File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/94/a94-1038_metho.xml
Size: 8,192 bytes
Last Modified: 2025-10-06 14:13:38
<?xml version="1.0" standalone="yes"?> <Paper uid="A94-1038"> <Title>Integration of example-based transfer and rule-based generation</Title> <Section position="4" start_page="196" end_page="196" type="metho"> <SectionTitle> 2 Example-based transfer </SectionTitle> <Paragraph position="0"> The transfer module outputs the TE structure from an input sentences by using translation examples.</Paragraph> <Paragraph position="1"> In the proposed translation method, translation examples are classified based on the string pattern of their SE and are stored as empirical transfer knowledge, which describes the correspondence between an SE and its TEs. The following is an example of the transfer knowledge about &quot;X to Y&quot; at the verb-phrase level.</Paragraph> <Paragraph position="2"> X to Y --~ Y' e X' ((go, Kyoto)...) Y' ni X' ((pay, account)...) Y' wo X' ((refer, announcement)...) The first possible TE is &quot;Y' e X&quot;, with the example set ((go, Kyoto)...). Within this pattern, X' is the TE of X, which expresses a variable corresponding to some linguistic constituent. (go, Kyoto) are sample bindings for &quot;X to Y&quot;, where X =&quot;go&quot;, and Y : &quot;Kyoto&quot;.</Paragraph> <Paragraph position="3"> Patterns in transfer knowledge are classified into different levels according to the scale of their linguistic structure in order to restrict the explosion of structural ambiguity, and an input sentence is decomposed into chunks by applying SE parts of transfer knowledge in a top-down fashion.</Paragraph> <Paragraph position="4"> Suppose that an input sentence is &quot;Please come to our company.&quot; SE parts of transfer knowledge are applied in the order, &quot;please X&quot; (simple sentence), &quot;X to Y&quot;(verb phrase), &quot;our X&quot; (compound noun), &quot;come&quot;, &quot;company&quot; (surface word), yielding the following SE structure: (Please ((come) to (our (company)))) For each chunk of the SE structure, the most appropriate TE is selected according to the calculated distance between the input words and the example words. The distance calculation method of (Sumita, 91) is adopted here. The distance between words is defined as the closeness of semantic attributes in a thesaurus.</Paragraph> <Paragraph position="5"> The SE structure chunks of &quot;Please come to our company&quot; are transferred to &quot;X' tekudasai&quot;, &quot;Y' e X'&quot;, &quot;watashi-tachi no X'&quot;, &quot;kuru\[come\] 1, and &quot;kaisha\[company\]&quot;. By combining these TE chunks, the following TE structure is obtained, which will be the input of the composition module of the rule-based generation model: (((watashi-tachi no (kaisha)) e (kuru)) tekudasai) In the above structure, the honorific word &quot;irassharu\[come\]&quot; is more adequate than the neutral word &quot;kuru&quot; from the point of view of politeness. The replacement of &quot;kuru&quot; with &quot;irassharu&quot; will be done in the adjustment module of the rule-based generation model.</Paragraph> </Section> <Section position="5" start_page="196" end_page="196" type="metho"> <SectionTitle> 3 Rule-based generation </SectionTitle> <Paragraph position="0"/> <Section position="1" start_page="196" end_page="196" type="sub_section"> <SectionTitle> 3.1 Composition </SectionTitle> <Paragraph position="0"> The composition module checks whether a transferred sentence is grammatically appropriate or not, and corrects grammatical errors produced by the structural gap. The composing method is almost the same as the syntactic analysis method. However, the process is much simpler, because the input string has the correct Japanese structure and the corresponding English expressions.</Paragraph> <Paragraph position="1"> The procedure is as follows: 1) Divide the sentence into clauses, using not only Japanese grammatical features but also the TE structure and its corresponding English expressions. 2) Analyze each clause using the Japanese syntax rule. 3) Check on Japanese grammatical constraints. If the process finds violations, it corrects them by using Japanese linguistic knowledge.</Paragraph> <Paragraph position="2"> Japanese sentences have a peculiar grammatical constraint. Some expressions cannot appear in a subordinate clause. For example, the postpositional particle &quot;wa(topic marker)&quot; cannot appear in a conditional clause. Table 1 gives examples of limitations on expressions. In Table 1, &quot;masu&quot; expresses an auxiliary verb, which indicates the level of politeness, and &quot;darou&quot; expresses an auxiliary verb, which indicates the speaker's supposition.</Paragraph> <Paragraph position="3"> The checking and correcting method is explained here, using the conditional clause &quot;((anata wa ryohkin wo shiharau masu) baai) \[you TOPIC fee OBJECT pay POLITE CONDITION\].&quot; First, the process checks on limitations for conditional clauses by referring to Table 1, so it understands that neither &quot;wa&quot; nor &quot;masu&quot; can appear in a conditional clause(X baai). Second, the process analyzes the clause, so it understands that the case of &quot;anata\[you\]&quot; is &quot;ga&quot;, and &quot;masu&quot; can be deleted. Finally, the process gets the right conditional clause &quot;((anata ga ryohkin wo shiharau) baai)&quot;</Paragraph> </Section> <Section position="2" start_page="196" end_page="196" type="sub_section"> <SectionTitle> 3.2 Adjustment </SectionTitle> <Paragraph position="0"> A sentence that is only grammatically appropriate, is not as natural as a colloquial sentence. The adjustment module refines the sentence by changing, 1 \[wl ... w,\] is the list of corresponding English words. Uppercase shows the meaning of a function word.</Paragraph> <Paragraph position="1"> Table h limitations on Japanese clauses example clause topic polite supposition &quot;wa&quot; &quot;masu&quot; &quot;darou&quot; agent recipient example for okuru\[send\] -- hearer o-okuri-suru hearer speaker okut-tekudasaru hearer -- o-okurini-naru adding, or deleting words. This module handles honorific expressions and redundant personal pronouns, which are important for generating natural Japanese sentences.</Paragraph> <Paragraph position="2"> Personal pronouns are usually redundant in Japanese conversations, because honorific expressions and modality expressions limit the agent of the action. The procedure is as follow: 1) Change the verb into the appropriate form based on the agent or the recipient, 2) Delete the redundant personal pronouns based on the verb form, or the modal, 3) Generate a final output by adjusting the morphological feature.</Paragraph> <Paragraph position="3"> This method is explained here, using the sentence &quot;((anata wa watashi ni youshi wo okuru masu) ka) \[you TOPIC I OBJECT form OBJECT send POLITE INTERROGATIVE\].&quot; First, the process changes the verb &quot;okuru&quot; into &quot;okut-tekudasaru&quot; by referring to Table 2. Second, it deletes the redundant pronouns &quot;anata wa&quot; and &quot;watashi ni&quot; . Finally, it generates the sentence &quot;youshi wo okuttekudasai masu ka \[form OBJECT send-RESPECT</Paragraph> </Section> </Section> <Section position="6" start_page="196" end_page="196" type="metho"> <SectionTitle> POLITE INTERROGATIVE\].&quot; 4 Evaluation </SectionTitle> <Paragraph position="0"> The prototype system was evaluated by using model conversations between an applicant and a secretary about conference registration. The model conversations consist of 607 sentences, and cover basic expressions. The system provided an average translation time of 2 seconds for sentences with an average length of 10 words, and produced a translation result for all of the sentences. 480 of the results were as natural as colloquial sentences and giving a success rate of 79%.</Paragraph> </Section> class="xml-element"></Paper>