GB2191611A - A man-computer data input technique - Google Patents

Abstract

The technique involves the provision of object labels 3 and reference labels 4. The labels each have a machine readable portion 2 and preferably also a separate portion 1 capable of being read by a human. The machine readable portion of an object label is read by a device 5 which subsequently reads the machine readable portion of one or more reference labels. The reference labels read for a particular object label are recorded thereby enabling subsequent identification of the object label by information from a reference label. <IMAGE>

Description

SPECIFICATION Man-Computer data input technique This invention relates to a method of accurately and quickly denoting Relations between items and item properties. The method uses 'tags' to denote item and property identities. A tag comprises a man-readable legend and a machine-readable version of the legend. A device is used to read, gather and process the information from the tags. The order in which the tags are read defines how the mappings defining the Relations are developed, as will be presently described.

The primary features of this method of gathering data are as follows.

(a) The mappings between items and properties are one-to-many and there is no limit to the number of properties to which an item can be mapped. In other words, an item can be associated with as many properties as are appropriate.

(b) There is no limit to the cardinality of the item and property sets. Indeed, new members of these sets can be created by preparing a new tag. This would typically involve writing a legend on a label which bears a new, unique machine readable code.

(c) The technique does not intrinsically require keyboard skills. The only necessary skill is that of manpuiating the reading device and understanding the method used to record the Relations.

(d) The data input can be performed anywhere that the reading device technology permits. If the items are pipe parts in an undersea oil field then this method could be used by a diver to build a database in situ for the Relations between parts, their state and position.

(e) The method is particularly suited to miniaturised data capture devices which have restricted userfeedback capabilities. Such devices enable the data to be gathered very conveniently at source, independently of the more powerful computing resources needed to process the information. Such deferred data entry is particularly well catered for by this data entry technique.

A specific application embodying the method will be discussed to illustrate the benefits of the technique.

This does not restrict the applicability of the method or imply the use of specific technologies.

An application will have a set of items divided into subsets, the nth subset being In ln={iii is an item) Each item is usually given one unique identity which corresponds with its spatial position and thus the subsets are usually disjoint. Each item may be associated with properties from property subsets Pm: P,=(plp is some property) The property subsets need not be disjoint and often are not. Neither set I nor P need have fixed cardinality.

The purpose of the input technique is to associate items with properties. These associations can be recorded by developing a Binary Relation Rk: iRkp where iEln and pEPm and n, m are values denoting the particular subsets of I and P. The kth Relation is something that has a binary nature, such as'... has content concerning ...' or '. . . was accessed at ...'. The reason for introducing separate subsets of I and P arises because a Binary Relation can be represented by a matrix with one boolean element for each of the pairings defined by the Cartesian product lx P. Each boolean element records whether i either is or is not related to p.In practise this matrix tends to be very sparse and, by creating the subsets In and Pm, only the 'populated' partitions need to be represented. Data input is also made more easy since, for particular items, only certain property subsets need be considered.

Having developed these relations over some period of time they are put to use by exercising the converse Relation Rk: pRk'i This makes it possible to find all the items which possess some property in the context of the Relation k. The implications of this will be illustrated in the application discussion.

To gather the information required to initiate, develop and augment the Relations Rk a record must be created containing: -the item identity, - a sequence of properties and - properties implicittothe context in which the information was gathered.

The latter refers to properties such as the date, the time of day, location or the person who gathered in the information. Such properties may be regarded as 'pervasive' in that, though they may change, they are implicitly added whenever a new item record is created. An essential characteristic of the record is that it does not have a finite size. This is a direct consequence of the one-to-many mapping which is a feature of the input method.

The record is created by using the reader device to create a data sequence which logically conforms to the following BNF grammar.

< data~sequence > ::= < data~record > < data~sequence > I < void > < data~record > ::= < start~of~record > < record~body > < end~of~record > < record~body > ::= < item~identity > < properties > I < properties > < item identity > < properties > < item~identity > ::= < item~set~id > < item id > < properties > ::= < property~identity > < properties > i < void > < property~identity > ::= < pervasive~property > I < immediate~property > < pervasive~property > ::= < property~form > < immediate~property > ::= < property~form > < property~form > :::= < property~set~id > < property~id > The (property~setid) corresponds to the subset Pm which forms part of the range of relation Rk. The (item~set id) corresponds to a subset In and is typically used to separate items in a coarse manner. For example, the databases used to record the Relations may be separated according to items with implied properties for convenience. Simpler applications may not use differing item and property subsets and thus some or all of the (. .#seUd)s may be void (where'. . ' may be either 'item' or 'property'). The only requirement that must be met is that the (. .~identity)s must be distinguishable.

