The functions itemdecay(time) and decay(time) can be any functions, but typically monotonically increasing functions are used so the newer events and history gets more weight. The functions are chosen so that the total sum of m(p) over all items is 1 .

∑_p m(p) = \ The relevance calculation process

There are many ways to do the computation described above. For example, the graph can be stored in a memory as an array of link arrays and the power method or any other numeric method can be used.

An example of a formula used in the revisiting random surfer model will now be explained.

The relevance calculation process 690 operates as follows according to an example embodiment. Figure 10a depicts as a flow diagram an example embodiment of the relevance calculation process 690. Information relating to the metadata graph 600 is read 1001 into the memory 702 if the information is not yet stored in the memory. Also the usage count of the whole metadata graph 600 is read 1002 into the memory 702. The total amount m(p) across the nodes is normalized to A. A can be 1 or some other number (for instance, a bigger number, such as amount of nodes^* 100, or 1/10^* max, where max is the maximum number that the system can represent, can be used for A to avoid rounding errors). Another option to avoid rounding errors may be to use logarithmic or some other scale in storage and perform calculations using those.

The algorithm used by the relevance calculation process 690 is, for example, as follows:

PR(p_i) = d₂ *m(p_i) +^l "^{l d2} + d_x

- p_j eLinkTo(pi )

(2)

where pi,p2,---,P/v are the metadata items or nodes, Lin To(pl) is the set of nodes that link to p,, L(p/) is the number of outbound links on node p , and N is the total number of nodes in the graph.

The algorithm works by iteratively calculating 1003 the page rank for one node at the time using the formula in Equation (2), and moving sequentially 1004 over the data set N times. N does not have to be very large, e.g., 20-50 may be sufficient, and the values may not change significantly after that.

Additionally, any other algorithm applicable in this context can be used to compute the PR vector, for example any algorithm presented in "A Survey on PageRank Computing" by Pavel Berkhin.

After the calculations have been performed the relevance values Pr(p) are stored into the database.

The relevance calculation process 690 may be run when there is a change in the metadata network, either because the history has changed, or because items are added, removed or modified. It is possible to speed up the runtime of the algorithm by using one of the various incremental page rank algorithms. It is also possible to use those variations together with this invention.

The relevance calculation process can be run completely in the user device 130, 140, 150, or completely in the server 160, 170, 180 or using some hybrid client-server solution. The relevance calculation process can also be provided as a network service for the user.

In a server or peer-to-peer setting, it is even possible to use the histories of many people in the relevance calculation.

Any optimization to the power method can be used with some embodiments of this invention. It is even possible to use completely different approaches to find the dominant eigenvector of the transition matrix. The results of the relevance calculation process 690 can be regarded as a relevance vector PR of size N in which the sum of items is 1 .0. The items of the relevance vector are stored in the database which can exist e.g. in the memory 702. Incremental Relevance Calculation

Incremental relevance calculation can be used to determine the relevance values for changed nodes, including new nodes, changes in links between nodes and changes to activations of nodes. Since any change in relevance or nodes may result changes throughout the metadata graph, for example any of the following approaches can be used based on how accurate results are needed, and how efficient the incremental process needs to be in terms of needed memory and processing power.

In the following the incremental relevance calculation 695 according to a first example embodiment will be described in more detail with reference to the flow diagram of figure 10b. According to an example embodiment one round of the page rank iteration calculation is run on the network, which consists of changed nodes represented by a changed nodes graph and their immediately linked nodes in the full graph. These immediately linked nodes can be called as border nodes. The iteration uses the current relevance value of the nodes in the full graph (r.relevance). Also the following variables are taken into account: the count of the nodes in the full network (r.linkcount), and the count of links between the full network and the new nodes (r.linksToChangedCount). The relevance values of the changed nodes are updated, but the relevance values of the border nodes are not updated.

The algorithm is illustrated in figure 3, while the pseudocode is illustrated in the following.

Initialize SUBGRAPH as: For each node cn in changed nodes graph CN (labelled with the reference numeral 300 in Fig. 3) do

Get all nodes r of cn. Related, which link to this node either from SUBGRAPH or the metadata network DB;

2 Store r.relevance and r.linkcount of cn. Related to

SUBGRAPH if not there

3. Increment r.linksToChangedCount by one

For each node cn in CN (changed nodes) do

1 . Get all nodes cn. Related, which link to this node from SUBGRAPH

2. Calculate_{DD /} \ , * ,_Λ \ - d_x - d₂ ^

= d₂ * m(cn) H - + f r. relevance

PR (cn) d_l¹

.related ^ .tOtdlC where r.totalC = r.linkcount + r linksToChangedCount .

