. ]]> The corpus element stores a whole corpus containing of resources (optional), meta data (optional) and corpusData entries (normally texts, other file types including video and audio files are supported as well). Usually, one would use the possibility to use one file containing the corpus root element which consists of several corpusDataRef elements (e.g. pointers to separate corpusData instances). The corpusData element stores all information regarding a single corpus file. Schematron primaryData constraint rule None of the provided primaryData elements is set as master file. The corpusDataRef element is used as a placeholder for a single corpusData entry saved externally. The metaRef element is used as a placeholder for external metadata. The primaryData element contains either one (or more) reference(s) to a normalized (in terms of whitespace) file serving as base data for the annotation, or -- when dealing with shorter texts -- the whole text as a string underneath the textualContent child element. The primaryDataRef element is used to refer to one or externally stored representations of the primary data used during the annotation process. The textualContent element stores the textual primary data. The checksum element can be used to preserve integrity of the primary data when dealing with multiple annotation layers. It is required to supply the algorithm that was used to calculate the checksum. The resources element is a container element for the resources used for segmentation or annotation of the primary data. Resources can be included in an XSF file (via the resource element) or referred to by the resourceRef element. The segmentation element is a container element for the segmentation units (segment, one or more) that are used to identify text spans in the primary data. A segment identifies a (character or time) span over the primary data. The creator attribute may be used for referencing a resource that was used during segmentation, in addition, the alt attribute may be used to mark this segment as an alternative segmentation to other segments. Schematron segment constraint rule End position of a segment must be higher than or equal to its start position. Start position of a segment must be higher than or equal to the start position of the complete primary data. End position of a segment must be lower than or equal to the position of the complete primary data. DiagnosticsError: End position of segment is lower than its start position. Start position: . End position: . DiagnosticsError: Start position of segment is lower than start position of primary data. Start position of segment: . Start position of primary data: . DiagnosticsError: End position of segment is greater than end position of primary data. End position of segment: . End position of primary data: . The resource element contains optional meta data and the resource used for segmentation or annotation. The resourceRef element is used for referencing resources used in the annotation process that are stored externally. The resourceContent contains the content of a resource. The annotation element contains annotation levels (and their respective layer(s)). The level element stores meta information and the actual annotation layer(s). A level refers to the conceptual model of information represented in markup. The import attribute should be used to mark other annotations levels that have to be established before this level can be used (i.e. this annotation level uses information derived from other levels). The meta element can be used to store meta information. Elements derived from other namespaces are allowed, e.g. OLAC metadata is recommended for textual primary data. Note that schema processing is set to lax, i.e. a meta information schema can be provided but is optional. In contrast to the level the layer element refers to the technical realisation (or XML serialisation) of markup. It contains the converted (i.e. standoff) representation of inline annotation layers and their respective elements and attributes. Note that schema processing is set to lax, i.e. an underlying schema (XSD) should be accessible, i.e., all layers should have an corresponding XSD associated, the only exception should be the all layer. The priority attribute can be used to prioritize annotation levels (in case of overlapping markup during an inline markup unification) -- the highest number correspond to the innermost markup. The creator attribute can be used to refer to resources that have been provided in an XSF file or to a user/agent who has made an annotation level. Keep in mind, that its type is NMTOKENS since resources belong to the corpus as a whole, not to a single corpusData file The inline element is used as root element for the outcome of an XStandoff2Inline transformation. The milestone element is modelled after the TEI milestone element and should be used when an XStandoff instance should be serialized into an inline annotation format. The xsf:segment attribute can be used to refer to the segment that was used in the XStandoff representation, the charPos attribute depicts the character position in the primary data where the overlap happened. The type attribute shall be used to specify the segment. Possible values are: 'empty' for empty segments depicting formerly empty inline elements, 'char' for (non-whitespace) character data of the primary data, 'ws' for whitespace character data, 'pun' for punctuation characters, 'dur' when using a duration (i.e. non-character based) segmentation, 'seg' for depicting a segment that is composed via other segments. In the very rare case that you want to store the textual content of a segment, that is the primary data that is annotated by the annotation elements referring to this very segment, use the content attribute. The content can be saved in terms of literal characters or in terms of Unicode codepoints. See the unicodeNormalizationForm attribute in the latter case. Some special characters (such as accented characters) can be represented in Unicode character sets by unique code points (precomposed) or with combining characters (decomposed). Since XSF requires a unique representation, a normalized form of Unicode text shall be used if a Unicode character set is used. Four Unicode Standard normalization forms are available: Normalization Form D (NFD), Normalization Form KD (NFKD), Normalization Form C (NFC), and Normalization Form KC (NFKC). Roughly speaking, NFD and NFKD decompose characters where possible, while NFC and NFKC compose characters where possible. Cf. Chapter 2, 'General Structure', Chapter 3, 'Conformance' and the Annex #15, 'Unicode Normalization Forms,' of The Unicode Standard for details. This information may be used by software processing XStandoff instances. The segments attribute can be used together with the mode attribute to create a new segment via referring to already established segments. The mode attribute marks segments that are created by referring to already established segments (cf. the segments attribute) either as continuous or disjoint. In case of multiple primary data files one primary data should be set as master using the role attribute. The master primary data file specifies the timeline/character stream that is used for creating segments. The unit used for delimiting the segmentation. XSF does not use the built-in data type duration, since it lacks some of the units used here. The offset attribute shall be used for specifying an offset (in the same unit that is used for delimiting segments) between different primaryData files. The URI specifying the location of the referenced file. This attribute is used to differentiate between relative and absolute URIs. Some special characters (such as accented characters) can be represented in Unicode character sets by unique code points (precomposed) or with combining characters (decomposed). Since XSF requires a unique representation, a normalized form of Unicode text shall be used if a Unicode character set is used. Four Unicode Standard normalization forms are available: Normalization Form D (NFD), Normalization Form KD (NFKD), Normalization Form C (NFC), and Normalization Form KC (NFKC). Roughly speaking, NFD and NFKD decompose characters where possible, while NFC and NFKC compose characters where possible. Cf. Chapter 2, 'General Structure', Chapter 3, 'Conformance' and the Annex #15, 'Unicode Normalization Forms,' of The Unicode Standard for details. These information may be used by software processing XStandoff instances. Enumeration of mime-types that can be used for primary data files. This attribute group should be imported into the schema files of the respective layers. It consists of the obligatory segment attribute and the optional dcrCategory attribute. Use this attribute for providing a link to a ISO Data Category Registry category (cf. ISO 12620 for details). The creator attribute may be used for referring to (non-) human agent (e.g. a resource) that was involved in creating an item. The alt attribute may be used for referring to an alternate version, neither in terms of isVersionOf, isBasedOn, isFormatOf or similar relations. The type of relation should be specified in the meta data entry of the respective element bearing the alt attribute.