For efficiency in storage, the pervasive properties may not be actually stored with each record but only appear in records when the property changes. For the same reasons the (. .#seLid# or even the (..~ides may be omitted if their is no ambiguity as to what the record body should contain for the particular application.

Other optimisations may be applied such as collapsing the (startof~record) with the (end~of~record), i.e.

making one or other void. Also the order of the elements in the sequence may be employed to substitute for particular parts of the record. The purpose of such optimisations is to reduce the number of actions required to create a record. The application example that follows illustrates this. However, removal of extra actions does reduce the redundancy in the gathered data and may degrade reading action error recovery.

The stream comprising the (data~sequence) will usually be forwarded to another computer device for subsequent processing, either immediately or at some later time. This is to enable the data 'pile' so gathered to be transformed into a database organised for optimal performance with queries based on the converse Reiation(s). A wide variety of suitable techniques existing using indexing employing B-trees or derivations thereof. However, some database models are not particularly well suited to the underlying one-to-many Relation mapping.

The database(s) will almost certainly use a different form of item and property coding than that used for the data input. An implementation might typically use a four digit code to stand for a (property~id) which would need converting to its alpha-numeric string equivalent, e.g. 89123'LUNG CANCER'. Thus, prior to forwarding to the database building and updating processes, the .Jd)swill need translating. For this purpose the following one-to-one mapping functions are defined.For the subset In: Fin: InoSn where S,=(sls is a database allowable item name) and for the subset Pm: Fpm: Pmewm where Wm=(w;w is a database allowable property name) The inverses of these functions may be used to create the tags necessary for the input method. Only the man readable portion of the tag is of importance to the user and it is thus essential that the legend should be consistent with its machine representation. Using the above example, the machine readable code 8912 E In and, since Fin is one-to-one, only stands for the property 'LUNG CANCER', the man-readable legend. Provided means are available for extending the definitions of Fin and Fpm, future tag requirements can be met by creating tags with preassigned unique codes but with no legend.The legend can be completed later. When the translation program finds that the code does not belong to its domain it is told what it represents (if necessary). The mapping function may then be permanently revised.

The following diagrams and tables relate to the above discussion and the ensuing description of an application which uses the input method.

Figure 1 is a data flow diagram showing how information gathered is treated and forwarded to some form of database, Figure 2 shows, in diagrammatic form, how keywords are associated with a document in a filing system application, Table 1 gives an example of the type of data sequence gathered by the reader device, Table 2 is a translation of the data sequence example of Table 1 as it might be interpreted by a typical database update program.

Figure 1 illustrates the man-computer data input technique. Items (3) are to be associated with properties (4) by means of the tags, each of which feature a man readable part (1 ) and a machine readable part (2).

Typicallly an item tag is attached to the physical entity for which it stands. The property tags may reside anywhere convenient, usually near the items with which they are most closely related. The reader device (5) is manipulated to generate a data stream (6) comprising of item and property identities read from the machine readable parts of the tags. The association between an item and its properties is formed by a tag reading sequence as illustrated by the reader device path (10). The data sequence may be stored in the reader device for forwarding to a translation process F (7). The particular function performed is to translate the item identities and property identities in the data sequence (6) into suitable codes for the database update process (8).The translation process F (7) is programmed with the mapping functions that transform particular identities into the database update codes. The database update process (8) takes the transformed data about items and their properties and records these in one or more databases (9) for later use by a suitable database interrogation process (assumed).

An application in which the input method described may be put to good effect involves the logging, keywording and retrieval of paper-based documentation. It is rarely practical to attempt to recode information which is already on paper into machine readable form for the purposes of indexing and retrieval. Indeed studies have shown that retrieval based upon full-text indexing is not as effective as it might seem because of contextual difficulties. Much documentation contains information which is very difficult to code anyway, such as drawings, photographs, marginalia and signatures. A practical alternative is to retain the paper (or an image of it) and create a database which only records the relationships between the paper items and their potential retrieval cues (properties).Such databases are sometimes termed 'phantom databases' since the actual data is not held in machine readable form.

