Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F40/00—Handling natural language data
  • G06F40/10—Text processing
  • G06F40/194—Calculation of difference between files
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F40/00—Handling natural language data
  • G06F40/10—Text processing
  • G06F40/197—Version control

Definitions

• the present invention relates generally to techniques for handling information objects (web pages, files, parts of files, etc.), and for navigating among these information objects.
• It aims in particular to propose methods in which a user can personalize “source” information objects, and where both the changes which have appeared in the source information object or objects and the personalization which he has effected can be found in a current view each time said user makes access to such an object.
• the present invention also relates generally to the methods implemented in an environment or computer system for assigning to information objects or resources relevance scores, and the methods for accessing resources in an environment or computer system based on such relevance scores.
• the main object of the invention is therefore to allow a user to create “relevance relationships” between a first resource 1 and other resources, in order to be able to find the first resource simply by accessing at least one relevant resource. compared to the first resource.
• one (or more) resource (s) relevant to the first resource can (not) be, not only one (some) of said other resources that the user has linked (s) ) to the first resource, but also: - a resource (s) having a relevance relationship with the first resource, created by another user in a network of relevance relationships shared between these two users, - one (of ) resource (s) for which the first resource is automatically deemed relevant by means of inference rules 2 which apply to the relevance relationships thus created, - one (or more) resource (s) for which (which ) the first resource has a high relevance score which was automatically determined by the process by assimilating the relations of
• resource an information resource.
• relevance relationship between two resources is meant a relationship indicating their relative relevance (one in relation to the other).
• An object of the invention is that, in addition to the relevance relationships created by the users themselves, the method can use relevance score propagation conduits created automatically from the interrelationships and the content of the resources.
• a method of the invention aims to automatically create conduits: between the co-cited resources 4 , between the resources linked by hypertext links, between the resources having distributions close to words, possibly by combining in addition with determined conduits with further other methods of calculating relevance between resources.
• the invention aims to allow the user: • to find the first resources; • as the user can share a network (relevance relationships and relevance score propagation paths) with other users, the process aims to allow the user to discover the first resources expressly linked by others users within the shared network; • Finally, the process also allows the user to discover other relevant resources from private or public resources (eg on the Web) to which good scores have spread via automatically placed conduits.
• the present invention further aims in particular to propose techniques in which a user can modify a source information object - he can for example associate annotations to it, or personalize part of it - and save the modifications as a page derived from the original page.
• the invention also aims to allow the derived page to be spontaneously presented to it by the system when it revisits the original page.
• the invention aims to determine, by exploiting the links which are found in the pages of the Web, the degree of relevance of the pages in relation to each other and in relation to the browsing context of the user.
• the invention aims to present the user with the most relevant derived page as well as annotations associated with other relevant pages and links to other relevant pages.
• Another object of the invention is to allow an annotation to be for example a link to an information resource such as a file, so that the user can thus find information resources simply by placing themselves in their context of relevance: when the user navigates and gets closer to the context of the resource he is looking for, we aim for a system which spontaneously suggests it to him.
• the invention thus aims to use the system constituting, for the user, an "associative memory".
• the invention aims, finally, to facilitate the manipulation of information objects by allowing selections of parts of information objects made in a graphical user interface to become in turn manipulated for example for the benefit of other objectives supra.
• the invention proposes a method implemented in a computer environment to generate a current view from at least one source information object liable to vary, characterized in that it comprises the following steps: establish and store a reference of the view to the source information object, use by a user at least one information object modification tool to produce an object from the source information object transformed information, save in a memory a set of difference information representative of the effect of modification actions performed by the user using the tool (s), and generate a current view from the or each object source information in its current version and said set of difference information.
• the information object modification tool includes a document editing program.
• the editing program is able to generate a log of the modifications made to each document, the recording step is implemented using the log data.
• the information object modification tool includes a plurality of modification operators that can be individually selected by the user.
• the current view consists of a reference to the source information object, said reference being associated with a containing information object.
• the method further comprises a step of establishing and storing a relationship between an information object containing directly or indirectly said view and the source information object.
• the method further comprises a step of establishing and storing a relationship between the view and the source information object.
• the method further comprises the implementation of an inference process based on a plurality of semantic relationships between views and source information objects.
• the reference of the view to the source information object is constituted by a request in response to which one or more source information objects are delivered.
• the request is a request applied to a search engine, the latter delivering the source information object or objects.
• the request consists of a group of information objects, and the method comprises a step of searching for at least one other information object having the best score of relevance with respect to said group.
• the method comprises a step of memorizing the current view generated, and said set of difference information contains one or more pointers to one or more information objects whose content and position are contained in the current view.
• said information objects are structured into hierarchical sub-objects, and in the case where said set of difference information contains difference information targeting a part of an object, then the step of generating the current view retains the latest version of the smallest sub-object containing such an object part.
• said difference information relating to a part of an object comprises an item of attribute change information.
• said difference information targeting a part of an object comprises object position information.
• the method is implemented as many times as there are sub-objects.
• the difference information set contains, for an object comprising a plurality of sub-objects, modified / unmodified attributes for each of these object objects.
• each object identifier is unique in the IT environment.
• each current view has at least one composite identifier comprising a modification identifier and an object identification to which the modification by the corresponding difference information set applies.
• a current view has an identifier referring to a plurality of modification identifiers and a plurality of object identifiers to which the modifications by the corresponding sets of difference information apply respectively.
• a corresponding composite identifier comprises an object identifier possessed by said other starting information object and a double identifier of sets of difference information.
• a method implemented in an IT environment for assigning relevance scores to information objects, characterized in that it comprises the following steps: - assign to at least certain pairs of objects information a capacity (conduit) for propagation of relevance score from one information object to another, - assign to at least one information object a relevance score, and - calculate, for at least one other information object, a relevance score from relevance scores of information objects having with said other information object a propagation capacity and values (sizes of conduits) of said propagation capacities.
• a propagation capacity is assigned to each pair of information objects, a propagation capacity possibly having a zero value.
• at least certain information objects are linked together by citing information object links to cited information object, and the propagation capacity between two information objects is calculated on the basis of a number criterion common citing information objects between these two information objects.
• the calculation step includes a sum of the relevance scores weighted with the propagation capacities.
• the method includes a step of developing, by the user, a propagation capacity between two information objects of a pair.
• each pair of information objects is assigned two propagation capacities, one in a first direction and the other in the opposite direction.
• the invention further provides a method for accessing information objects in a computer environment, characterized in that it comprises: - the creation of a group of initial information objects, - the implementation implementation of the method as defined above to identify other information objects having the highest relevance scores with respect to the group of initial information objects, and - the display of elements representative of said objects of information identified for access to these by the user.
• the method includes the prior step of implementing a search process to develop the group of initial information objects.
• the group of initial information objects is developed from a manual and / or automated classification of information objects in an operating system of the IT environment or in a computer application.
• the method further comprises: - the determination of a refined relevance relationship between at least one of the identified information objects and at least one of the initial information objects, and - the graphic representation of said relationships of relevance refined.
• the display step comprises: the determination of the existence of non-zero propagation capacities between the starting information objects and said identified information objects, and - the graphical representation of links between said identified information objects and the original information objects as a function of the propagation capacities.
• the method further comprises: - determining the existence of a zero preexisting propagation capacity between at least one initial information object and an identified information object, and - modifying this zero propagation capacity in a non-zero propagation capacity.
• the set of information objects is formed by a basic network comprising information objects and the corresponding propagation capacities, and by modifications / additions / deletions made to this basic network to form a modified network, and the implementation of the method defined above is carried out from the group of initial information objects and said modification / addition / deletion data.
• the invention also proposes a method for accessing information objects in a computer environment, characterized in that it comprises: - the implementation of the method as defined above for determining absolute relevance scores of the 'set of information objects, - the implementation of a search process using a request, - the sorting of information objects satisfying said request according to their absolute relevance score, and - the display of '' representative elements of information objects sorted for access to them by the user.
• a method is also proposed according to the invention for accessing information objects in a computer environment, characterized in that it comprises: - the creation of a group of initial information objects, - the implementation of a relevance score calculation method to identify other information objects having the highest relevance scores compared to the group of initial information objects, - the creation of an object additional information representative of the group of initial information objects, and - the creation of propagation capacities, established as a function of said relevance scores, between the identified information objects and the additional information object.
• the invention also provides a method, implemented in a computer environment comprising information objects having references to other information objects, and in particular first content having at least one reference to second content such as a link to the second content or such as a transclusion of the second content, for the management of such references as a function of the existence of modifications made to the information objects by the user, characterized in that it comprises the steps following: - viewing by the user of a second current content via a reference possessed by a first content, in a mode allowing the modification of said second content; - when said second content is modified in said mode, creation of a new version of the second content and replacement of said reference by a reference to this new version.
• said reference is a query such as a query in a standard language, to a database.
• said new version of the second content is represented by a structure of markers containing the differences between said new version of the second content and said second current content, so as to present to the user a second content in which said differences have been applied to the version second content.
• the invention also proposes a method implemented in a computer environment for selectively accessing a derived information object created in the computer environment from a starting information object, characterized in that it includes the following steps: (a) developing relevance scores of the original information object and the derived information object using a given relevance determination mechanism; (b) in connection with the access by a user to the initial information object, whether or not providing the user with information on the existence of the derived information object or access to said object of derived information, based on said relevance scores.
• step (a) also draws up a relevance score of at least one other information object
• step (b) further comprises whether or not providing the user with information on the existence of said other information object or of access to said other information object, according to said relevance scores.
• step (a) also develops a relevance score of at least one other information object, and the method further comprises, in connection with access to the initial information object or to the derived information object, a step (c) of providing or not providing the user with information on the existence of said other information object, as a function of said relevance scores.
• step (a) also develops a relevance score of at least one other information object, and the method further comprises, in connection with access to the initial information object or to the derived information object, a step (c) of displaying or not displaying at least part (annotation) of the content of said other information object, as a function of said relevance scores.
• step (b) is implemented in relation to the existence of a modification made to said derivative or initial information object since its previous supply to the user.
• step (c) is implemented in relation to the existence of a modification made to said other information object since its previous supply to the user.
• the starting information objects being connected by directional links to together form a starting oriented graph
• the method further comprises the step of constructing a virtual derived graph formed from the starting oriented graph and virtual links starting from information objects citing the initial information object and directed to the derived information object, and the relevance scores are developed on the basis of said virtual derived graph.
• the derived information object is an information object having at least one view towards at least part of the initial information object.
• the derived information object is an information object having at least a copy of at least part of the original information object.
• the step of constructing the virtual derived graph also takes into account virtual links starting from information objects quoting said other starting information object and directed towards said other derived information object.
