<?xml version="1.0"?><!DOCTYPE article SYSTEM "/project/take/software/searchbench_offline_processing/paperxml_generator/aclextractor/src/python/../resource/dtd/paperxml.dtd"><article><header><firstpageheader><page local="1" global="24"/><title>The Standard of Chinese Corpus Metadata</title><author surname="Tingting" givenname="He"><org  name="Huazhong Normal University" country="China" city="Wuhan"/></author><author surname="Xiaoqi" givenname="Xu"><org  name="Huazhong Normal University" country="China" city="Wuhan"/></author></firstpageheader><frontmatter><p><b>The Standard of Chinese Corpus Metadata*</b></p><p><b>He Tingting</b></p><p>Huazhong Normal University</p><p>tthe@mail.ccnu.edu.cn</p></frontmatter><abstract><b>The normalization of corpus metadata plays a key role in building sharable corpora. However, there is no uniform specification for defining and process­ing metadata in Chinese corpus nowa­days. This paper introduces a metadata system we've proposed for Chinese corpus. 46 elements are de­fined in all, which can be divided into 6 classes: information about copyright, information about background of lin­guistic material creator, information about medium of linguistic material, information about the content of lin­guistic material, information about collecting linguistic material, and in­formation about management of lin­guistic material. To distinguish one element from another, or our elements from someone else's, we provide a po­tent description method, where 10 sub­sections are designed to describe the detailed properties for each element.</b> </abstract></header><body><section number="1" title="Introduction"><p>"Metadata" is first defined in computer science. It plays an important role in the management of electronic resources, especially the huge infor­mation from Internet. By cataloguing the web pages, we can obtain a better search more effi­ciently. Nowadays, metadata becomes a popular tool to describe administrative information about all kinds of resources. It defines schemes for resource description, and also provides universal mechanism for resource retrieval.</p><p>In corpus linguistic, metadata description has existed for a long time, and is generally referred <b>Xu Xiaoqi</b> to heading information. By defining metadata, more accurate and profuse annotation contents can be provided for corpus, such as, information about time, area, author and etc. However, there is no uniform specification for processing meta­data in Chinese corpus at present. Thus, we de­fine a core metadata set for Chinese corpus and normalize the description of set element. Basing on the Dublin Core metadata, which is widely accepted in philology, the definition takes much attention on the linguistic characteristics of Chi­nese corpus, and is compatible to the OLAC metadata standards as well. Both creator and users of the corpus can get regulations of textual description and annotation strategy from this standard.</p><footnote>Xu_xiaoqi@hotmail.com</footnote><p>In section 2, we discuss some referenced standards and resources, including DC and OLAC metadata. Section 3 presents a frame­work within which we design our metadata, and lists the main problems to be solved. Section 4 summarizes our metadata description and re­ports some further development of the standard. Conclusion is drawn in section 5.</p></section><section number="2" title="Related metadata resources"><subsection number="2.1" title="Dublin Core metadata"><p>Dublin Core Metadata has been present in OCLC / NCSA ( National Center for Super­computer Applications ) Meta Workshop in 1995.It's a standard for cross-domain informa­tion resource description, and has no fundamen­tal restrictions to the types of resources to which the metadata can be assigned. DC metadata de­fined 15 core elements, which are maintained and managed by DCMI (Dublin Core Metadata Initiative). The core elements are listed in table 1.</p><p>* The paper is supported by National Language Application Project of Tenth five-years plan of China, (Grant No. ZDI105-B01, ZDI105-43B);</p><p>National Natural Science Foundation of China (NSFC), (Grant No.10071028);</p><p>Ministry of education of China, Research Project for Science and technology, (Grant No. 105117).</p><page local="2" global="25"/><p>In DC metadata, each element is described in 10 property items that defined in ISO/IEC 11179.They are: "Name", "Identifier", "Ver­sion", "Registration Authority", "Language", "Definition", "Obligation", "Datatype", "Maxi­mum Occurrence" and "Comment". However, 6 items among them have settled value for each element as following: Version:1.1</p><p>Registration Authority: Dublin Core Meta­data Initiative Language:en Obligation: Optional Datatype: Character String Maximum Occurrence: Unlimited</p><p>DC metadata is an important reference for the definition of Chinese corpus metadata. There are at least two reasons for this.