In the above described embodiment values of any other nodes will not be recomputed, so it may not be very accurate but it can be quite fast.

According to a second example embodiment the first method is applied sequentially over CN and all nodes directly linked to CN. This method can be slightly more accurate than the first method but can be slower than that.

According to the third example embodiment the relevance calculation is performed as follows. Similarly to the first method the SUBGRAPH is constructed. Initially the SUBGRAPH contains all changed nodes. Then SUBGRAPH is expanded by following links from current nodes of SUBGRAPH to the full graph. A criterion for stopping the expansion can follow the dampening factor for the relevance computation by iteratively applying (d1 +d2)/s.outlinksCount to the nodes linked from s (a node in SUBGRAPH; s.outlinksCount is a function of s, e.g. the amount of outlinks from s) so that each link decreases the effect of change in CN, and when the effect becomes smaller than a threshold or the SUBGRAPH larger than a size threshold, expansion is stopped. Other criteria (e.g., the previously computed relevance value) can be used as well. The rest of the nodes (nodes in full graph but not in SUBGRAPH) from the full graph are aggregated into one node A and a transition matrix G^* for the union of SUBGRAPH and A is formed. The aggregation of markov chain states is a well-known principle in literature. Then, using e.g. the power method, the static end-state distribution PR^' of the transition matrix G^* is calculated. The resulting PR^' may be scaled so that the probability distribution matches the full graph. Finally the relevance values PR^' are saved. This method can be slightly more accurate than the first method but can be slower than that.

According to a fourth example embodiment the full relevance calculation 690 is performed as disclosed above but all the changed relevance values may not be stored. The decision on which relevance values will be stored may be based on, for example, the quantity of change of a relevance value. In other words, the changed relevance value of a node is compared with a previous relevance value of that node and if the change exceeds a certain threshold, the re-calculated relevance value will be stored into memory. Otherwise the relevance value is not changed. This approach may provide quite accurate results but may slow down the calculation and conserve more memory than the first and the second approach.

It should be noted here that the above described methods for the incremental relevance calculation are only examples but also other methods appropriate for this purpose may be used as well.

User Interface Processes The relevance value of the items can be used in various ways in a user interface 201— 204. The following sections can be considered as examples of user interfaces, but the invention may also be applicable in other user interfaces, which use the relevance as filtering, sorting, emphasizing or clustering criteria. Each of the usages is depicted in the example user interface wireframe in the example depicted in figure 4. It should be noted that in a real user interface, not all of these usages of relevance are necessarily utilized simultaneously, for instance the sorting could be based on time. In the user interface of figure 4 some imaginary notifications 401— 406 are depicted demonstrating the various ways the computed relevance can be used. Each of the rows and even the subitems, such as thumbnails, can be activated e.g., with a pointing device, leading into a detailed view of that item, with possible actions. In this example events are sorted so that the most relevant ones are on top. Some of the items have been emphasized. For example, Marie's photos 403 have a preview because they have been detected to be more relevant to John's photos 405 which are not provided by preview images. Further, some less relevant items from various persons have been grouped into a single area 406 (clustering). Such notifications are not shown whose relevance is below a threshold value. This is illustrated as an empty rectangle 407 in figure 4. Filtering by Relevance

The computed relevance value can be used as a filter in the user interface for example using one or more of the following exemplary ways of filtering:

A pre-defined cut-off (threshold) value may be used, which can also be a function of the network size and/or history size and the recentness (decay). The filtering is possible to be implemented so that a certain percentage of the items are shown. For instance, 25% of the most interesting items could always be shown. It is also possible to use a variable cut-off value. For instance, based on statistical properties of the relevance distribution within the network, the cut-off value can be determined dynamically. One option is to use the probability P_gi₀bai_read, which is the global probability of the user reading an update. For instance, if the user has read 40 % of the updates in the past, P_gi₀bai_read would be 0.4 and 40% of the most interesting updates would be shown. The other option is to give the user the possibility to filter more or less on demand.

For example, relevance filtering can be used in a user interface to filter out states, which do not refer to relevant items or while summarizing. Also, one option for the summarization would be the visual representation of the most relevant state in that cluster.

Clustering by Relevance

Similar items can be collected together to form a cluster so that only information of the cluster is shown. The clustering process can be affected by relevance values. Most relevant items would go to separate clusters, and the less relevant items could be clustered together. E.g., "John called you once at 15:15, and Mary called you twice at 16:03, and there are 4 other phone calls.". In that example, there are three clusters and the 4 phone calls which have a lower relevance value are collected in the same cluster.