• the step of constructing the virtual derived graph takes into account the virtual links starting from information objects citing the objects d 'respective initial information and directed to the derived information object.
• a method implemented in a computer environment for personalized access to information objects comprising initial information objects linked by oriented links to form together a starting structure, characterized in that it comprises the steps consisting in: (a) creating information objects added to the starting information objects, to form a derived structure, (b) storing the derived structure under form of a reference to the starting structure, of the added information objects and of links between the starting structure and the added information objects, (c) memorizing, in association with the derived structure, specific research information relevant information objects, (d) search for relevant information objects in the derived structure using specific information from re seeks.
• steps (a) to (c) are implemented repeatedly to form a plurality of derived structures by adding information objects to the starting structure or to preexisting derived structures.
• step (c) of memorizing specific search information takes place in the form of a reference to specific search information already associated with the structure or to a preexisting derived structure, and specific search information added.
• the invention additionally provides a method implemented in a computer environment for selectively accessing one of a plurality of information objects from a starting information object, characterized in that it comprises the steps following: (a) elaborating relevance scores of the initial information object and other information objects using a given relevance determination mechanism, said other information objects comprising objects of independent information and information objects derived from the original information object and / or these independent information objects, (b) in relation to access by a user to the original information object , whether or not to provide the user with information on the existence of another information object or access to said other information object, according to said relevance scores.
• step (b) is implemented in relation to the existence of a modification made to said other information object since its previous supply to the user.
• the relevance determination mechanism is based on analyzes of links between information objects, at least one of the derived information objects is obtained by reproduction of a starting information object or by insertion, in a any information object, a view or a copy of the whole of said initial information object, and the method comprises a step of creating virtual links directed to said derived information object and starting from each information object having a link directed to the initial information object considered.
• the relevance determination mechanism is based on analyzes of links between information objects, at least one of the derived information objects is obtained by inserting, in any information object, a view or d '' a copy of only a part of a starting information object, this part being or containing itself a view or a copy of the whole of an indirect starting information object, and the method comprises a step of creating a virtual link directed to said derived information object and starting from each information object having a link directed to the initial indirect information object considered.
• the method further comprises, in connection with access to the initial information object or to another information object, a step of displaying or not displaying at least part (annotation) of the content of 'a third information object, according to said relevance scores.
• the display step is also implemented according to the existence of a modification made to the third information object since its previous display.
• the present invention also provides a method implemented in a networked computing environment for displaying a personalized information object obtained from an original information object, characterized in that it comprises the following steps: load the original information object from a site making said original information object available to the user, save in a memory a work information object obtained from '' a copy of the original information object, implement information object modification tools to produce from the said working information object a modified information object, save in a memory a set of information representative of the differences between the original information object or the work information object and the modified information object, and on subsequent access to the information object original, generate, from the original information object in its current version and of said set of information representative of said differences, a new modified information object.
• a preferred but non-limiting aspect of this process is as follows: * the work information object comprises a plurality of content belonging to the original information object and identifiers assigned to the respective content, and the differences represented by said set of information are differences identifier by identifier.
• a method of manipulating information is proposed using a graphical user interface, said interface comprising means allowing selection of a portion of given information, characterized in that it comprises a step of converting 'a selection made using said interface into a later manipulable object.
• the conversion step is implemented only in response to a user instruction triggered using a pointing tool used to make the selection.
• the method further comprises a step of selective display of a visual indicator showing the existence, among displayed information, of an information object resulting from a previous conversion from a selection.
• process or the term “system” will be used in an essentially undifferentiated manner, depending in particular on whether one refers to the actions implemented by the process of the invention or the computer system that supports these actions.
• object designates any information resource having an identity. For example: • a computer file • an element of a document in XML format • a directory • a paragraph in a Word document • a selection of part of a Word document • a web page • an entry in a directory (this entry representing a particular person). An object can contain others, thus forming a hierarchy 6 .
• relation covers any form of reference to one (or a set of) “source object (s)”, which reference is associated with a “connected object”.
• a relation can thus be for example: • a semantic relation from one object to another. For example, o an “author” relationship of a document (the linked object) written by a certain person, to the object (source object) representing this person;
• resource and “information resource” are also synonymous - for example in UP (Unifor Resource Identifier).
• transclusion or inclusion by reference 7
• HTML A special case of relation is transclusion (or inclusion by reference 7 ).
• img the "img" tag in HTML allows you to include an image by reference.
• transclusion is represented by the option "Create a link" from the menu that appears when a Word document is dragged and dropped from a selection of part of the document by holding the right button of the mouse supported.
• Transclusion consists in creating a view ("the result") of at least one object ("source”) placed in an object ("the container”).
• Source objects can be distant. They are either identified directly or implicitly as the result of a search query 10 .
• the source (of a transclusion) means "the source object or the set of source objects (of said transclusion)”.
• FIG. 1 schematically shows an example of a series of enriched transclusion operations 13 .
• A, B and C There are three documents, respectively named A, B and C. Initially these 3 documents are empty except A which contains a section (whose identifier is 1) in which there is a paragraph (whose identifier is 2).
• paragraph 3 First of all the user transcludes paragraph 2 in document B (transclusion 3). The result is paragraph 3: 2. Then he transcludes the latter in section 1 of document A, and places it just before paragraph 2 (transclusion 4). Finally he transcludes section 1 (taken entirely) in document C (transclusion 5).
• document C contains section 5: 1 in which paragraphs 5: 4: 3: 2 and 5: 2 are presented.
• the transclusion enriched with a view of an R2 directory would offer the advantage of being able to modify (add, rename, delete, etc. .) objects (files, sub-directories) in this view without modifying the content of the source of this view (that is to say without modifying the original content of the R2 directory - see Figure 2). If later a new object is added in R2, it will also be 1 added in this view of R2 (as in the simple transclusion).
• the invention provides in particular the implementation of an editor (or explorer) of objects presenting the objects in a tree-like manner (in the style of current file explorers) which allows on the one hand to carry out enriched transclusions, and on the other hand to create relationships between objects (for example to link a file located in a directory (or a directory view) with another file located in another directory in the same tree structure - see Figure 2) , each relationship optionally giving rise to a channel for the propagation of relevance scores.
• the user will create relations of the type “comments” (i.e. such object is a comment of), “confirms”, “contradicts”, “is against”, “is for”, “supports”,
• such an object explorer is able to present relations connecting an object taken as a whole to a part of another object.
• an object not explicitly presented such as a selection of only part of a document (the document being represented by an icon 17 in the object editor) can be linked with an object presented explicitly such as 'another document or directory. The user can then “zoom in” on the object not explicitly presented to refine the granularity of presentation.
• conduits are placed (automatically, for example by measuring the similarity of word distribution, or manually by the user) between web objects and the user's private objects, some of which can, for example, be transposed from other users' private items. This is how relevance scores for web objects spread to user 1's private objects. Private objects and / or objects of the Web can thus be considered relevant and will then be presented to the user, according to his current context of navigation 19 or exploration of objects.
• new relevant found objects can be automatically presented in the context of an object editor (such as a file explorer described above; for example new documents can be inserted in the (enriched) view of the R2 directory) .
• object editor such as a file explorer described above; for example new documents can be inserted in the (enriched) view of the R2 directory
• relevance relationships are automatically placed with existing objects (Figure 2 illustrates the automatic placement of a relevant object and its relationships in the Rl / réelle-de-R2 / R3 directory. ).
• the user can choose how the result of the transclusion should be presented, the trivial case being to insert all of the content. At the other extreme he can ask to have only a hypertext link (and therefore that the content is only accessible by a click on the link). The user can request that certain changes are made to the content source 20 before it is inserted into the object to be presented. He can choose a transformation from a pre-established list (which he can complete as needed), for example applying a function which selects the first n characters of each source object.
• the user can also define the transformations to be performed by acting directly on the result of the transclusion, using the appropriate content manipulation tool 21 (the tool he would use if he wanted to modify the source).
• the results of the actions he performs are saved in order to to be able to apply these transformations automatically the next time the object is displayed or updated. For example if the user exchanges two parts of a transcluded object, when the content of this object changes at the source, the two parts will keep their new position but their content will always correspond to that of the source.
• the current version of a transclusion (or "current view”) is the representation of the result of the transclusion, as it was presented the last time. This current version is kept, and is updated by the system each time the user accesses it.
• each transclusion is associated an object of differences (or "diff object"), which indicates the source or sources of this transclusion, as well as the transformations to be applied to it.
• the object of the differences is based on the current version 22 and therefore does not need to contain "verbatim" the contents and the positions of the objects to be transformed.
• the diff node indicates the source (ref) and the result (id) of the transclusion (or the identifier of the transclusion itself). Note that the source can be identified by a reference or a request.
• the identifiers “id” refer to the current version of the transclusion.
• FirstSentences is an operator that selects the first sentences of the content of an object. Modify indicates that the content of the object must be replaced (the new content being stored in the current version).
• these transformations can advantageously be carried out at the source level in order to avoid transferring unnecessary data.
• the object of differences refers to the objects of the current version (that is to say to their respective identifiers) to define the transformations which are applied to them. Move indicates that the object has been moved (the new position is also stored in the current version).
• Move, Modify and Create were created by the user by acting directly on the objects, in the same way as if he modified a local version (the process allowing this is described later in the section "Manipulation of objects").
• the user can request that the content of the source of a transclusion is not included at all, but that a simple hypertext link be placed as a result, pointing to the source or sources.
• the object of the differences, in this case, would look like this:
• Each transclusion can automatically be associated with a semantic relationship linking the result of the transclusion to its source, and expressing the nature of the changes made to the content. By default, it is an equivalence relation, that is to say that the meaning of this content is not affected.
• the user has the possibility of adding another semantic relation which applies from a direct or indirect container of the result of the transclusion, to its source.
• Figure 3 presents a transclusion whose source is the object obtained in a directory, representing the author of a document, and whose result is placed in this document (and transformed so as to show only a photo 23 ) . It may be advantageous for the user (in particular when carrying out searches, as described below) to place the semantics "author" at this transclusion. In this case it will ask to apply the relation to the document itself.
• the user may wish to have only a semantic relationship between two objects, without modifying the appearance of the objects concerned.
• Figure 4 shows an object representing an article of law and another object being a legal commentary of this article of law. If the user wishes to inform the system of this relationship without modifying the appearance of these objects, he will transpose the legal commentary in the article of law (at point "X” in Figure 4), select the desired semantic relation (“commented by”), and request that the result of the transclusion be “not presented” (and therefore naturally remove the equivalence relation from the result of the transclusion). This amounts to applying a transformation which removes the entire content of the object.