</p><p><b>(1) </b>Both DC and corpus metadata are designed for large-scale users, who are not always pro­fessional catalogue person. Thus apprehensi­ble and general are two pivotal aims to achieve.</p><p>(2) DC metadata has been mostly assigned to electronic text from Internet webs, which are primary source of linguistic material as well. Therefore, it's expected that the corpus can be used directly without reannotation if they are annotated with DC metadata before.</p></subsection><subsection number="2.2" title="OLAC metadata"><p>The OLAC ( Open Language Archives Com­munity Metadata) metadata set is based on the Dublin Core metadata set. In order to meet the specific needs of the language archiving com­munity, the OLAC metadata set qualifies with three kinds of qualification: element refinement, encoding scheme, and content language. With these three attributes, an element in OLAC can indicate more information than the same one in</p><p>DC does. Take the element "Date" in OLAC for example, with the element refinement, it can represent either date of create, or date of issue, or date of modification in different occasions</p><p>The elements in OLAC are listed in table 2,and we can see that it uses all the 15 elements in DC. Element in OLAC are described in 5 property items which are "Name", "Definition", "Comments", " Attributes" and " Examples".</p><p>Genre= Prose Style= narrative Mode= Written Topics= Literature Medium= Textbook Name= Sex=</p><p>Nationality^</p><p>Language=Chinese</p><p>Publish House= National Institute for</p><p>Compilation and Translation Publish Place=Taiwan Publish Data= Title= starlight <b>Research data</b> <b>on large-scale corpus meta-</b></p><table caption="Table 3 Example of metadata describing in Sinica corpus"></table><doubt alpha="0.0" length="3" tooSmall="False" monospace="0.0">2.3</doubt><subsubsection number="2.3.1" title="Sinica corpus metadata"><p>Sinica corpus is developed and maintained by Institute of Information Science and CKIP group in Academia Sinica. It's designed for analyzing modern Chinese. Texts are collected from dif­ferent areas and classified according to five cri­teria: genre, style, mode, topic, and source.</p><table caption="Table 2 Elements in OLAC, this set uses all fifteen elements in DC." class="main" frame="box" rules="all" border="1" regular="False"><tr class="row"><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>Elements about</b></p></td><td class="cell"><p><b>Elements</b></p></td><td class="cell"><p><b>Elements about</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>Resource Con-</b></p></td><td class="cell"><p><b>about</b></p></td><td class="cell"><p><b>External At-</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>tent</b></p></td><td class="cell"><p><b>Copyright</b></p></td><td class="cell"><p><b>tribute descrip­tion</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Title</p></td><td class="cell"><p>Creator</p></td><td class="cell"><p>Date</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Coverage</p></td><td class="cell"><p>Publisher</p></td><td class="cell"><p>Identifier</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Description</p></td><td class="cell"><p>Contributor</p></td><td class="cell"><p>Format</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Language</p></td><td class="cell"><p>Rights</p></td><td class="cell"><p>Format.cpu</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Source</p></td><td class="cell"><p></p></td><td class="cell"><p>Format.markup</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Relation</p></td><td class="cell"><p></p></td><td class="cell"><p>Format.os</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Subject</p></td><td class="cell"><p></p></td><td class="cell"><p>Format.sourcecode</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Subject.language</p></td><td class="cell"><p></p></td><td class="cell"><p>Format. encoding</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Type</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Type.data</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Type.function</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td></tr></table><table caption="Table 1 Fifteen core elements in DC, which are divided into 3 classes." class="main" frame="box" rules="all" border="1" regular="False"><tr class="row"><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>Elements</b></p></td><td class="cell"><p><b>Elements</b></p></td><td class="cell"><p><b>Elements</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>about Re-</b></p></td><td class="cell"><p><b>about Copy-</b></p></td><td class="cell"><p><b>about External</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>source Con-</b></p></td><td class="cell"><p><b>right</b></p></td><td class="cell"><p><b>Attribute de-</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>tent</b></p></td><td