The task of the proposed input method is to facilitate the creation of the database(s) containing the Relation information. The traditional method would involve the properties of the items described by keywords and then record these using conventional keyboard input. However, this would not be regarded as an acceptable technique for many users who may not be able to use, or inclined to use a keyboard. Even if a keyboard is readily to hand, using it too freely may give rise to some of the problems encountered with full-text retrieval systems.

The input method uses tags to code the documents (items) and the retrieval cues (properties). The document tags give each a unique identity and denote its position in the spatial storage system. The cue tags stand for keywords that the information user would wish to assign to their documents. A practical implementation could typically use printed barcodes as a cheap, reliable and flexible method of representing the machine readable parts of the tags. Barcodes are easily created using commonly available computer controlled printers on ordinary paper. Other implementations may choose to use other techniques such as magnetic stripes, though these are less convenient to produce. The reader device for the implementation using barcodes might be a convenient, pen-sized barcode reader with a non-volatile buffer in which the scanned information can be stored sequentially.

For this application illustration I is divided into two subsets, Ii corresponds to 'product information' and 12 to research papers'. For convenience the two sets of documents are stored separately and the Relations databases are also separate. The set P is divided into subsets as follows, - P1 for product descriptive keywords, - P2 for research area descriptive keywords, - P3 for product manufacturer names, - P4 for information source names and types, - P5 for dates, - P6 for personal identities, - P7 for document destinations.

The Relations of interest are R1, the Keyword and Logging Relation, and R2, the Tracking Relation. R1 is updated each time a new document is added to the filing system and its range is P1#6. Its converse Relation R'1 enables documents to be found which have a particular keyword associated with them andlor which were logged on a particular date. For the database for Ii the members of P2 are not so likely to feature and similarly for 12 and P1. R2 is perhaps more useful in retrieval situations than R'2 as it records who had a particular document and when; its range is P#7. Slnce P5,6 are common to both ranges and do not change often, these would qualify as pervasive properties.

Figure 2 shows how a new document (11) is introduced into the filing system. A new document tag (12) is attached from the prepared list of tags (3) in the user's filing kit (15). The reader device (5) is used to read the document code on the tag. The filing kit contains lists of properties tags (4). The user builds the Relation Ra by reading as many of the tags from these properties tags as are appropriate. The pervasive properties, one from each of subsets P5 and Psr would be entered from lists in the filing kit when necessary if they are not automatically added by the device itself. The document, identifiable as now belonging to the filing system by its tag, is now stowed in its correct position in the filing system store (17).If the document tags are serialised within the subset then this would correspond to the next available space for that type of document ('product' or 'research'). This would typically be on the front of more or less chronologically 'stacked' documents. When a document is retrieved from the file (18) then its access is recorded by reading its document tag (19) with the reader device. This information is used to add to the Relation R2.

Table 1 shows a typical example of a reading sequence. Several optimisations to reduce tag reading effort have been utilised and these are explained in the commentary in the Table. The following points should be noted from these examples: (1 ) The (item~setid) I also acts as a sentinel to mark the start of each (data~record). Thus the (item~identity) must always be the first tag scanned.

(2) When there are no keywords associated with a (item~set id) then the (data~record) belongs to R2.

(3) P6 and P7 come from the same tag list but are differentiated by there (property~set id). To create an element of P7 an extra prefix is added to the P6 tag (the ' > ' symbol).

The read data is then subject to the mapping functions, either immediately or subsequently, giving rise to the information stream illustrated in Table 2. This format of information would then be suited to forwarding to a database updating program. Since the two databases are kept separate one of the tasks of the ensuing update program will be to parse the data stream to separate the records relating to each of the databases.

Additionally the records relating to each of the two Relations, R1 and R2, must be separated by the parser.

These Relations are thus appropriately updated for each database for each broad category of documents.

Data Sequence Commentary N0068 Reader device owner id (pervasive) D021486 Date stamp (pervasive) T1400 Time stamp (pervasive) 181067 Document number (introduction) 691 371 Keywords 992 180012 Document number (retrieval) Destination prefix N0037 Destination name 151005 Document number (retrieval) Destination (same as previous) 181068 Document number (introduction) 371 225 Keywords T1415 Time stamp (update) 281 364 I51039 Document number (introduction) 980 Keyword 181021 Document number (retrieval) I Table 1 Example scanned data sequence (... indicates possible presence of other data) field content (comment) ITEM = PDD/P/1067 KEY = Printer KEY = Laser KEY = Zapper Inc. (new keyword) BY = PWB AT = 14: :00 ON = 14 Feb 1986 ITEM = PDD/P/0012 TO = JB BY = PWB AT = 14:00 ON = 14 Feb 1986 ITEM = PDD/R,/1005 TO = BY = PWB AT = 14:00 ON = 14 Feb 1986 ITEM = PDD/P/1068 KEY = Laser KEY = Coupler KEY = Plastic KEY = Newsclip BY = PWB AT = 14:15 On = 14 Feb 1986 ITEM = PDD/R/1039 KEY = Menu BY = PWB AT = 14:15 ON = 14 Feb 1986 ITEM = PDD/P/0021 BY = PWB AT = 14:15 ON = 14 Feb 1986 Table 2 Example translation of data sequence