• he may request to have only a hypertext link 25 , as explained previously.
• the user interface thus makes it possible, within the framework of a single interaction (consisting in defining a transclusion), to specify both the hypertext link and the added semantic relationship.
• the information objects can themselves be characterized semantically.
• the semantic characterization of an object consists in indicating to the system what (s) information (s) this object presents (ie "what this object is talking about”). This can be done by specifying keywords (or descriptive text), or by creating relationships with objects representing semantic properties. This latter approach is facilitated by instantiating structured concepts 28 which offer the advantage of better structuring the added information.
• a relation to a transclusion is optional and one can therefore define a relation without associating it with a transclusion (for example when characterizing an object semantically, as described in the following section).
• the user interface visualization objects will then present the hyperlink as well as for example the title of the semantic relationship in question.
• the user interface (or programmatic, in particular within the framework of a Web service) could also present this semantic relation within the framework of a resource graph (for example according to one of the RDF or OWL recommendations of the W3C) and allow the user (or a program) to navigate from resource to resource. So in the example of the author relationship between a document and the resource representing its author, the interface allows you to navigate from the document to its author and from there to the author's address or to other documents of the same author.
• At least one concept is associated with at least one other determined object, by a method of determining relevant objects from a set of starting objects, as being relevant with respect to a (or des) object (s) determined according to step b1, the system offers him said (or said) concept (s) associated (s) for semantically characterizing the object to be characterized; 3. then for each associated concept selected by the user, the latter characterizes the object in question by instantiating it and by entering its properties 30 , for example by simply filling in fields in a form associated with said selected associated concept.
• the object to characterize is a document concerning case law.
• the user therefore uses the keyword "jurisprudence” to characterize it.
• the system recognizes the concept with which this keyword has been associated (suppose that for example the word "jurisprudence” is the very wording of the concept in question).
• a form for collecting additional information 31 is thus associated with this concept.
• the system offers the user to fill in this form to specify and complete the information: what is the type of object of the conflict, who is the plaintiff, who is the court, what is the jurisdiction, on what date the decision has been returned, etc.
• the user thus creates relationships between the object to be characterized and the objects corresponding to the properties entered (in the "form" in question). Like all relationships, they are likely to spread relevance scores as described later in this report.
• Objects can be created, modified and deleted, using the usual tools of the user.
• the user can also create, modify and delete transclusion results and semantic relationships between different objects, as described in the previous section. Finally, it can directly act on the content of the result of a transclusion in the same way as on local content.
• the transclusion operation can be implemented by a drag and drop operation.
• a drag and drop operation For example, the selection of part of a resource (viewed in the context of a resource manipulation tool such as Word, Excel or Frontpage, or even in the context of a directory and file explorer) being made , dragging and dropping while holding the right mouse button pressed will result in the presentation of a menu (context menu) in which the user can choose the option "Transclude here".
• the system will offer the user to choose a transformation (for example in XSLT, in particular in the context of Frontpage) from a set of transformations available if necessary to immediately modify the presentation (adapt it) to this location. Then the user can still modify the result of the transclusion interactively as already described ...
• a selection can be "reified". To reify a selection is to define it as an object (in the sense of an identified object, as defined in the introduction).
• the essential advantage of reified selections is that they can be distinguished from other selections and thus encourage reuse. It is advantageously possible to implement an option according to which the system automatically reifies any selection which serves as a source of transclusion or which is linked (by a relation among a set of types of relation specified beforehand) or according to still other criteria (in particular avoiding overlaps). The user will thus be able to avoid recreating a selection which had already been created on the same part of the resource or almost (in particular this is important to take advantage of the conduits of propagation of scores connecting this resource to the others).
• a handle titled "reified selection” - or any other distinctive means that we will assimilate here to such a handle - can be displayed at the selection level to allow it to be identified.
• a button or a menu option is offered in the toolbar to display (or alternatively hide) all the "reified selection” handles of a displayed resource, the purpose being of course able to select a pre-existing reified selection (if there is at least one).
• the selection is made conventionally by dragging the mouse so as to cover the selected part, the left button being pressed (this conventionally has the effect of highlighting the selected part); Then the reification itself is done by clicking on a menu option presented in the toolbar, or by pressing the right mouse button, which must first be positioned on the selection in question.
• the reification can be done directly within the framework of the very selection operation: by dragging the mouse so as to cover the selected part the right button being pressed (which will have the effect of highlighting the selected part) and by selecting the "Reify” option from the "context menu” (conventionally displayed when using the right mouse button).
• a "reified selection" handle is then displayed at the selection level to serve as a guide. It is useful to distinguish the reified selections actually used (in particular those which serve as a source for a transclusion or which are linked by relations being of a certain type among types previously specified) from other selections, and especially from those which do not not used at all. Indeed, the user will be able to more easily reuse those which are already used and thus avoid cluttering the resource in question. These can be distinguished by graphic means and, to lighten the presentation, under the control of the user, the system can automatically delete the reified selections that were not used at the end of a session or for example after a given period of time.
• Clicking on a "reified selection” handle has the effect of selecting the object it designates (and making the corresponding system highlight the corresponding selection, that is to say typically having "reverse” video "the part in question of the resource viewed).
• the system identifies and presents to it the closest reified selection (geographically) in the displayed resource ( however, this requires switching to a special mode).
• the system could automatically identify the most relevant reified selection (s), that is to say which correspond most closely to its "user profile” (as described in the "Personalization Method” section).
• the user can transpose part of an article viewed in Word by means of a drag and drop to a web page viewed in Frontpage.
• the system will offer to choose a transformation (in XSLT) from a set of transformations available for presentation at this location. Then the user can still modify the result of the transclusion interactively as already described. Change the result of a transclusion
• a query can therefore combine several properties, using Boolean operators.
• the network of conduits is an oriented network whose vertices are objects and the edges are conduits to which a numerical value called size is attached.
• a conduit between 2 objects an ability to propagate a relevance score from one object to another.
• the propagation capacity is called the size of the duct.
• a request is made up of a set of objects weighted by numerical values 34 .
• a positive value indicates a "score source” and a negative value indicates a "score absorber”.
• Each object distributes its score along its outgoing conduits (distributed according to their sizes).
• a score source produces score for itself.
• a score absorber spontaneously decreases his score.
• the method seeks to give scores to each object so that each object receives and creates as many scores as it destroys and gives, that is to say that we seek a situation of equilibrium.
• the query (all the sources and absorbers of score) describes the domain where the user wishes to obtain results (therefore the objects that the user wishes to favor or disadvantage), the conduits represent relevance relationships (a conduit of A à B means that if A is interesting then B will be interesting too, to a certain extent which is indicated by the size of this line).
• the result found by the method therefore corresponds to the objects which are relevant to the request, according to the network of conduits. Indeed, an object having incoming paths from numerous objects relevant to a request must also be relevant.
• xi is the score of the object.
• the exponent + represents the value after update.
• s; is the value of the source or the absorber of score for the object (zero if i is not a source or absorber of score).
• the conduits can be obtained by several means.
• a transclusion indicates a certain relation of relevance, even in the absence of explicit semantics; this is particularly useful in the case of hypertext links between web pages, because the system is thus able to analyze the web automatically, without any other intervention.
• the system can be extended, depending on the applications, with other “pipe installers”, that is to say functions which produce pipes between objects.
• conduits between objects which cite the same objects the size of which is equal to the proportion (weighted) of outgoing links in common.
• a difficulty is that it will be necessary for each type of semantic relationship to give the size of the conduit to associate with it, and for each installer of the conduit to give the weighting of its conduits vis-à-vis the other posers. These weights represent the importance in terms of relevance of these different kinds of conduits.
• each object When one wishes to obtain scores in an absolute way (without particular request), one can place a source of score with each object (thus with each iteration each object distributes a part - the same one for all the objects - of its score to all the other). This will have the effect of ensuring that each object has a non-zero score, and therefore of maintaining the differences between the different objects, as opposed to the case without sources of scores, where entire regions may end up with zero scores as explained more high, and therefore where the differences between objects disappear.
• This profile is in fact a weighted average of the interest of the user, the weight decreasing the more we go towards the past.
• the present interest can be estimated by considering the objects consulted by the user and the queries he makes (so when the user makes a query, the scores of the results will also be the scores of the objects in his present interest. also serves objects visited by the user, these can be considered as requests and the same processing can be carried out.)
• the weighting is carried out by making an average between the average interest of the user at the previous instant and the user's current interest.
• P (i) the score of object i in the user's profile, and p (i) the score of object i in the present interest of the user
• P (i) will be updated with s P (i) + (1-s) p (i), s being the weighting factor (constant giving the “weight of the past”; this value may depend on the information used to determine the interest In particular, a user request will leave less weight to the past than a simple page visit. 35 )
• An object can have optional children, so they should only be displayed if the system judges that they correspond to the user's profile. It can also have exclusive children between them, i.e. groups of objects whose one that most closely matches the user's profile will be displayed.)
• An object is said to correspond to the profile of the user when the score of this object in relation to the profile of the user is high.
• Each object has an identifier.
• Locally created objects have an atomic identifier, which is a whole number unique to the system.
• Each transclusion also has an atomic identifier, which is also unique.
• Objects seen through a transclusion have a composite identifier, noted A: B, where A (the head of this identifier) is the identifier of the transclusion, and B (the tail of this identifier) is the identifier of the object at the source of this transclusion.
• Parentheses are used to clearly delimit the head and the tail when necessary. In the absence of parentheses, the sub-identifiers extend as far to the right as possible. For example,
• a transclusion A is here considered on the one hand as a set of objects "seen through” A, and on the other hand to a set of modifications concerning these objects. These modifications include • modification or deletion of an object seen through A (therefore whose identifier begins with “A:”) • displacement of an object seen through A, whose parent before and after displacement is also an object seen through A. • creation of an object (locally created or seen through a transclusion different from A) as a child of an object seen through A. • displacement of an object seen through A as a child of object created or seen through a transclusion different from A.
• a particular case is as follows: when the user performs a transclusion, the algorithm, to perform this operation, will first create the transclusion but at this instant the latter n is not yet accessible (because no object seen through this transclusion is yet placed as a child of an accessible object.) The second step is then nothing more than a displacement of one of these inaccessible objects as a child of an accessible object.
• a polynomial simplification algorithm consists in replacing all the products of sums by sums of product, then for each x to group the terms in d x .
• the two forms therefore have the same density, as expected.
• Each object can have a set of sources, which is the set of objects of which it is the image by a transclusion, implicit or not.
• Each source is associated with the corresponding transclusion.
• 4: 3: 2 in the example above, has two sources, 3: 2 and 4: 2, through transclusions 4 and 4: 3, respectively.
