Root element of each document that comprises the Dataset. The Dataset is comprised of multiple InkML and UPX documents organized in a directory structure, with perhaps common information stored in a separate UPX document and referred to by other UPX documents. Specifies metadata related to the dataset as a whole, a high-level characterization of the Dataset. Definitions block containing information about writers, sources of annotation and annotation hierarchies referred to in this document [OPEN: Model for sharing definitions across Dataset] Detailed labeling of digital ink traces, organized in a customizable hierarchy. Gestures and other notations can also be accommodated in the same structure. Version of schema (format) used in this document, of the form X.Y, where X is the major version and Y the minor version. Backward compatability between major versions is not guaranteed. Name of the dataset. Used to refer to the dataset, for example the "IUF-train_r01_v07" collection. High-level category of dataset similar to InkML contentCategory. Examples: 1. text/language [/script][/sub-category] (e.g., text/jpn/Kanji, text/en/ssn) 2. drawing [/sub-category] Sub-category can be similar to Unipen categorization (numerals, isolated characters, words etc) or correspond to semantic categories such as SSN. [OPEN: Standardization of syntax. Also how to handle documents wherein multiple scripts or in general multiple (sub)categories are represented] Dataset version number of the form X.Y, where X is the major version and Y the minor version. Date of publication of dataset in format YYYY-MM-DD Contact info for dataset related queries, such as name, email, address etc. Name of the organization where the data came from Description of data collection recording setup and conditions [OPEN: Are there standard bits of information we would like to capture here, and should we have corresponding subelements ?] Reference to inkSource element from InkML, describing the device used for capture. Nature and structure of the data as a whole. The dataInfo element gives details about the data such as quality of writing, groundTruth of what is captured, the data collection methodology,etc. Reference to datasetInfo element defined elsewhere (e.g., in a common location). This is overridden by any information provided as sub-elements. Identifier of datasetInfo element that can be used in references Unique identifier for this dataset, derived from its name. General top-level description of content of dataset, including information about script(s) used, writing styles, quality of writing. This is only meant to provide a brief consolidated picture of the data in the dataset, and is not a substitute for detailed annotation of ink. Number of writers contributing data Overall quality of the handwriting data captured in the dataset Overall writing style of the handwriting data captured in the dataset, an enumeration of discrete, cursive,mixed, unknown. In addition, style labels may be specified as annotation of ink. Contains a verbatim transcription of the reference text. Alternatively, Describes the corpus/source from where the data-list was created. e.g. TDIL corpus, EMILLE corpus, manually generated data-list. Reference to file containing transcription of the referenced text. Brief description of groundTruth. Profiles of writers who contributed data to the dataset Reference to block of writer definitions stored elsewhere. In general, writer information is collected at the time of data collection and stored in a common place. Identifier for this block of writer definitons (may be useful for defining groups of writers) Profiles of sources of labels (human or machine) used for annotating the dataset Definitions of various annotation schemes used in the dataset [OPEN: Would we need more than one ?] Information about each writer whose writing samples appear in the dataset Personal profile of writer (relevant to writing) Familiarity with different types of capture devices used (Possible values: Poor, Average, Good) Type of device, e.g. TabletPC, PDA used for data capture, used to retrieve the skill of the user on a specific device [OPEN: Should we have an explicit reference to specific InkML captureDevice ?] Familiarity with different scripts featured in data collection Familiarity of the writer with drawing Familiarity with entering music scores Familiarity with mathematics notation Familiarity with chemistry notation The unique identifier of this writer definition. The identifier of this writer (typically, unique within a dataset). Handedness: R for Right Handed and L for Left Handed Highest level of formal education Writer gender Profession of the writer [OPEN: Should this be a closed list such as standard occupation codes ?] Native region of the writer. The convention is Country/State/[City,Town,Village] Date of birth of the writer in format YYYY[-MM[-DD]?]