Emphasizing by Relevance

Relevance can be used to alter to visualization of the items or the clusters and summaries of item clusters. First, after grouping or clustering items, the relevance value of the items within a cluster can be used to determine how the cluster is represented. For instance, only the title (summary) of the cluster could be shown ("Marie has 16 new updates in a social music site"), or some amount of the items (1 ..n, n being the size of the cluster) could be shown fully depending on relevance.

Second, single items could be rendered differently based on the relevance value. For instance, very relevant shared images from a favourite contact could be shown with the image title in addition to the thumbnail, while less relevant images would be shown just as thumbnails.

Sorting by Relevance Relevance can be used when sorting a list of items. For instance, search results could be sorted so that the most relevant items would be shown first. The user interface effect, which follows from the fact that the user's history is used in the relevance calculation, is that items, which are similar to what the user typically views, may be shown on top of the search, also even if the user has never viewed them before.

For example, an example of the search depicted in Figure 5a uses the sorting by relevance. For instance, if the user have often opened the image "Mac User", that image will have a high relevance value, and will be on top of the search hits, which are sorted by relevance, even if the user just presses "m".

Another example user interface using relevance is depicted in figure 5b, which shows updates differently based on their relevance, and filters irrelevant updates. It should be noted that the examples in figures 5a and 5b are rather an illustration on how relevance calculation can be used in a real user interface.

Using the Relevance to Determine Caching

Some of the items may be stored in a cache memory 709 of the device 700. The cache can either have a static size or the size can vary based on available local resources or even usage patterns. Each binary object cache entry may be appended with a relevance value, which is copied from the item that references this binary item. If there are many items referencing the same item, a maximum or average of these values may be used. The cache can also implement a time to live value, which can be used. The time to live can override the relevance value or the relevance value can override the time to live value. In the below examples time to live overrides the relevance value:

In a first example situation the cache is full and a new binary object is loaded. From the cached items, whose time to live is expired, evict from the cache the binary object whose cache value cache(object) is the smallest

In a second example situation an automatic pre-caching is performed before going offline, or when a cheap network is detected. The pre-caching is performed for such cached items, whose time to live has been expired. If there are items in the system whose relevance is higher and whose binary objects are not currently in cache, those binary objects are preloaded from the network and necessary amount of cached objects which have the lowest relevance and expired time to live are deleted from the cache.

Using the Relevance to Determine Synchronization Priorities

Since the relevance value models the global probability that the user will activate and display that item, the solution is the use relevance to determine the set of synchronized items so that the most relevant items are synchronized with higher probability.

One option of selecting the items to be synchronized is as follows: If K_aii is the set of all items considered to be synchronized, and the changed items is Kchanged, and N is the amount of items that can be synchronized because of time or bandwith limits for example, then K_synchronized is selected using e.g. the following formulas in which d is in the range of 0...1 and denotes how much new items will be emphasized in the synchronization.

Nnew = d^* (N - length(K_changed));

Irrelevant = (1 -d)^* (N - length(K_changed)); Knew = Nnew first items Of (Kail INTERSECT (Krelevant U K_changed)), SOrted ΪΠ timestamp order, newest first; Relevant = N_relevant first items Of (K_a|| INTERSECT (Knew U K_changed)), SOrted ΪΠ relevance order, most relevant first;

Ksynchronized^— Kchanged U Krelevant U K_new- Current user interfaces allow some customization by the user (e.g., which types of items to show), but they do not automatically adapt of personalize to the user's behaviour. By using the present invention the user interface of the device can be automatically personalized and adapted based on the user's habits. It also generalizes the habits to other similar items by using the metadata connections. Therefore it is more likely that the information, which is important or relevant to the user is noticed and found by the user.

One advantage of incremental relevance calculation is that the relevance value can now be used for much wider amount of use cases since the full relevance calculation, which may require much computing power, is not required. For instance, newly arrived items can be filtered for relevance immediately.

The advantage of the mathematically modelled approach is that the algorithm may always converge, and any optimizations described in the literature can be used.

The advantage of caching based on relevance is that the cache is also personalized to the user's usage habits. It can also precache items that are predicted to be needed by the user in the near future.

The advantage of synchronization based on relevance is that since only the relevant items are synchronized, battery and network and time may be saved compared to full synchronization. Also, the server-side changes to the items that are relevant to the user can be synchronized faster and more reliably. The combinations of claim elements as stated in the claims can be changed in any number of different ways and still be within the scope of various embodiments of the invention. As used in this application, the term 'circuitry' refers to all of the following:

(a) to hardware-only circuit implementations (such as implementations in only analog and/or digital circuitry) and

(b) to combinations of circuits and software (computer programs) (and/or firmware), such as: (i) to a combination of processor(s) or (ii) to portions of processor(s)/software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone, a server, a computer, a music player, an audio recording device, etc, to perform various functions) and

(c) to circuits, such as a microprocessor(s) or a portion of a microprocessor(s), that require software or firmware for operation, even if the software or firmware is not physically present.