• A is a source of B by the transclusion C
• the identifier of B is equivalent to C: A.
• Each object is associated with two types of information: its position and its content 36 .
• the position of an object is given by its parent and its predecessor (the predecessor being used only to determine the position of the object among the children of its parent).
• the notation ! A means that information was created locally by object A.
• Each object will store, in this form, the origin of its content and the origin of its position.
• the object is an element and its content is the "innertext"; indeed the attributes and the child elements are considered here as being child objects (contained in the object in question) therefore treated in their turn as objects each having a content and a position.
• the attributes present additional constraints, such as for example the fact that the algorithm will have to merge the same attribute coming from two different sources.
• the calculation of the origin of the position of an object is carried out in an identical way to the calculation of the origin of its content with the difference, of course, that the origins! Oj considered are the origins of the positions of these sources .
• the parent and the predecessor of the object will not, however, be equal to the parent and predecessor of O j , but to their image by the path T j: C j: So if the parent of O j is P then the parent of l 'object to calculate will be TCP, and the same for the predecessor. If the predecessor thus obtained is not a child of the parent thus obtained, then this information of predecessor will be ignored (and the object will for example be placed at the end of the list of children of its parent), [local 5]
• FIG. 8 presents an example of the behavior of this algorithm, for the calculation of the positions of the objects (since the calculation of the content is not detailed here, step 4 does not appear).
• the 5: 1 transclusion result also gives its position, in fact it is also considered displaced 38 , since its parent is not the corresponding by 5 of the parent of 1 (not shown in the diagram in order to avoid overload it.)
• Transclusion 5 specifies that 5: 3 is moved as the first child of 5: 1.
• transclusion 6 moves the objects 6: 5: 1 and 6: 5: 4.
• any source of information for this object will be the image of the same first source through a sub - all of these transclusions (or their images through a transclusion).
• the polynomial associated with a subset of these Ti is obtained if one replaces Ti by t 1 and one carries out the sum of all these terms.
• the subset Tl T3 gives the polynomial t + t 3 .
• One source has priority over another if the polynomial associated with the set of Ti which separates it from the object of which it is a source is greater than the corresponding polynomial of the other source, for t very large.
• the algorithm for constructing the object of differences will have to place in this object of differences the information that the root of the result of this transclusion is displaced.
• the terminator as the number that comes last in an identifier. For example, the terminator for 2: 3 is 3, the terminator for (4: 5): ((3: (8: 9): 5) :( 4: (7:11))) is 11, the terminator for 5 is 5.
• Ci If the different sources (Si is the source and Ci the path, Ci: Si is equivalent to O).
• Tn T (nl): ...: T2: Tl: S, where the Ti are transclusions of atomic identifier.
• S is the object at the very end of the chain, also atomic.
• the requested object is the image of another object by a transclusion
• the following objects must be recalculated with the local algorithm: • The object itself (in order to obtain its content) [global 1] • For each source (in the case of implicit transclusion), the images of all the descendants of this source by the transclusion corresponding to this source [global 2] • The set of objects created and moved by the transclusion [global 3].
• 4: 1 is obtained through a transclusion. We apply the local algorithm on 4: 1 to obtain the content [global 1]. 2. 4: 1 has only one source, 1, by the transclusion 4. We will therefore recalculate the image by 4 of all the descendants of 1, so 4: 3: 2, 4: 2, 4: 6, 4: 3: 5 and 4: 5 [global 2]. 3. 7 is the only object created or moved by 4, we recalculate it [global 3]. 4. The descendants thus obtained for 4: 1 are 4: 3: 2, 4: 2, 4: 6, 4: 3: 5, 4: 5 and 7. We continue the global algorithm on each of them [global 4]. 5. 4: 3: 2 is the image of an object by a transclusion.
• the local algorithm does not need to be called on it because it is already done [global 1]. 6.
• the set of transclusion pairs: source of 4: 3: 2 is 4: 3: 2 and (4: 3) :( 3: 2). We will therefore call the local algorithm on the descendants of 3: 2 seen through 4, then on the
• T2 has priority (closer) to Tj if i ⁇ j.
• T2 is closer and therefore more priority than T4. Therefore a path which borrows T2 and T3 will have priority compared to a path which borrows T4, for example.
• a path that passes through T3 and T4 will take priority over a path that passes through T2 and T5, because of T5 which has less priority than all the elements in ⁇ T3, T4 ⁇ .
• the present invention allows a user to create relevance relationships 40 between a first object 41 and other objects, in order to be able to find the first object simply by accessing at least one relevant object with respect to the first object.
• Said (or said) object (s) relevant to the first object can be, not only one (some) of said other objects that the user has connected (s) to the first object, but also: • one (s) object (s) having a relevance relationship with the first object, created by another user in a network of relevance relationships shared between these two users, • one (s) object (s) for which (which) the first object is automatically considered relevant by means of inference rules 42 which apply to the relevance relationships thus created, • one (or more) object (s) for which (which) the first object has a score of high relevance which was determined automatically by the method by assimilating the relations of relevance to capacities for propagation of relevance score from one object to the other (that is to say to conduits).
• the method uses conduits created automatically from the interrelationships and / or the content of the objects.
• the method automatically creates conduits, in particular 3 • between the cocitated objects 44 , • between the objects linked by hypertext links, • between the objects having distributions close to words.
• conduits are created in particular between web pages likely to interest users.
• conduits or relevance relationships between web objects and private objects (eg company documents) are created and thus allow the relevance scores of web objects (pages to be propagated to the latter).
• Web which are calculated automatically by analyzing the structure of the links and their content.
• Finding / discovering an object by accessing i.e. by visiting or expressly selecting
• relevant objects are done, for example, in the following scenarios.
• Tracing relationships between objects can be done in one of the following ways: • by explicitly indicating to the process the identifiers of the objects in question, or by first accessing a first object then by indicating to the process the identifier of a other object (to specify its relation to the first); • or can be done within a document, via a transclusion of each object that we want to link, that is to say by creating in a document references to the objects in question then by tracing relationships between these references in the document itself, using an editor; • or even in a mixed way (editor to object or object to editor).
• Figure 11 (which consists of a variant of Figure 2) illustrates such relationships presented (and made manipulable) in a typical specialized editor that we propose to use in a business intelligence activity on the Internet.
• Each block could represent an enriched transclusion, for example from a web page.
• By clicking on each of the arrows (which represent a relevance relation) the user can specify which semantic relation it is, and the given semantics will then appear when the mouse cursor is positioned on the arrow in question (this is also not illustrated in Figure 11).
• Any semantic relationship can be a relevance relationship and new relevance relationships (or conduits) derived automatically, semantics can be added manually or by means of user-configurable rules. The user can thus discover new semantic relationships, highlighted by the process.
• the result of the process is a set of new objects (new blocks are added to FIG. 11) which are effectively linked with the objects viewed by the user (arrows connect these new blocks with the existing blocks).
• the result of the method must be ordered first according to the user's wishes (for example, private objects first) and then sorted by relevance score (this method is described below).
• the process analyzes the web and calculates relevance scores for the pages. To do this, it calculates a network of conduits.
• a conduit between two pages expresses the fact that if the first is relevant (for the current request or in absolute terms) then the second will also be.
• the conduits also have a "size" which expresses the intensity of this propagation.
• the default conduit size between two web pages is the number of inbound links common to those two pages. It is possible to add (or enlarge) conduits following the links
• hypertexts 48 or from text analysis for example, a conduit is placed between two pages, the size of which increases with the resemblance of their word distributions), or by various other methods.
• the user can himself modify this network as desired, by adding / removing these conduits or by modifying their size.
• Process 4 continues inside the company's documents. For example: • The user can add a conduit from a web page to a company document, or from one company document to another. In general, a conduit can be placed from or to a reified selection of a document or page part.
• the user can perform a transclusion in a document whose source is a reified selection, which will also create conduits in both directions between the source and the result of this transclusion. • The user can set a conduit between two results transclusion, which will be replaced by default 50 a conduit between the respective sources of these transclusions. 51 • Conduits are placed between the Web and private documents according to their proximity to word distribution.
• selection will designate all the possible cases of reified selection of part or all of the content of a document or of a page 52 .
• a selection is therefore an object
• the user interface allows it to draw, between selections, relationships (as illustrated in FIG. 11) and to explain their meaning 53 ; these relationships generate conduits 54 .
• the user can configure the order in which the results are presented to him by default, according to the location (for example that his private documents are shown to him before the other documents of the company, or conversely that the documents found in the web be shown first) and score.
• the system can be placed in client or server mode.
• server mode it manages a server network
• client mode it manages a modified network which is a set of modifications made to the server network of a particular system.
• system in this document implies that it manages a single network.
• the user 50 may choose where the conduit will actually be placed (in each case or by reconfiguring the interface).
• the user can put a conduit from or to the result of a transclusion which by default will be replaced by a conduit from or to the source of this transclusion or its upstream transclusions if there is one ( cascade transclusions).
• the source and result of a transclusion are selections; the two ends of a conduit are selections.
• such relationships may indicate, for example, that a selection confirms / contradicts / causes / influences / explains / explains / translates / etc another selection.
• conduits are created according to rules that can be modified by the user.
• advanced users can be offered an API to define duct size calculation systems themselves, in particular according to the semantics of the content of the selections or the relationships between the selections.
• 55 See section "Request Handler" In client mode, the user creates and maintains his network, in particular by transcribing selections
• the network is maintained thanks to the fact that each conduit thus created is actually created in the network, and only one reference is inserted in the document in question.
• the network in which the conduit in question is actually created can be that of the system in client mode or directly that of the system in server mode.
• a user can make all or part of his network available to other users (his subscribers). When he modifies his network, his modifications will be directly visible and effective for his subscribers.
• a request representing a set of selections can itself be a node in the network.
• Conduits connect it to the selections having had the best scores for a previous execution of this query (their sizes being a function of these scores).
• the user can trace or modify relationships or paths from or to a set of selections (having constituted a request to the system).
• the query can be executed either by means of the relative score calculation algorithm expressed below, or by another method, in particular by the relative distillation method described in this report.
• each system manages a single network.
• part of the conduits in a network can be imported from another network and automatically updated from the latter 56 .
• Figure 10 an architecture 57 in graph structure where the Web is crawled by a basic system managing a
• Base network from which other systems (Company networks) 58 import conduits and add their own selections and conduits; and from these other systems, custom networks in turn import conduits that can be manipulated by users.