class="cell"><p></p></td><td class="cell"><p><b>scription</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Title</p></td><td class="cell"><p>Creator</p></td><td class="cell"><p>Date</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Subject</p></td><td class="cell"><p>Publisher</p></td><td class="cell"><p>Type</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Description</p></td><td class="cell"><p>Contributor</p></td><td class="cell"><p>Format</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Language</p></td><td class="cell"><p>Rights</p></td><td class="cell"><p>Identifier</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Source</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Relation</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Coverage</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td></tr></table><page local="3" global="26"/><p>Therefore, this corpus is a representative sample of modern Chinese language. Metadata in Sinica corpus lays special emphasis on describing the linguistics information of lin­guistic material, such as "Mode", "Style", "Me­dium", and "Topic". An example of metadata describing in Sinica corpus is given in table 3.</p></subsubsection><subsubsection number="2.3.2" title="National modern Chinese corpus meta­data"><p>National modern Chinese corpus is the largest balance corpus in China at present. The selec­tion of linguistic material follows the principles of commonality, description and practicability. In order to reflect the panorama of modern Chi­nese, a lot of work has been done on designing balance gene. And the finally selected samples have a wide span on time, domain and medium.</p><p>Metadata in National modern Chinese corpus pay much attention on copyright information and publish information of linguistic material. Furthermore, both a global serial number and a category number are designed to identify a cer­tain sample.</p></subsubsection><subsubsection number="2.3.3" title="BNC metadata"><p>The British National Corpus (BNC) is a 100 million word collection of samples of written and spoken language from a wide range of sources, designed to represent a wide cross-section of current British English, both spoken and written. Each text in BNC has a TEI header to indicate the identification and classification of individual text, special details such as speakers', and the housekeeping information. The defini­tion of text classification is meticulous. For spo­ken text material, age, sex, and class of respondent are all make sense as well as the do­main, region and type of the content. And for written text material classification, age, sex, type of author, audience, circulation, status, medium, and domain are laid emphasis on. However, some classification were still poorly defined and partially populated, such a "dating"(date of copy or date of first publication?) and "domain" (has something different with text-type?).</p></subsubsection><subsubsection number="2.3.4" title="Metadata in balanced corpus"><p>In recent years, the awareness that text is not just text, but that texts comes in several forms, has spread from more theoretical and literary subfields of linguistics to the more practically oriented information retrieval and natural lan­guage processing fields. As a consequence, sev­eral test  collections  available  for research explicitly attempt to cover many or most well-established textual genres, or functional styles in well-balanced proportions</p><p>In practice, choosing balance gene is a profes­sional work that needs a scientific programming strategy. Sinclair suggested a minimum set of balance gene for general corpus in 1991 that indicates a popular classify principle for linguis­tic: the style of linguistic (on-the-spot record or literature); the form of linguistic (formal or in­formal); the medium type of linguistic (from book or magazine or paper); and the age, sex of the author. From the Sinica corpus and National modern Chinese corpus we've discussed above, we can see that the gene of time, style, area and subject are most in frequent use, which become our crucial reference for metadata designing.</p></subsubsection></subsection><subsection number="2.4" title="ISO1179 standard"><p>ISO1179 is an international standard about de­veloping metadata. There are 6 parts in this standard, which are considered as our basic rule to follow.</p></subsection></section><section number="3" title="Framework of metadata description"><p>We describe metadata information from three aspects, which we consider as: content structure, syntax structure and semantic structure. Content structure is used to decide the elements in a metadata set. Syntax structure introduces a model or syntax to represent metadata, while semantic structure declares the signification concourse of elements.