File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/intro/03/n03-4013_intro.xml
Size: 2,649 bytes
Last Modified: 2025-10-06 14:01:49
<?xml version="1.0" standalone="yes"?> <Paper uid="N03-4013"> <Title>pre-CODIE - Crosslingual On-Demand Information Extraction</Title> <Section position="3" start_page="0" end_page="0" type="intro"> <SectionTitle> 3 System Architecture </SectionTitle> <Paragraph position="0"> Pre-CODIE is implemented as an integration of several modules, as shown in Figure 1: translation, information First, the system takes the narrative description of the scenario of interest as an English query, and the Translation module (off-the-shelf IBM King of Translation system) generates a Japanese query. The IR module retrieves a set of relevant documents from Japanese Mainichi Newspaper from 1995. Then, the Pattern Acquisition module produces a list of extraction patterns from the relevant document set sorted by their relevance to the scenario (Sudo et al., 2001). Pre-CODIE asks the user to assign each placeholder in the patterns to one of the slots in the template. Finally, the Extraction module performs the pattern matching with slot-assigned patterns to each text in the relevant document set and generates a filled Japanese template, which is translated slot-by-slot into English for the user.</Paragraph> <Paragraph position="1"> From the user's point of view, pre-CODIE works as follows with the screenshots in Figure 2.</Paragraph> <Paragraph position="2"> 1. Query: The user types in the narrative description of the scenario of interest, one phrase in &quot;description&quot; text-box and more detail optionally given in &quot;narrative&quot; text-box: &quot;Management Succession: ...&quot;.</Paragraph> <Paragraph position="3"> 2. Configuration: The user adds and/or deletes the slots in the template; Add &quot;Person-In&quot;, &quot;Person-Out&quot;, &quot;Post-In&quot;, &quot;Post-Out&quot;, and &quot;Organization&quot;. 3. Slot Assignment: The user assigns a slot to each placeholder in the pattern by choosing one of the slots defined in step 2; Assign &quot;(be-promoted (PERSON-SBJ))&quot; to &quot;Person-In&quot;.</Paragraph> <Paragraph position="4"> Also, the user can see the example sentences with the match of the pattern highlighted. This will make it easier for the user to understand what each pattern aims to extract.</Paragraph> <Paragraph position="5"> 4. Extraction: The user gets the extracted template and repeats this procedure until the user gets the right template by going back to step 3 to change and/or add slot assignments, and by going back to step 2 to delete and/or add slots in the template.</Paragraph> </Section> class="xml-element"></Paper>