knowledge from a linguistics database almost automatically.</Paragraph>
    <Paragraph position="1"> We have a linguistic database which includes 60 keyboard dialogues. The dialogues include 70,000 words in total and the number of different words is more than 3000. These dialogues are analyzed and managed by a linguistic database (P-).</Paragraph>
    <Paragraph position="2"> We extracted local cohesive knowledge from 60 dialogues which include 350 verbs and 1000 nouns. First we made a table which includes each verb and its noun. Then we extracted constraints of local cohesive knowledge to make the pair from the table. Constraint pattern (a), a:!~ shown in Figure 3, was obtained automatically from the data and patterns (b) and (e) were generated from pattern (a). We obtained 24531 assertions of&amp;quot;local cohesive knowledge&amp;quot; for types 1, 2 and 3, and 651 a.,.;:sertions of&amp;quot;local cohesive knowledge&amp;quot; for t.ypes 7, 8 and 9, We have learned that local cohesive knowledge is very sparse. Therefore the volume of &amp;quot;local cohesive kzmwledge&amp;quot; is not a problem.</Paragraph>
    <Paragraph position="3"> We have implemented the fi'amework as a module of a  i(1&amp;quot; I sle.eletml ) = ; ( 1 ) skeinton i {iifl I qUiD ; ='toil &lt; {l'; SU B,I ;,, f OBJ 2i,(f,. Oi~j ~ &gt;'\], \[(ft SU ILj )=:l:e, i fe pRI!;D)__ ~ \] ' ;~N.B) !fx lq(I';I)) - -@, \[fit ()F;J2)=: f.a, Ila Pllt&amp;quot;,ll) = ~\], ; I~ men c,s an ellipsis, {I\['I O ~J = \[I. &amp;quot;1 P x 'H) ='nu d el&amp;quot;, t~ MOI))= fn, I'~ Pl{i,il)l='c~edi'~card'  (N.B) I lere the local cohesive kni)wlcdge ( 1 ) is k el)resented us Li&amp;quot;() rel/t esentat ion, the h)cal cohesive knowledge {2). It, is equivalent. In Ihe imph!Inentatil/n Ihe \[,I;'G style was used, Figure 5 Examples of a pair of skeletons and their local cohesive knowledge.</Paragraph>
    <Paragraph position="4"> context process in a dialogue machine-translation system. The system is built on a Lt?G based machine-translation system (3). It has 200 grammar rules and more than 3000 words. It transfers Japanese sentences into English ones. It was implemented in Quintus Prolog on a SUN-4 system and its program size was 3.4MB.</Paragraph>
    <Paragraph position="5"> An example is shown in Figure 5.</Paragraph>
    <Paragraph position="6">  (1) kurejitto kaado (credit card) no (of) namae (name) o (OBJ) oshiete kudasai (Could you tell).</Paragraph>
    <Paragraph position="7"> \[ = Could you tell me the name of credit card. 9\] (2) motte (have) inai (not) no desu ga (copula).</Paragraph>
    <Paragraph position="9"> in the sentence (2) there is an ellipsis. It means &amp;quot;kurejitto kaado (credit card)&amp;quot;. It points to the modifier in the previous sentence, &amp;quot;kurejitto kaado no namae ( name of credit card)&amp;quot;. In this approach, as a results of analysis, the skeletons of two sentences are obtained as shown in Figure 5. The pair of skeletons are satisfied with the local cohesive knowledge (c) in Figure 5. Then the ellipsis is obtained as a 'credit card'.</Paragraph>
  </Section>
  <Section position="5" start_page="0" end_page="0" type="metho">
    <SectionTitle>
6. CONCI:USIONS
</SectionTitle>
    <Paragraph position="0"> To build a &amp;quot;contextual robustness&amp;quot; system, we proposed a context-processing mechanism which analyzed the context with &amp;quot;local cohesive knowledge&amp;quot;. In order to apply the model into a machine-translation system, the knowledge needs to be produced effectively. Therefore we defined 18 types of&amp;quot;local cohesive knowledge&amp;quot; and used this definition to abstract knowledge from a linguistics database almost automatically. Some of the 18 types were implemented on a machine translation system. The other types were not generated, because they includes synonyms. In l, he future, we will construet them with a thesaurus and also extend the context processing algorithm to process more complicated phenomena such as parallel phrases.</Paragraph>
  </Section>
class="xml-element"></Paper>