</p><p><b>Metadata description</b></p><p><b>FIG.1.The framework of develop metadata</b></p><p>A consistent strategy is essential when these three structures are used to define metadata. Our research is to solve three problems especially.</p><subsection number="3.1" title="Element selection"><p>Elements in metadata set are used to describe a resource from different aspects. Thus, the selec­tion or designing of elements becomes an impor­tant issue. When the selection depends on the experience of corpus creator rather than a nor­mative rule, it's hard for the metadata to assert the resource sufficiently.<page local="4" global="27"/></p><p>We referenced a lot from DC and OLAC metadata. For the universal use of these two metadata standards and the similarity between DC metadata and corpus metadata we discuss above, we finally used all the fifteen elements defined in DC standard. However, some ele­ments are refined or splitted into several new elements on the basis of the old definition. For example, the elements "Date" is extended as "Indite Date", "Issued Date", "Created Date" and "Modified Date", thus more detailed and definite information of date can be obtained for either a single sample or the whole corpus. And the same case for the element "Language" from DC. We defined three kinds of information about language to describe both creator informa­tion and content information.</p><p>To fully consider the linguistic characteristics of Chinese corpus, we've introduced several popular metadata elements in balanced corpus, such as style, mode, medium and so on, which are also important balance gene for corpus de­signing.</p><p>Therefore, we define 46 elements in all, which can be divided into 6 classes. They are: information about copyright, information about background of linguistic material creator, infor­mation about medium of linguistic material, in­formation about the content of linguistic material, information about collecting linguistic material, and information about management of linguistic material. Most elements we defined are intellectual metadata, while some structural metadata, access control metadata and critical metadata are included as well.</p></subsection><subsection number="3.2" title="Description field"><p>Metadata is structured data about data. It's usu­ally expressed with several property fields or subsections, which is regarded as its own data structure or syntax. Different metadata system may use different way to describe and naming its elements, thus it's hard for metadata com­munion or understanding the same element from two dissimilar systems.</p><p>A unified method for description is helpful, and it's expected to be succinct, general and dis­tinguished. Our standard has provided 10 fields for a metadata description. Some are obligatory, that is to say you must give a value to such fields in order to confirm an element. And some are optional for the individuation use. This seems to do a better work than DC, while 6 fields in it always have settled value. We spe­cially introduce two subsections for naming an element, thus elements can be distinguished ex­actly from either "Name" or "Long Name" field. We have exhibited such format in XML (eXten­sible Markup Language), and created the DTD (Document Type Definition) file for it as well.</p></subsection><subsection number="3.3" title="Semantic description"><p>Semantic structure defines the detailed value of metadata, and finally affirms how to use it. Value land should be carefully considered to avoid confusion use. Many famous metadata standards have formed a maturity definition of elements. For example, DC use ISO 8601 to de­fine element "date", Dublin Core Types for ele­ment "Resource Type" and URL or ISBN to define element "Identifier". We took much ac­count of the linguistic characteristics of Chinese corpus and some value are assigned refer to lin­guistics literature.</p></subsection></section><section number="4" title="Metadata standard"><subsection number="4.1" title="Element set"><p>Our corpus metadata set is based on the Dublin Core metadata set and uses all fifteen elements defined in that standard. We've summarized some annotation items in other large-scale cor­pus and developed an element set listed in table 4.There are 46 elements in all. They are ex­pected to describe the resources from six aspects.</p><subsubsection number="4.1.1" title="Information about copyright"><p>The intellectual property right of corpus is copy­right. According to the copyright law, corpus must show clearly its copyright information when being published or promulgated. Metadata in this class is about corpus' created or issued information, mainly including:</p><p>■ Title: the title of original linguistic material, such as books, articles, webs and so on.</p><p>■ Source Identifier: the tag of source linguis­tic material, such as ISBN for books and URL for webs.