• the region denoted “CRAWLED” represents the part of the basic network resulting from the crawl of a subset of the Web of interest to a company; the region
• MUTUALIZED represents the part of the basic network resulting from the crawl of a subset of the Web of interest to two companies at the same time (which offers the advantage of sharing the crawling of this subset by the basic system) ;
• IMPORTED represents the part of the company's network made up of conduits imported from the basic network; the region
• ADITIONAL represents the additional conduits connecting the company's resources to each other as well as to web pages in the IMPORTED region.
• conduits and scores are generally neutral and uniformly applied in the basic network as well as in company networks, while they can be special (i.e. personalized) in personalized networks. Some query results can be delivered to custom networks (as introduced above).
• conduits from one network can be imported from any other network 59 .
• certain conduits of the network managed by a given system 60 can be references to conduits of networks of other systems (their sizes thus being traced
• Part of a network can be an enriched transclusion of part of another network.
• the user of a personalized network potentially benefits from all the enhancements made upstream in the graph, which may for example include selections in the form of advertising.
• a user can contribute modifications he has made from his network, upstream in the graph, and the owner of the network upstream from the transclusion in question can accept or not this contribution. Users can thus communicate relevant resources.
• Each system is able to estimate for itself the knowledge it has of the domain of each selection that is part of a query (this is described later in the section “The Distiller”). He can thus decide to delegate all or part of a request to a system which has the knowledge in question and with which it is connected in Peer-to-Peer (as described below).
• inertia a is a constant between zero and one. 0 gives none inertia (risk of oscillation) and 1 prevents any variation (large values slow convergence and 1 prevents any convergence)
• the well-known “PageRan” algorithm propagates scores only along hypertext links. Propagating scores along the conduits, for example if they are calculated from co-quotes, promotes selections that would not otherwise have been offered.
• a selection is highly cocited with a selection having a good score and in the second case several selections with a good score have a link to the same selection.
• a mixed algorithm that is to say to propagate the score both along the co-quotes and links, will discover the pages of interest in both cases.
• conduit makes it possible to combine several methods of searching for relevance relationships (including manual modifications) in order to obtain better results.
• the network of conduits is independent of requests 69 . Users can choose to share the same network and all the modifications made to this network by different users can be used at the same time.
• transclusions The exchange of networks between users is facilitated by the implementation of transclusions: a user can exploit the network of another user by performing a transclusion in a document in his system, and he can then modify it, reduce or enlarge it to suit its needs, simply by editing the document containing the result of the transclusion.
• the equations are linear in s (ie the scores obtained for the linear combination of several queries will be equal to the corresponding linear combinations of the scores obtained for the basic queries) which means that queries can be broken down into more basic queries which can then be recomposed differently. Thus one can calculate and memorize these elementary requests then obtain the composite results without having to re-carry out the calculations.
• the scores x t for R 3 are the average with the same weightings of the x, - correspondents of R ! and R 2 .
• Linearity is used in particular in the presentation of the results to the user and in the Peer-to-Peer delegation process described below.
• Each network represents the relevance of the selections each compared to each other, for a user or for a set of users, whatever their requests. However, different networks (for different interests) can be created by (or for) " ⁇ my users. a set of “authority” selections with high absolute scores.
• the score of a selection for a given request can be obtained by calculating for each selection in the request the amount of score x a that it would send to each selection authority, then averaging the scores x a weighted by the scores obtained by these selections a. So, by precalculating for all a the distributions of scores x a and x a , page scores can be calculated very quickly.
• navigation context can be used when giving relevant selections to the user.
• the navigation context is the set of selections that the user has seen or selected with a weight describing the importance it has in the present context.
• the context is taken into account, at each user request, by carrying out a weighted average between the context and his request before processing it.
• ⁇ is the set of score sources, i.e. ⁇ 2 -> X; 3 ⁇ -> ⁇ (as explained before, their sum must be worth 1).
• the interface allows you to request that the relationship between two selections be displayed in a document, whether it is a semantic relationship, a path imposed by the user or an implicitly calculated value.
• the interface allows you to perform a transclusion, the result of which is placed in a document.
• the interface will superimpose the relations starting from or arriving at the source (and if necessary from the source of this source and so on) and from or to result of this transclusion.
• the user can then modify the relationship in the document, and the change will be automatically reflected in the network.
• a reference will be sent to the server so that the latter can take it into account in the score calculations.
• the relationships that the user creates (or modifies, deletes, etc.) will be stored in the system, whether in client or server mode. If in client mode, the interface offers a way to send these conduits to the server (so that they are used for other client systems on the same server).
• This interface can be integrated into a standard editor such as Word.
• the user can also request to see the entire network of relationships (both those added by the user and those calculated automatically) in the vicinity of a selection.
• the relationships from or to the sources, of a transclusion will be “descended” in the graph presented.
• the user can add / delete semantic conduits and relationships.
• the user can request to see only certain relationships, i.e. activate or deactivate the display of relations obtained by subscription (selectively subscription by subscription), activate or deactivate the display certain types of relationships semantics, and so on, all the means of calculating the sizes of conduits can be selected or not to give rise to the display of relations.
• the user can configure the way in which the default sizes are calculated.
• the interface allowing to do this is presented in the form of a list of ways to calculate the sizes of conduits (cocitations, hypertext links, conduits given explicitly, subscriptions, etc.), on each line (except in the case of subscriptions and explicitly given conduits) the user can indicate the weighting and if it is not zero activate or not the display of the corresponding conduits.
• an option makes it possible to make them effective or not for the calculation, and here again if they are active an option makes it possible to display them or not in the representation of the network. More concretely this can be achieved with a two-column display, the first allows to weight (or activate / deactivate) the corresponding calculation and the second allows to display or hide the relationships in the representation.
• the user can select a set of selections (transclusions or not) as a request and ask the system to suggest selections relevant to this set. He then receives a list of selections which he can use to enrich his document.
• each selection is associated with a path which is a list of names such as "/ companyA / divisionB / departmentC /”. Selections found on the web will have a path such as "/ web / http: // ".
• the system represents in the interface in the form of a tree all the nodes existing for example in the company, and the user can choose if he wishes to see only the nodes which contain results.
• the user can open or close the nodes and reorder the children of a node and the system reacts by presenting the results in the same order as the nodes in the tree and by sorting according to their scores the results found in a closed node .
• the modifications carried out on the tree are memorized, that is to say that the tree will be presented in the same way at the next request.
• the user can also request that only the results found in certain nodes of the tree be returned to him (he will notably be limited to the paths containing selections to which he has access).
• the system offers the user tools to manage the context (for example to delete selections, to reset it, to temporarily deactivate it, to save it in order to resume it later, etc.).
• This part of the system explores the network of hypertext links on the web at the request of the other modules. It will also periodically check whether the sets of incoming and outgoing links of the selections have changed, and more often check the selections receiving requests from other modules.
• the scoring algorithms described above have the disadvantage that when the pages of a set are strongly linked to each other, any page outside the group will have a bad score relative to the pages of the group.
• Page 1 should get a good score relative to page 2 because, apart from the navigation links, all pages pointing to 2 also point to 1. However, the equations will only see the fact that, among the eight pages pointing to 2, only three also point to 1.
• the internal links to the site link very close pages (because the two ends are linked in common by many pages) while the others link pages of little proximity (few pages, in fact none .in the drawing) point to both the source and the target of this link.
• the vertices of U are copies of the vertices of AN, i.e. pages.
• Each edge (link) is associated with a weighting which is equal to the complement (1-x) of the proximity (homogeneity) of the two linked pages, calculated in the network U.
• this network is lazily built , that is, nodes and links are created when they are requested.
• the cardinality of a set of links in AN is replaced by the sum of the weights of these links. (For example we replace the number of reverse links by the sum of the weights of the reverse links).
• the sum of the scores for reverse links is replaced by the sum of the products of the scores and the weighting of the corresponding link.
• the cardinality of the union of two sets of links is calculated as follows. The contribution of a link present in only one of these sets is equal to its weighting. The contribution of a link present in the two sets is equal to the logical sum of their weights, i.e. l- (l-a) (l-b).
• the cardinality of the intersection of two sets of links is the sum, for all the links present in the two sets, of the products of their weights.
• This module is responsible for calculating the sizes of the conduits between selections (depending on the configuration by the user, by cocitation, along hypertext links, by text analysis, by translating semantic relationships, by taking a value imposed by l 'user, or conduits obtained by transclusion).
• the system will memorize in the network the conduits obtained by transclusion, and these are updated when the user views the documents containing these transclusions.
• the system will choose, in order of priority: 1. A pipe size or a semantic relation given by the user 2. A pipe size provided by transclusion 3. The pipe size calculated automatically ( cocitation summer)
• this module In addition to asking, the sizes of conduits, we can also ask this module how a particular conduit has been calculated in order to display it in a document (for example what type of semantic relation has been placed or if it is is a fixed value imposed by the user).
• the system can be configured to introduce artificial nodes into the network, each representing a request. Bidirectional conduits between the request and the resulting selections (those with the highest scores) are then laid, the sizes being a function of the scores for the request in question.
• conduits with special requests, such as those implementing a “relative distillation” method described below.
• this module When the user has chosen to publish part of his network, this module will be responsible for constructing the data to be sent, i.e. all the sizes of conduits either themselves obtained via a transclusion or explicitly provided by l 'user.
• This section describes an accessory component of the system allowing it to be used for web monitoring.
• the system must be able to suggest pages that have just been created, even if the relevance of the page based purely on hypertext links seems low. Indeed, if the page has just been published on the web 74 , it is likely that it is known to few people and that consequently few pages point to it, even if in reality it is a very good page. It is therefore necessary to calculate its relevance in another way, so as to give them a chance to appear among the suggestions to the user.
• Each page is associated with a state which can be:
• One scalar product of two distributions is equal to the sum for all the words of the products of the frequencies of appearance in one and the other document. The number will be close to one if the documents use the same vocabulary and zero if not.
• W represents the set of selections
• x (p) represents the absolute score of the selection p
• x p (q) represents the score of the selection q relative to p.
• r (p) x (p) - r max (max (x ⁇ (p), x p ("))), which means the absolute score minus (with a weighting ⁇ ) the greatest relative score involving p and a selection of A (which represents the “distance” of the candidate selection p with those already taken), ⁇ is a constant.
• a small value of ⁇ indicates that we prefer to have selections with a good absolute score and a large value indicates that we prefer to have one of the selections distant from each other.
• A is initially empty. Then as long as we wish to make the set A grow, we add to it the selection of W having the greatest representativeness.
• ⁇ . (I) be the score of selection i for the query consisting of the only selection y, calculated using the algorithm described in the introduction.
• the score will give too small a value but it is not important because it is assumed that the selections of A are chosen so as to be in the points (of the domain of interest to the user) easily affected by score streams in the conduits.