Claims (24)

1. An information retrieval system comprising: a.plurality of labels each having a machine readable portion; a plurality of references each having a machine readable portion and each adapted to receive an identifier which identifies the reference and which is capable of being read by a human; a device for reading said machine readable portions; and automated storage means for storing information resulting from use of the device to read labels and references.

2. A system as claimed in claim 1, further comprising additional storage means for storing details of respective identifiers applied to the references.

3. An information retrieval system comprising a plurality of labels each having a machine readable portion; a plurality of references each having a machine readable portion and a portion capable of being read by a human; a device for reading said machine readable portions; and automated storage means for storing information resulting from use of the device to read labels and references.

4. Afiling system comprising: a plurality of object labels each having a machine readable portion; a plurality of reference labels each having a machine readable portion and a portion capable of being read by a human; a device for reading the said machine readable portions; and automated storage means for storing information concerning relationships between individual object labels and one or more of said reference labels, said information resulting from use of said device to read the object label and the said one or more reference labels.

5. A system as claimed in claim 3 or 4, further comprising buffer storage means which accumulate information from the said device and subsequently transfer the information to the said automated storage means.

6. A system as claimed in any of claims 3 to 5, wherein each of the object labels is adapted for attachment to a physical object.

7. A system as claimed in any of claims 3 to 5, wherein each object label includes a portion capable of being read by a human.

8. A system as claimed in any of claims 3 to 7, wherein the said machine readable and human readable portions of each reference label are constituted by a single legend.

9. A system as claimed in claim 7, wherein the said machine readable and human readable portions of each object label consist of a single legend.

10. A system as claimed in any of claims 3 to 7, wherein said machine readable portions are in the form of bar codes and said device is in the form of a bar code reader.

11. A system as claimed in any of claims 3 to 7, wherein said machine readable portions are in the form of respective magnetic stripes and said device is in the form of a magnetic stripe reader.

12. A kit for building a computer assisted filing system, said kit comprising: a plurality of object labels each having a machine readable portion; plurality of reference labels each having a machine readable portion and a portion capable of being read by a human; and a device for reading the machine readable portions of the labels.

13. A kit as claimed in claim 12, further comprising buffer storage means for temporary storage of information provided by said device.

14. A method of storing information concerning the attributes of objects comprising: providing a plurality of machine readable object labels, providing a plurality of machine readable references, assigning a number of said references to respective attributes, providing each assigned reference with an identifier which identifies the respective attribute and which is capable of being read by a human; assigning a number of said object labels to respective objects; machine reading an object label; machine reading one or more assigned references; and recording which references were read for a particular object label.

15. A method as claimed in claim 13, further comprising recording the assignment of the references and including in the recordal of which references were read details of the assignment of those references.

16. A method of storing information concerning the attributes of an object comprising: providing a machine readable label on the object; providing a plurality of machine readable references which describe respective attributes each reference having a portion capable of being read by a human; reading the machine readable label on the object; reading one or more of the machine readable references; and recording which references were read for a particular object.

17. A method of storing documents comprising: providing a plurality of document labels and a plurality of reference labels, each label having a machine readable portion and each reference label having a portion capable of being read by a human; providing an object label on each document to be stored; reading said machine readable portion of the label provided on each document to be stored; reading the machine readable portion of one or more of said reference labels; and recording which reference labels were read for each document.

18. A method as claimed in claim 16 or 17, wherein said recording comprises a temporary storage while information is accumulated for a number of object labels with subsequent transfer to permanent storage.

19. A method as claimed in any of Claims 16 to 18, further comprising the provision for each object label of a portion capable of being read by a human.

20. A filing system substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.

21. An information retrieval system substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.

22. A kit for building a computer assisted filing system, substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.

23. A method of storing documents, substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.

24. A method of storing information substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.