</p><p>■ Indite Date: is used to describe the writing time of original linguistic material, or the</p><page local="5" global="28"/><p>recording time of the oral linguistic mate­rial.</p><p>■ Issued Date: describe the publish time of a given linguistic material.</p><p>■ Copyright: show the composer, publishing company or the web site of the original lin­guistic material.</p><p>■ Resource Type: resource's physical type can be various, such as papery, electronic, recordy, or kinescope.</p><p><b>4.1.2 Information about background of lin­guistic material creator</b></p><p>We pay some attention on the individual infor­mation of corpus' creator, because it's helpful for analyzing linguistics characteristic about corpus. Such information includes native lan­guage, born place, age, sex as well as creator's name. For corpus' creator is not always a single person, we define "Agent Type" to clarify such instance, and introduce other creators in "Con­tributor".</p></subsubsection><subsubsection number="4.1.3" title="Information about medium of linguistic material"><p>Information about medium of linguistic material provides detailed data of publish region, influ­ence area, circulation extent and so on, which are all important to evaluate the corpus' balance gene.</p><p>■ Medium Type: linguistic material is usually selected from different published medium including paper, book, magazine, web or else.</p><p>■ Publish Type and Publish Area: respec­tively indicate the geographical area type or size, such as national or local, and the idio-graphic cantons the area covers.</p><p>■ Publish Period and Amount: respectively show the publish frequency and copies of the publication.</p></subsubsection><subsubsection number="4.1.4" title="Information about the content of lin­guistic material"><p>Information about the content of linguistic mate­rial describe corpus from the point of view of linguistics, such as mode and style. And other elements in this class focus on two things, that is what the material expressed and how it ex­pressed. For example:</p><p>■ Subject: is used to express the theme of the linguistic material, while "Description" gives some further detail of what is talk about.</p><p>■ Markup Language: is especially defined to indicate the coding language of electronic resources.</p></subsubsection><subsubsection number="4.1.5" title="Information about collecting linguistic material"><p>Corpus is not a simple set of corpus. When se­lect linguistic material, many factors are consid­ered. We discuss the information about collecting linguistic material in written corpus and oral corpus respectively.</p><p>In written corpus, elements mostly describe the information of material sample, such as how to abstract the sample or how long the sample linguistic material should be. Oral corpus has its particular way to collect materials, so we de­scribe them from the scene character of the interlocution.</p></subsubsection><subsubsection number="4.1.6" title="Information about management of lin­guistic material"><p>Information about management of linguistic ma­terial record data for corpus management and further-processing. Most elements are designed for system administrator and it's recommended that the data is user- sightless. Such as Tag in­formation of linguistic material:</p><p>■ Identifier: defined for system to identify each linguistic material from this unique identifier.</p><p>■ Sample Name: the material title in the cor­pus .It can either be the original title of the material, or new name the corpus' creator gives afterward if it has.</p><p>Log information of linguistic material process­ing are defined for corpus updating and backup, such as input type, annotate type, create date and modified date. And system information, such as operation system (format.os) and CPU (for-mat.cpu) are defined to describe the running en­vironment of the corpus.</p><page local="6" global="29"/><p><b>Table.4.We define 46 elements in all. They are expected to describe the resources from six aspects.</b></p></subsubsection></subsection><subsection number="4.2" title="Subsections"><p>To distinguish one element from another, or our elements from someone else's, we provide a potent description method. Ten subsections are defined as mutual attribute field. Each metadata element can be described with these subsections selectively or whole.</p><doubt alpha="40.0" length="10" tooSmall="False" monospace="0.0">4.2.1 Name</doubt><p>Unique for each element. Used as identifier when preserve data. Name is a sting of English letter.</p><subsubsection number="4.2.2" title="Long Name"><p>Displayed as full name in Chinese.</p></subsubsection><subsubsection number="4.2.3" title="Definition"><p>Semantic content of an element.</p></subsubsection><subsubsection number="4.2.4" title="Comments"><p>Extra or special explanations are put in com­ments.