• the distiller is therefore responsible for memorizing and updating the values of the functions x a and x a for each selection of A.
• a request to this component consists of a set of selections, possibly a set of modifications to be made to the server network (this is used by the client mode described below), as well as a set of filters on the selections to return (which will be used in particular to return only the selections to which the user has access), and it responds with all the selections having the highest relevance score for this query (the score for each selection returned is also returned).
• the system can be used in two ways, either it does all the calculations itself (server mode) or it uses another system by providing the modifications (ie modifications / additions / deletions) that the user contributed to the other system's score propagation network (client mode).
• this module will simply send the request back to a server, along with all the modifications made on the score propagation network.
• this module In server mode, if the request does not request modifications to the score propagation network, this module simply calculates the function x R defined above. When a filter on selections has been given, only the scores of selections satisfying the filter will be calculated
• this module will modify the request so as to simulate the effects of the changes made to the server network.
• This algorithm works even for selections which, in the server network, have no incoming or outgoing path, unlike the modified network, which will occur in particular when a user of the client system has created selections for which he is the only one. to have access and therefore for which he is the only one to have put conduits. Since it itself depends on the result, this query will be iterated approximately, that is to say that you must have an estimate 80 of the result of the query to calculate the modified query, and that the results of the modified request are used to calculate a more precise modified request, until convergence 81 of the x values which can then be returned in response to the request. The closer the first estimate of the query is to the true value, the shorter the convergence will be, however whatever the initial values the algorithm will tend towards the same value.
• the request will consist of selection a.
• the score of q for this query is q s 0.1988, and so on.
• the diagram in Figure 24 represents all the data managed by the system as well as the dependencies which connect them.
• rectangles represent classes.
• a name inside a class represents data attached to each instance of the class.
• the asterisk * represents multiplicity.
• An arrow represents a dependency between two data (called source and target of the dependency).
• a dependency entirely contained in a class links attributes of the same instance of the class.
• a symbol • or! at one end of an addiction indicates the strength of the addiction, weak or strong respectively.
• a dependency is strong if its triggering makes the target data inaccessible until its update is completed, and weak if the target data remains accessible during its update
• the force symbol is on the source side if it is a push dependency, which means that the target is updated when the source changes.
• the ® symbol placed on a dependency indicates that the dependency is only checked if the last update of the target is old enough.
• a dependency whose source is the symbol indicates that the update of the target is carried out automatically at regular intervals.
• a dependency shown in dotted lines indicates that the update is done synchronously, that is to say that a modification of its source does not end until the update of the target is completed (and can be implemented by a simple function call, unlike the other dependencies which must be implemented by separate threads).
• the client mode system has no authorities and all of its requests are delegated to its server. In addition, it works on the same server network as its server and it is therefore not necessary for it to calculate the conduits obtained automatically: it can download them from the server. This leaves only the manually defined conduits, the semantic relationships and the subscriptions to manage.
• the system in server mode offers a function to make all the known pages grow as well as the set A. This operation must typically be carried out at startup, because at this moment the set A and the set of known pages are empty .
• the user gives a set of selections from which the extension will be made.
• the system then performs the relative algorithm for calculating the scores 82 starting from this set, and extending it along the conduits starting from selections having the highest score (which, via the dependencies, will cause a search for reverse links and before).
• the relative algorithm for calculating scores and the extension of the network are thus executed in parallel.
• border here means selection whose the outgoing duct set has not been calculated.
• the system will then be able either to destroy a possible set A already existing and to launch the algorithm of selection of the set A described in the section “The distiller” above, or simply to add new selections in this set, by resuming the algorithm from the set
• the error on the result and therefore the knowledge of the system on the request can be measured using the terms ⁇ 3r a ( z). Note that if the request is accompanied by a modification of the network, it is important to first calculate the modified request to take these modifications into account before calculating these terms.
• the user who is authorized to cause an extension of the network, requests suggestions for his document.
• the request modified by these three conduits will have a weight s not zero on the two web pages in addition to the two transclusions (see the algorithm in the section "The request handler"), and if no extension is made, only suggestions starting from the first web page will be given
• the user will define the domain he wishes to remove from his network by making a request, for example with a document containing transclusions, or with a set of selections given directly, etc. He can then tell the system to delete the authorities that have been used a lot to produce the results, ie the selections of A for which ⁇ R. • x a (i) is large enough (the user can either specify a threshold or an isR proportion of authorities to be deleted).
• the threshold we can display a graph indicating the proportion of authorities that would be deleted according to the threshold. If the domain he gave is tightly closed (i.e. weakly connected to the rest of the network) the graph will present a flat around the threshold to be chosen (because the threshold must be lowered a lot so that authorities in another area begin to be affected by their tower). This is illustrated in Figure 26.
• a peer-to-peer system can be used to delegate parts or all of the request to other systems.
• the user can link his system to other systems, thus forming a peer-to-peer network.
• the “neighboring” systems are those which are directly connected by this network. When the system chooses to delegate part of a request to other systems, it will ask its neighbors if they know a system capable of handling all or part of this request, that is to say which has knowledge values. ⁇ x a (i) large enough. ae
• a system If a system is able to handle part of the request, it informs the system where the search started (by providing it with its knowledge values for these selections), and queries its neighbors for the part of the request it did not know.
• the request will flow to the neighbors of these neighbors and so on until for all selections of the request a system has responded, or until the entire network has been traversed without success (which is detected if no response is received after a certain period).
• the first system (the one that initiated the search) can then distribute the request between the different systems that responded, then, thanks to the linearity of the algorithm, the different results can be recomposed directly.
• systems can memorize a table of the requests which most often reach them, by associating with it the system which had responded positively. This will allow requests to reach systems capable of handling them more quickly.
• each Distiller will ensure that the scores x a and 3c a are updated for their portion of the set A (according to the dependencies of the diagram in Figure 27).
• Another aspect of the present invention aims to propose techniques in which a user can modify an original page - he can for example associate annotations to it - and save the modifications as a page derived from the original page.
• the invention aims to determine, by exploiting the links which are found in the pages of the Web, the degree of relevance of the pages in relation to each other and in relation to the browsing context of the user.
• the invention aims to present the user with the most relevant derived page as well as annotations associated with other relevant pages and links to other relevant pages.
• Another object of the invention is to allow an annotation to be for example a link to an information resource such as a file, so that the user can thus find information resources simply by placing themselves in their context of relevance: when the user navigates and gets closer to the context of the resource he is looking for, we aim for a system which spontaneously suggests it to him.
• the invention thus aims to use the system constituting, for the user, a
• a first system implementing the method for selectively accessing one of a plurality of resources from a starting resource, characterized in that it comprises the following steps: (a) developing scores relevance of the starting resource and other resources using a relevance determination mechanism, said other resources comprising independent resources and resources derived from the starting resource and / or from these independent resources, ( b) in connection with the access by a user to the starting resource, whether or not providing the user with information on the existence of another resource or access to said other resource, as a function of said relevance scores.
• this system selects or not the content of a another resource to be presented, depending in particular on the relevance score of said other resource.
• support groups Users have declared a number of interests, known as "support groups," to the system. Each support group is initially represented by a set resource identifiers (typically URLs of popular web pages). Let us call R this set. We define 83
• R ⁇ + ⁇ by the set of resources having a link to at least one R "+ resource.
• Each R " + resource is associated with an authority score.
• the relevance score of a resource 86 for a given support group is the authority score 87 thus calculated for this resource.
• the relevance score of a resource P for a given support group can be determined more simply by counting the number of times that P is quoted 88 (by the resources of R " ) with the resources of R characterizing this support group 89
• Each resource derived from a 90 starting resource is constructed • by reproducing (and possibly by modifying) a starting resource • by inserting into any resource a view 91 (possibly modified) of a starting resource or an element a starting resource.
• the starting resources are linked by directional links to form together a starting oriented graph.
• the starting resources are for example web pages and the starting oriented graph is then the Web).
• the system creates a virtual link - directed to said derived resource (L5 directed to P7) - starting from each resource citing a starting resource (P2) having served to construct said derived resource.
• a first page is co-cited with a second page if there is a third page which has a link to the first page and a link to the second page.
• a resource can be derived from several starting resources at the same time, for example it can contain several views of different starting resources.
• the system creates a virtual link (L3) - directed to said derived resource (PI) - starting from each resource citing an indirect starting resource (P3).
• Figure 29 illustrates the inheritance of real and virtual links, as part of an example showing nested views.
• the page P4 contains a view of the page P3, this view itself containing a view of the page P5 as well as a view of the page P6.
• the view of P5 takes P5 entirely, it contributes (to page P4) the virtual link L1.
• the view of P6 contributes L2 and the view of P3 contributes L3.
• the page P4 inherits 93 from the links L1, L2 and L3.
• the page P2 contains a view on the part of the page P4 which contributes the virtual links L1, L2 and L3, this view in turn contributes the corresponding virtual links (also denoted L1, L2 and L3).
• page P7 contains a view of P2 (taken entirely), it inherits all the links (real or virtual) directed to P2, namely Ll, L2, L3 and L5. Furthermore, the page PI has a view of part of the page P2, this part being a view of the page P3 taken entirely and therefore contributing to its link L3. PI thus inherits the virtual link L3 (via P2). However, PI does not inherit from the other links of P2 (L5, Ll and L2) because neither it reproduces P2, nor a view that it contains takes
• the system thus constructs a “virtual derived graph” formed from the starting oriented graph and virtual links.
• page is thus meant “resource” (of any kind) to which a link can be directed (so that the resource is activated when the user clicks on said link).
• This mechanism is based on the notion of set of link sources for an element 94 .
• a page A is a source of links for an element B if any new link (real or virtual) pointing to page A also points (in the form of a virtual link) to B (and therefore to the page containing B as well than to all views containing B), as described above.
• Any element can have sources of links but only one page can be a source of links.
• a page A is a direct link source of an element B if it is a link source of B and that there is no page X such that X is a link source. of B and that A is a source of links of X.
• the extended set 95 of inbound links on a page is thus equal to the union of the actual inbound links on that page and the union, for all direct link sources, of their extended sets of inbound links.
• a derived resource inherits a link (virtual or not) from a citing resource and directed to a cited resource, means that for the system a virtual link exists (implicitly) from said citing resource to said derived resource.
• extended links we mean real links and virtual links. The construction of these extended sets is therefore separated into 1. the construction of these sources of direct links and 2. the inheritance of links proper.
• each instance of the system has a function returning the set of direct link sources for a given element.
• This function will return the union of all the pages of which a (full) view is contained in the requested element and sets of direct link sources of the parts of pages whose view is contained in the requested element.