? Examples: 2006-09-20, 2002, 1986-12. The age of the writer in duration format. Examples: P24Y (24 years), P12Y2M12D (12 years, 2 months and 12 days), P453M (453 months), P243DT9H26M12S (243 days, 9 hours, 26 minutes and 12 seconds), or PT315784938472610300S (315784938472610300 seconds). A date that can be restricted to only year, years and months, or years, months, and days. Examples: 2006-09-20, 2002, 1986-12. Style of writing [OPEN: Are more styles needed ?] Frequency of the usage of the script over last year Level of skill of the writer in writing particular script Whether native writer of this script Name of the script [OPEN: Naming convention such as ISO-15924] Whether native writer of this script Frequency of the usage of drawing. Level of skill of the writer in drawing. Contains descriptions of annotators (human or machine) of the Dataset. Each labelSrc element captures information regarding the origin of a particular set of labels, type of labels or other meta-information associated with annotations contained in the dataset. Reference to block of label source definitions stored elsewhere. In general, information about label sources is stored in a common place. This is overridden by any information provided as subelements. Identifier for this block of label source definitions Name of label source Organization that label source represents Contact information for label source: Name, email, etc General description of the label source (more suitable fo rmachine sources) Identifier for this label source, refered to by label Type of annotator: machine or human Type of label provided by label source [OPEN: Should this be a closed list ?] Encoding used for labels of this type (e.g. UNICODE, ISCII, etc.) Description of annotation scheme Reference to block of annotation definitions stored elsewhere. In general, annotation hierarchies are defined once and stored in a common place. Identifier for this block of annotation definitions Description of levels of the annotation hierarchy, now supports heterogenous hierarchies, wherein each node can have children of different types.[OPEN: Do we require to explicitly specify parent-child relationships?] Identifier of the annotationScheme hLevel specifies a collection of ink traces at a certain hierarchy level specified in the corresponding annotationScheme. Each hLevel can refer to a different annotation scheme, uiInfo/imgInfo, writerDef,.. for even more flexibility. Description of the writing area or field used to capture ink. All measurements are in writing area's resolution [OPEN: Completely OPEN !] Background image information. Description of hwData block Reference to the annotation scheme used Identifier for hwData block Labels describing the ink in this block. Each label is of a specific type (e.g. segm, truth, quality, etc) and from a specific label source defined earlier. Labels are in general overridden by labels at lower levels (sub-blocks). A label of type "segm" refers to the block structure (itself the result of segmentation/annotation). [OPEN: Alternative ways of representing alternative block structures from different label sources] Description of the writing area or field used to capture ink. All measurements are in writing area's resolution [OPEN: Completely OPEN !] Background image information. Leaf-level references to digital ink traces (list of InkML tracerefs) Identifier for block Reference to hLevel defined elsewhere. [NOTE: In order for parsers/validators to extract information regarding, e.g., writer, dataset definitions, or annotation scheme, the referenced hLevel context should be considered.] Reference to writer of ink in block (overridden by writer references at lower levels) Type corresponding to basic unit of annotation Alternative label values with rank and score. Alternates can be generated from a recognizer (in semi-automated labeling) or can be specified manually specifies the rank of the annotation Refers to machine annotation. A measure of the confidence of the annotator on the annotation. Unique ID for label Reference to label source defined previously Category of label, e.g. segm, truth, quality, stype. Must be one of the categories of labels generated by this label source. Timestamp of annotation event. Useful for generating history of annotation. The format is UTC The uiInfo element is used to capture the User Interface information of the data collection software. The xOrigin and yOrigin Positions of horizontal guidelines, such as a baseline, topline etc (each position separated by a comma) Position of vertical guidelines such as box boundaries, comb lines ( positions separated by commas) X-coordinate of top-left corner of the writing field, in ink coordinate system Y-coordinate of top-left corner of the writing field, in ink coordinate system. Width of field (using the resolution of the input device, not that of the display) Height of field (using the resolution of the input device, not that of the display) Defines a region of interest (roi) that contains the corresponding ink traces.[OPEN: Do we need attributes like originX, originY, orientationX, orientationY ? This may be required for processors to map ink trajectories on image regions.] Defines a sequence of operations performed on the image before it was displayed to the user. Images may be cropped, rotated, or scaled. An identifier for the image information contained in this element. A URI to the image source data. A reference to the label source, i.e., the program or person who makes the link between online trajectories and roi information. An identifier for the image information contained in this element. The points defining the roi-polygon. The bounding box of the crop process in image coordinates. The scaling factor in the x coordinates of the image. If this attribute is not present, a default value of 1 will be assumed. The scaling factor in the y coordinates of the image. If this attribute is not present, a default value of 1 will be assumed. The rotation angle (anti-clockwise) in radians. The pivot (rotation point) in image coordinates, If this attribute is not present a pivot at (0,0) is assumed An identifier for the image preprocessing sequence contained in this element. A two dimensional coordinate. Examples: "0 0", "-12.32 23.11". Defines a list of coordinates defining a polygon. Examples "0 0 5.2 8.2 9.3 3.5 5.8 -2.4". List of label types (e.g. segm, truth, style, quality, script, etc.) produced by label source