</p></subsubsection><subsubsection number="4.2.5" title="Value Land"><p>Specify the possible value land of metadata.</p><doubt alpha="40.0" length="10" tooSmall="False" monospace="0.0">4.2.6 Type</doubt><p>A "Type" subsegment can be either "basic ele­ment" or "file citation", while "file citation" denotes that the element's definition has referenced some content from another file.</p></subsubsection><subsubsection number="4.2.7" title="DefinedIn"><p>Indicate where the metadata has been ever de­fined. It may be from DC, OLAC or user-defined.</p></subsubsection><subsubsection number="4.2.8" title="Obligation"><p>Elements could be either obligatory or optional. When a metadata is obligatory, it must be used in corpus.</p></subsubsection><subsubsection number="4.2.9" title="Publish Date"><p>Indicate the publish date of the metadata</p></subsubsection><subsubsection number="4.2.10" title="Publish File"><p>Indicate the name of file in which the metadata first defined.</p></subsubsection></subsection><subsection number="4.3" title="Metadata description"><p>The way to describe a metadata element is to assign semantic content to each subsection we defined before.<page local="7" global="30"/> There are 46 elements in all, and we give description of two elements to show the model.</p><table class="main" frame="box" rules="all" border="1" regular="False"><tr class="row"><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>information about copyright</b></p></td><td class="cell"><p><b>information</b></p><p><b>about background of linguistic material creator</b></p></td><td class="cell"><p><b>information about me­dium of linguistic material</b></p></td><td class="cell"><p><b>information about the content of linguistic material</b></p></td><td class="cell"><p><b>information about col­lecting linguistic material</b></p></td><td class="cell"><p><b>information about man­agement of linguistic ma­terial</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p><b>Written corpus</b></p></td><td class="cell"><p><b>Oral corpus</b></p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Title</p></td><td class="cell"><p>Agent Type</p></td><td class="cell"><p>Medium Type</p></td><td class="cell"><p>Mode</p></td><td class="cell"><p>Abstraction Type</p></td><td class="cell"><p>Environment</p></td><td class="cell"><p>Identifier</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Source Iden­tifier</p></td><td class="cell"><p>Creator</p></td><td class="cell"><p>Publish Type</p></td><td class="cell"><p>Style</p></td><td class="cell"><p>Position</p></td><td class="cell"><p>Event</p></td><td class="cell"><p>Sample Name</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Indite Date</p></td><td class="cell"><p>Sex</p></td><td class="cell"><p>Publish Area</p></td><td class="cell"><p>Subject</p></td><td class="cell"><p>Words Amount of Resource</p></td><td class="cell"><p>Place</p></td><td class="cell"><p>Input Type</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Issued Date</p></td><td class="cell"><p>Native Place</p></td><td class="cell"><p>Amount</p></td><td class="cell"><p>Description</p></td><td class="cell"><p>Words Amount of Sample</p></td><td class="cell"><p></p></td><td class="cell"><p>Annotate Type</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Copyright</p></td><td class="cell"><p>Native Lan­guage</p></td><td class="cell"><p>Publish Pe­riod</p></td><td class="cell"><p>Language</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Software</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p>Resource Type</p></td><td class="cell"><p>Age</p></td><td class="cell"><p></p></td><td class="cell"><p>Markup Lan­guage</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Create Date</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p>Contributor</p></td><td class="cell"><p></p></td><td class="cell"><p>Relation</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Modified Date</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Source</p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Copyright after Annotation</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Description of Copyright</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Limitation</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Format</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Format.cpu</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p></p></td><td class="cell"><p>Format.os</p></td><td class="cell"></td></tr><tr class="row"><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td><td class="cell"></td></tr></table><p><b>Descr </b>ipt ion 1:</p><p><b>Name: </b>Agent Type <b>Long Name: â'J</b></p><p><b>Definition: </b>The way in which corpus' crea­tor organized.</p><p><b>Comments: </b>The creator of a corpus may be one person, several persons or an organization.