• the set of extended links is then calculated as defined above, from these direct link sources.
• Figure 30 shows an example of eight pages.
• the first (PI) contains an entire view of the second (P2) as well as a partial view of the third (P3).
• P2 has an entire view of P4 as well as P5.
• the part of P3 seen by the second view of PI contains whole views of P6 and
• P7. P3 also contains a view of P8.
• the sources of direct links to PI are P2, P6 and P7.
• the extended set of links from PI is the union of the actual incoming links to PI with the extended sets of links from P2, P6 and
• the system selects or not another resource to present, according to the relevance scores.
• a relevance score was assigned - explicitly or implicitly - to each resource for each support group existing in the system. Recall that in the case where for a given support group the score has not been explicitly assigned, it is considered to be zero; the system knows in advance what resources whose score is not zero: those who belong 97 to the set R "of at least +
• the system selects or not the content of another resource to be presented if necessary in place of (or in relation to) the content of each view, according to the scores relevance.
• the selection of the resource P t most relevant to the view in question is then based on the logical sum (or the maximum) for j of the product of the score of this view for a support group E j with relevance score XEJ (PI) -
• the browsing context represents the user's current interests when browsing and visiting resources one after the other.
• a user's browsing context is defined by associating a score with each support group. You can initialize the navigation context with a score of 1 for all the support groups to indicate the absence of context. To update the navigation context, when a new PQ starting resource is visited by
• the user has the means, for a particular view, of indicating whether or not he wishes the view to be replaced by the system (if the user does not specify, each view will follow the default behavior of the system).
• the selection of said other resource P, the most relevant with respect to the starting resource Po, taking into account the navigation context, is done as follows: If X E J (P ⁇ is the relevance score of P, - compared to the support group Ej, the system selects the resource P, - whose relevance score in relation to Po and to the navigation context, calculated as being the logical sum 106 (or the maximum) for, / ' of K E j -XEJ (PO) -X E J (P ⁇ , is the highest.
• P t Information of the existence of P t is signaled to the user 107 if his score Si is sufficient, or the system can be configured so that (in particular if the logical sum for j of KEJ -XEJ (PO) -X ⁇ jfPi) is greater than the logical sum for./ ' of K E j -XEJ (PQ) -XEJ (PO)) Pi can be directly presented to the place of P ⁇ . 108
• Each resource created by the user is associated with an editing context D.
• his editing context is initialized to the browsing context of the user at the time of this creation.
• its new D + editing context becomes the weighted average of the old D editing context and the current K navigation context (when the new
• D + KK + KD, where K is a number between 0 and 1 representing the weight of the present compared to the past.
• the user will for example visit a page of an online book sales site and decides to monitor an element of a page (PI) presenting science fiction books. Later, when he visits another page (P2) which is relevant to the same field (of science fiction), the system presents him with the page (PI) of the bookseller whose element he had selected to monitor. However, the latter (PI) is only presented to it if the content of this element to be monitored has changed since the last time the user visited it.
• a view of a page (P3) accessed by the user reflects the content of a page
• the system offers the user a means of declaring that he has already seen a page (and that he therefore does not want the system to spontaneously offer it to him unless its content has changed).
• the system regularly uploads to the server all pages "viewed" by users and compares them to their previous downloaded version. When the system observes a change in a page, it will record with this page the date when the change was observed.
• the user can specify particular page elements whose change must be monitored.
• the system stores in a table the list of elements to be monitored (globally for all users) for each page already viewed, with for each element the last date when a change was detected 110 .
• the system regularly uploads to the server all the pages containing at least one item to monitor and compares them to their previous downloaded version. When the system observes a change, it will record the When it is a question of presenting an "other page" (more relevant) to the user, the system will select if necessary (according to the scores) another page in which at least one element that the user wanted monitor has changed since the user last viewed the page (or the item in question specifically).
• Each modification of an original page is shown to the user in solidarity with the page in which said modification was saved, except for the annotations 111 which can also be shown autonomously (ie adjacent or superimposed on another page displayed). 112
• the system is capable of presenting relevant annotations 113 , in addition to the methods already described.
• these annotations can in particular: • be displayed adjacent to the displayed page, • occupy spaces on the page reserved for advertising, • position themselves according to specifications given by the user when the annotation is created , • or be arranged in a special way specified or chosen by the user (by selecting from display configuration options).
• the user selects part of the page (by dragging the mouse there with the button pressed) • and / or provides the system with one (or more) comment (possibly containing a link ), • and optionally gives a set of keywords (or a display position).
• a derived page (by modification of the original page) containing this information is then created.
• the system searches the page for the occurrence of one of these keywords and displays the selection and / or the comment adjacent to the position of this keyword in the least intrusive way possible: for example, the keyword is simply underlined and the annotation appears when you bring the mouse cursor over it (by roll-over).
• the system is able to select only annotations whose content has changed since the last time the user saw them (as described in the section "selection of resources whose content has changed").
• the system When a user declares to have seen a page (or more precisely, an element under surveillance), the system will memorize the date of this instant, which makes it possible to know if the element has changed since the moment when the user saw. Thus the system is able to display only pages of which at least one element under surveillance has changed.
• Annotation is implemented as a special case of modification of a page.
• each annotation associated with a page is recorded as part of a page derived from this page (ie derived by reproduction / modification).
• Creating an annotation can consist of simply highlighting part of a page.
• the user selects part of a page (by dragging the mouse over it, the button being held down) and declares to the system that it is an annotation associated with this page.
• the system then creates a page derived from this page, in which the location of this annotation in the page is specified.
• the user can also modify the content of the annotation thus created, or even create any annotation containing new information, for example a link on another resource (see the last paragraph of this section).
• Said derived page created then contains in addition these modifications or this new information.
• annotations contained in pages that are relevant to the current page visited and to the context By clicking on an annotation with the right button of the mouse the user can (select in the contextual menu the option allowing him to) view the entire page containing this annotation 114 .
• An annotation can be simply textual, or visual or audible, for example an oral message or a video (or have any other effects on the senses) and can include one or more clickable link (s) by l 'user.
• annotations are able to include executable code directly and thus allow to extend the user interface.
• An annotation can contain a link to a resource 115 .
• the annotation could consist of an icon representing a document (and containing a link to it), an icon that the user associated with the page by dragging and dropping (from this icon) from his desktop.
• the system offers the option of downloading (to the system server) the resource at the end of this link, so that, when the annotation containing this link is presented, the user can automatically return this resource on his client computer and open it directly, simply by clicking on it (as he usually does to open a document).
• I (E) is the cardinal of the intersection inbound links from the pages of E, U (E) of the meeting, S (E) the cardinal inbound links from the page that counts the least, B (E) from the one that counts the most, constant.
• the introduction of the weighting by ⁇ is necessary because otherwise the equation does not make a difference between the case of a page having few inbound links but all of which are inbound links from the other pages of the set (should give a high homogeneity but gives a low value) and the case where all the pages have a lot of incoming links but only a small subset is common to them (the equation correctly gives a low homogeneity).
• the correction with the weighting by ⁇ increases the score in the first case (because the smaller the ratio the stronger the correction).
• the factor ⁇ indicates the strength of the correction. The larger it is (maximum 2) the stronger the correction. 0.6 is probably a good value.
• Equation by the reasons to love. where p (S) is the ratio between the cardinal of the union of the incoming links of pages of S divided by the cardinal of the workspace (the set of pages which we consider to have been able to point to a page of S, that is to say that they know). Since it is not possible to obtain the exact size of this workspace, we will generally take all of the pages known by the system.
• a pivot score and an authority score are associated with each page and the aim of the algorithm is to find values for these scores such as 1.
• the pivot score of a page is equal to the sum (except for a proportionality factor ) the authority scores of the pages to which it has a link 2.
• the authority score of a page is equal to the sum (to the same proportionality factor) of the pivot scores of the pages which have a link to it.
• R is defined by the set of pages having a link to at least one page of
• Each page of R ' is associated with a pivot score.
• Each R '+ page is associated with an authority score.
• the aim of the algorithm is to find values for these scores such as 1.
• the pivot score of a page is equal to the sum (except for a proportionality factor) of the authority scores of the pages to which it has a link 2.
• the authority score of a page is equal to the authority relative to R for that page.
• the system When the user requests to modify a page using an external tool, the system will create a derived page 117 .
• Legacy XML contains all the raw data that has been created for a site
• XML here replaces the“ current version ”described in the first part of the report). For example, in the case of a page containing a view towards another page of the site, the contents of these two pages are in XML legacy, but not the fact that the first contains a view towards the second.
• the "XML of the differences" (it is the equivalent of the object of the differences described in the first part of the report), contains all that is missing in the legacy XML to describe the site completely, that is to say does not contain any content (text, etc.) but rather everything that is reference between elements, the location and the reference of all views. It will contain the derived pages and a description of the differences (hence its name) present in the derived pages compared to the original pages 118 . The structure of these XML files will be described in the following sections.
• XML file in particular for XML legacy, can be stored and organized in various ways, not necessarily in the form of separate XLM files, for example by grouping them all in a single file, under a same node, or in an XML database.
• this section describes how to build a description of a page derived from a version modified by the user (using any commercial editor) but a specialized editor, allowing in particular to create views, can also modify these structures.
• page derived from a page X designates only a modified version of X (and not a page containing a possibly modified view of X).
• Either the referenced element has a unique identifier, and the reference is completely defined by this identifier.
• the difference XML describes the dynamic part of the system, in particular everything that is referenced. It describes what data to use and how to modify it in order to build the structure
• Each node in the differences XML corresponds to a node in the resolved structure and specifies
• the first information is specified using the name of the element and possibly a ref attribute which indicates a data source.
• the differences XML file can contain the following elements 119 : • deported: This means that the content of the element must appear as a copy of the element pointed to by the attribute ref (with all children) • imported: This state is similar to deported, except that the element is a version derived from the source (pointed to by ref), so it can be a modified version of the source (ie in addition to ref, it can be defined by modifications to bring to the source). • created: Indicates that the element was created from scratch by the user. It contains a ref attribute indicating where to take the element in an XML legacy. (All children of the node pointed in the legacy XML will be taken with) • rejected: Indicates that the element should not be displayed.
• the second information (on the location) can be defined as follows: • Using a reference of the replacing attribute: If a child of an import has this attribute, the element corresponding to the source of the import containing this element will be replaced by that one. In particular, if the element is in the rejected state, this attribute means that the element pointed to by replacing must be deleted in the resolved structure (a replacing attribute always refers to a child of the most import source. close containing the element having this attribute) • Implicitly: If A immediately precedes an element C, in the resolved structure it will also immediately precede this element C. If an element A is a child of an element B, this means that in the resolved structure element A will also be a child of B.
• markers are just references to elements from the source of an XML import or legacy (a marker always points to a child of the reference nearest created or imported node containing this marker)
• Figure 31 shows an element 10 being an import of an element 1.
• Element 1 has a child 11 which itself contains children 2, 3 and 4, and element 3 contains children 7, 8 and 9
• element 10 contains a child (marker) pointing to 3
• this marker contains an element (12) for example in the imported state followed by a marker pointing to 8.
• the resulting structure (see Figure 32) is therefore as follows: 10 has a child 10:11 which itself contains three children 10: 2, 10: 3 and 10: 4, and 10: 3 contains four children 10: 7, 12, 10: 8 and 10: 9. Note that the element 10:11 is missing from the XML file even in the form of a marker. Finally, if an element must replace an element in the source of an import while being placed elsewhere (the new location is implicitly specified), the reference to the element to be replaced is in the movedRepIacing attribute and replacing n is not present in the item description.
• Each XML legacy defines an XML structure without references, which can be placed as is in a resolved structure (in fact all the elements of a resolved structure will be found either in an XML legacy or in an external (html) page).
• All the elements of an XML legacy will preferably contain an identifier in order to allow XML differences to refer them, for example to indicate where to place a view in this structure.
• the P page without the view will be in an XML legacy
• the paragraph added by the user will be in another legacy XML. Indeed it is not possible to place the new paragraph and the rest of the page in the same XML legacy because they are separated by nodes from elsewhere (the source of the view).
• Figure 33 shows an example of a small website illustrating the role of the different files.
• the replacing attribute only means replacing a node (without its children), it is not necessary to make a copy of 4 (and if 4 changes in the original page in the future, the changes will be reflected in the modified version).
• Figure 34 shows what happens if the user moves 3 as a child from his div node:
• the moved element is placed at this location (if 2 had not been modified, it would have been necessary to create a marker for 2 at the location of this created node.
• Figure 35 presents the case where the user creates elements (for example a paragraph containing an image) inside this element 3.
• markers As already said briefly, a marker always refers to an element of the closest import or creation, unlike references of moved elements which apply exclusively to the nearest import (in fact, to move an element in a structure created, just move it into the XML legacy directly).
• the system makes it possible to create views on pages to other pages. From an XML point of view, this amounts to adding a deported node to the location in question in the XML legacy.
• the system will create a derived page (therefore an import) from the source of this view, containing the modifications made by the user, and change the definition of the view to point it to this new version derivative.
• the user has made modifications on a web page and the system must now build the definition of this derived version.
• the elements of the original file all have a unique identifier, and the system considers that these attributes are preserved when the user makes modifications to the elements.
• placing an element in a certain state means creating an XML element whose name is that of the state in question. Indeed, a production of this algorithm will always go to the XML file of the differences, because it is an import.
• the first step is to establish a table indicating, for each element of H, the element or elements of H 'having the same identifier, then to build the set of elements of H' which either do not have an identifier or have one who is absent from H.
• the second step is based exclusively on the structure (ie which node is the successor / parent of another) and on the correspondences established in the previous step. • Any element of H 'found in case A is placed at the root of the XML structure as a rejected element.
• the element For the elements found in case B or C, the element must be compared with that of the original document (H), that is to say its internal text and its attributes. If this content differs, the element must be placed in the created state and its replacing attribute (or movedReplacing if the previous step had placed it in the moved state) must be pointed to the original element. The new content of the element must be placed in a new XML legacy and pointed to by the ref attribute. Otherwise, if the element was placed in the moved state in the previous step, leave it in this state. Otherwise, if the element had not received a status, it can be deleted from the structure.
• H original document
• the element must be placed in the created state and its replacing attribute (or movedReplacing if the previous step had placed it in the moved state) must be pointed to the original element.
• the new content of the element must be placed in a new XML legacy and pointed to by the ref attribute. Otherwise, if the element was placed in the moved state in the previous step, leave it in this state. Otherwise,
• the parent of an element D is also in case D, they can share the same XML legacy, and only the parent must be present in the XML of the differences. If the predecessor or a child (1) of an element D (2) having been thus hidden in the difference structure is not the case D then it is necessary to create a marker pointing to 2 in the XML legacy, which allows to place 1 in the resolved structure.
• Figure 36 shows an example where five authors of pages derived from an original page suggested their modifications.
• the first user suggested removing the first item from the page 2.
• the second user suggested replacing this first item with another version 3.
• the third user suggested an item to add before the second 4
• the fourth user suggested another item, also to be added before the second. 5.
• the fifth user suggested replacing the first element with yet another version, and in addition to adding a pair of elements (the first of which is a version derived from the second) before the fourth
• Figure 37 shows what the user sees upstream (i.e. the author of the original page):
• the user has the option of accepting either the two pages (the first of which is a version derived from the second) but cannot accept the first without accepting the second.
• a propagated element with a ref attribute pointing to the propagated element will be placed in the place corresponding to the upstream (possibly using markers, if the page upstream is it - even a derived page).
• this element will be placed as a child of the element corresponding to the downstream parent, and immediately before the element corresponding to the downstream successor.
• the propagated element will be placed as a child of this element.
• This algorithm is used both to process pages derived from html pages (or more generally any XML structure without identifiers) and pages derived from XML pages (provided with identifiers), the only difference being the nature of the references to the original page, which are references relating to the position or the content in the first case and references to identifiers in the second case.
• This part of the system is called the Resolver.
• the resolver is a system capable of finding an element given its identifier, and, once an element is known, can return part of the children selected by filters and a maximum depth.
• the content of an element in the resolved structure is defined by its state, that is to say by the name of the element in the difference file. • If it is a created element then the content is the XML node (located in an XML legacy) pointed to by the ref attribute. • If it is a moved, imported or propagated element then the content of this element is the same as that of the element pointed by its ref attribute. Note that when an imported element is encountered, the list of differences described by this import must be stored in an associated structure, as described in the previous section. • If it is a rejected element, this means that the element must not be considered to exist (requesting this element produces an error)
• Movements or creations of elements within this list of children that is to say all the elements immediately preceding (in the difference file) an element of this list of children, and n 'having no (moved) replacing attribute pointing to an existing element at the source of the import 3.
• Movements or creations of elements as children of the element whose children are being constructed that is to say all the elements that must be placed using their parent in the difference file (which can be a marker for example), and whose successor does not exist in the difference file or does not exist at the source of the import, and finally having no (moved) replacing attribute pointing to an existing element at the source of the import. It is therefore necessary to apply all the differences corresponding to one of these criteria, and remove them from the list of differences still to be applied. ...
• the first thing to be done by the resolver is to locate the element whose identifier is given in the request. 123
• glassbox glassbox server, glassbox derivation, delegation to another glassbox, etc.
• glassbox server glassbox derivation, delegation to another glassbox, etc.
• each support group (of a glassbox) stores the initial associated links (denoted R), the union of their incoming links (R " ), and the union of the links before R " ,.
• An import can specify that only one element, its children and its ancestors must be returned, using a derivedOfElt attribute (this is described in more detail in the context of the second system). Here's how it should be handled by the resolver. If we are trying to build a child of an import whose attribute exists and is distinct from the value of ref, we must first find (by a call to the resolver) the element pointing to the target of derivedOfElt, then connect this import element by building the chain of parents until reaching the imported element. If the algorithm seeks to directly reach a descendant of this imported element via the identifier, we can determine if it is a child of derivedOfElt by placing the latter in M and the import in R rather than in M.
• a page i.e. outbound and inbound links
• These sets should be updated regularly.
• a page ie its front and back links
• Figure 38 shows schematically the architecture of the complete system
• 126 D is the symmetric of the algorithm described in the previous section where, instead of R-, R- + and R - + -, we take R +, R + - and R + - + and we evaluate to what extent candidate pages R + - cite the pages of R + cited by the pages of R.
• GlassBoxServer The data structure associated with a system.
• Supports all the support groups of the system.
• SupportGroup a support group, with a reference to the associated support element.
• Req The initial associated links of the support group.
• ReqB the set R " , ie the pages pointing to at least one initial associated link.
• Hub a pivot page for a support group. We store its pivot score in memory.
• ReqBF the set R " + .
• Authority an authority page for a support group. We know its authority score by the links and (in case the page is new) its score by the words.
• ReqBFB the known subset of R "+" (the system searches for internal links on a page only if the page is in the initial links of a support group or has a good authority score for at least a support group)
• Corpus This is the set of known pages in the system.
• Page A page found in the corpus. It has an identifier by which it is referenced from other points in the data structure. Freshness indicates if it is a new page, and transToOldDate indicates if necessary on what date the page will become old.
• F d All the links starting from a page. We also store the weight of the link (for antinepotism)
• Words The word distribution of the page.
• Word A word with a frequency associated with it (the sum of the squares for all the words gives 1)
• AltUrl If the system has detected that several different addresses correspond to the same page, this element will store the other url of the page.
• Friendship Gathers information corresponding to the peer-to-peer network of servers of which this server is a part.
• Neighbors These are all neighboring servers, i.e. to which this server can send requests (see description below).
• FriendServer Describes a neighboring server.
• FriendCache A cache of the most recent and most requested requests, associated with the glassbox server that could have answered them.
• FriendCacheEntry A cache entry, specifying the request and then the server that was able to respond to it, and the date since which this entry was in the cache.
• Us geEntry Gives the number of requests for this url on a given date. The interval between the oldest and the most recent entry in the Usage list is subject to a maximum and when it is about to be exceeded it is necessary to remove the entries that are too old.
• the system operates on the basis of a network of dependencies.
• the system builds a network of dependencies, and according to requests and needs, following the arcs of this network, decides which part of the system should be built / updated.
• dependency target depends on a (called dependency source) whether a change in the data associated with b may imply a change in the data associated with a.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Health & Medical Sciences (AREA)
• Artificial Intelligence (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Computational Linguistics (AREA)
• General Health & Medical Sciences (AREA)
• Physics & Mathematics (AREA)
• General Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
• User Interface Of Digital Computer (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=34575706

Family Applications (1)

Country Status (3)

Families Citing this family (77)

Citations (2)

Family Cites Families (11)

• 2004
  • 2004-10-25 EP EP04805302A patent/EP1719061A2/de not_active Withdrawn
  • 2004-10-25 US US11/667,936 patent/US8495099B2/en active Active
  • 2004-10-25 WO PCT/FR2004/002745 patent/WO2005045698A2/fr active Application Filing

Patent Citations (2)

Non-Patent Citations (4)

Also Published As

Similar Documents

Legal Events