</p><p><b>Value Land: </b>Defined as {sole, multiple, corporate, unknown, unclassified } according to the design of BNC corpus. "unknown" represent this property has not been obtained, while "un­classified" indicate the organized form has not been defined in our value land (equivalent to NULL in relation database).</p><p><b>Type: </b>basic element <b>DefinedIn:</b><b> </b>user-defined <b>Obligation:</b><b> </b>commendatory <b>Publish Date:</b><b> </b>2005.1.</p><p><b>Publish File: </b>Standard of corpus metadata Descript ion 2:</p><p><b>Name: </b>Mode <b>Long Name:</b><b> </b>ig- f$ <b>Definition:</b><b> </b>Type of writing <b>Comments:</b><b></b></p><p><b>Value Land: </b>{kouyupingshi, kouyuyishu, shumianpingshi, shumianyishu} <b>Type: </b>basic element <b>DefinedIn: </b>user-defined <b>Obligation: </b>commendatory <b>Publish File: </b>Standard of corpus metadata</p></subsection><subsection number="4.4" title="Encoding syntax"><doubt alpha="33.3" length="9" tooSmall="False" monospace="0.0">4.4.1 XML</doubt><p>XML is a widely used language for defining data formats. It provides a very rich system to define complex documents and data structures. As long as a programmer has the XML defini­tion for a collection of data (often called a "schema"), they can create a program to process any data format according to those rules. We've defined our metadata format by using XML in several schema files.</p><subsubsection number="4.4.2" title="Example of schema"><p>The following codes define the element "Style", which is proposed to describe the style of article material in corpus. We defined four types of mode in our standard: narrative writing, expository writing, argumentative writing and practical writing. In the schema, they are represented as "jixuwen", "shuomingwen", "lunshuwen", and "yingyongwen" respectively.</p><doubt alpha="51.3" length="39" tooSmall="False" monospace="0.0">&lt;?xml version="1.0" encoding="GB2312"?&gt;</doubt><doubt alpha="85.7" length="7" tooSmall="False" monospace="0.0">&lt;schema</doubt><p>xmlns="http ://www.w3 .org/2001/XMLSchema"&gt; &lt;annotation&gt; &lt;documentation&gt;</p><p>CMD Schema for Article Style, seiga, 1/7/05 &lt;/documentation&gt; &lt;/annotation&gt;</p><p>&lt;simpleType name="CMD-Style-Code"&gt; &lt;restriction base="string"&gt; &lt;enumeration value="jixuwen"/&gt; &lt;enumeration value="shuomingwen"/&gt; &lt;enumeration value="lunshuwen"/&gt; &lt;enumeration value="yingyongwen"/&gt; &lt;/restriction&gt; &lt;/simpleType&gt; &lt;/schema&gt; <b>Code1.</b><b>define element "Mode" in a schema 5 Conclusion</b></p><p>Metadata is "data about data". In our norm, a lot work has been done to describe such data. We normalized data item, naming regulation, data type and data width.46 metadata elements have been defined to register information of resource, within which 15 belong to DC Metadata, 3 be­long to OLAC Metadata, 15 belong to both DC and OLAC Metadata, and 28 are user-defined. According to this, we tag each metadata by its "DefinedIn" item. Corpus designers are able to choose element during the annotation. They can also add new elements to satisfy various re­quirements basing on this standard.</p><p>The standard use English string when de­nominating metadata element, because some software cannot support Chinese variable. We develop DTD files and an assistant software (FIG.2-FIG.4) for the convenient of corpus an­notation. By filling blanks with some metadata information in this software, users can directly get the XML code of an annotated corpus. We are to do some further experiment on corpus annotation and corpus management. With vari­ous information the metadata interpreted, more works may lead to resources discovery and con­tent rating.</p><page local="8" global="31"/><p>[6]What is BNC.</p><p>http://www.hcn.ox.ac.uk/BNC/what/index.html<u> </u>[7]XML http://www.w3<u> .org/TR/RE-XML </u>[8]Sun xiaofei. XML and modern digital library. Modern books information,2000,(4)</p><p>[9]Cui gang,Sheng yongmei. Annotion of corpus. Pekin : Tsinghua. University press, 2000, 15(1) <b>FIG.</b><b>2Interface to input the information about copyright, background of linguistic material creator and medium of linguistic material</b> <b>FIG.</b><b>3Interface to input information about the content of linguistic material and collect­<u>ing linguistic material</u></b> <b>FIG.</b><b>4 Interface to input information about management of linguistic material</b></p></subsubsection></subsection></section><references><p>[1]He tingting.Study on corpus. Doctoral dissertation of central china normal university 2003.4.</p><p>[2]Dublin Core Metadata Initiative.</p><footnote>http://dublincore.org/index.shtml</footnote><p>[3]OLAC Metadata Set . http://www.languagearchives.org/OLAC/olacms.html.</p><p>[4]International Organization for Standardzation .http://www.iso.org/iso/en/ISOOnline.fromtpage<u>.</u></p><p>[5]The standard and normalization construction of digital library. http://cdls.nstl.gov.cn/</p></references></body></article>