This definition of 'circuitry' applies to all uses of this term in this application, including in any claims. As a further example, as used in this application, the term "circuitry" would also cover an implementation of merely a processor (or multiple processors) or portion of a processor and its (or their) accompanying software and/or firmware. The term "circuitry" would also cover, for example and if applicable to the particular claim element, a baseband integrated circuit or applications processor integrated circuit for a mobile phone or a similar integrated circuit in server, a cellular network device, or other network device.

The computer programs may be stored in the memory of the user device (e.g. a terminal, a mobile terminal, a wireless terminal etc.) and/or in the memory of the server, for example. The computer program may also be stored in a data carrier such as a memory stick, a CDROM, a digital versatile disk, a flash memory etc.

Claims

Claims:

1 . A method comprising:

- nodes representing the items; and

- connections between nodes representing links between the items;

- said determining a relevance value comprising using information of previous activations of nodes.

2. The method according to claim 1 comprising:

- collecting information regarding items the user has selected for viewing by a device; and

- using the collected information in determining the relevance value.

3. The method according to claim 1 or 2 comprising:

- examining the relevance values;

- selecting from the metadata graph one or more items having the highest relevance values; and

- providing information on the selected items to the user.

4. The method according to claim 3 comprising clustering items of which information is provided to the user on the basis of the relevance values.

5. The method according to claim 4 comprising:

- clustering items of the highest relevance values into separate clusters; and

- clustering items of lower relevance values and having similar type into the same cluster.

6. The method according to any of the claims 1 to 5 comprising calculating the relevance value by using the equation

in which

di is a first probability,

62 is a second probability,

p is a node index,

m(p) is a value indicative of the previous access frequency of the node

P.

N is the total number of nodes in the metadata graph

P ,P2,—,PN are the metadata items,

LinkTo(pi) is the set of nodes that link to p,, and

L(pj) is the number of outbound links on node p ,

7. The method according to claim 6 comprising:

- repeating the calculation of the relevance value for each node in the metadata graph; and

- repeating the calculation of the relevance values of the metadata graph at least once.

8. The method according to any of the claims 1 to 7 comprising recalculating at least one relevance value when there is a change in the metadata graph.

9. The method according to claim 8 wherein the change in the metadata graph includes at least one of the following:

- information of previous user activations of nodes has changed, or - an item or a link between items is added, removed or modified.

10. The method according to claim 8 or 9, wherein the recalculation is performed by using the equation

_{DD /} \ , *_{( Λ} l - d_l - d₂ ^ ( r. relevance

PR (cn) = d₂ * m(cn) H ^ + α ^

N r in cn. related -tOtdlC in which

di is a first probability,

62 is a second probability, cn is the changed node,

m(cn) is a value indicative of the previous access frequency of the node cn,

N is the total number of nodes in the metadata graph

r.relevance is the current relevance value of the node in the metadata graph,

r.linkcount is the count of the nodes in the metadata graph,

r.linksToChangedCount is the count of links between the metadata graph and new nodes,

r.totalC = r.linkcount + r linksToChangedCount

1 1 . The method according to any of the claims 1 to 10 comprising:

- storing a copy of information of at least one item of the metadata graph in a device;

- determining, whether said information needs to be synchronized with the original information of the at least one item; and

- using the relevance value to select the items to be synchronized.

12. The method according to any of the claims 1 to 1 1 comprising:

- determining a threshold value; and

- using the threshold value to select items to be provided to a user interface process.

13. The method according to claim 12 comprising:

- calculating the number of items in the metadata graph; and

- using the threshold value as a percentage value, wherein the number of items to be provided to the user interface process of all items in the metadata graph is determined by the threshold value; and

- selecting those items which have the highest relevance value to be provided to the user interface process.

14. The method according to claim 12 comprising:

- comparing the threshold value with the relevance value of an item in the metadata graph; and

- providing the item to the user interface process if the comparison indicates that the relevance value exceeds the threshold value.

15. The method according to any of the claims 1 to 14 comprising using the relevance value in determining how much information of an item is provided to a user interface process.

16. The method according to claim 15 comprising:

- providing all information of items having the highest relevance values to the user interface process; and

- providing only a part of the information of items having lower relevance values to the user interface process.

17. The method according to any of the claims 1 to 16 comprising providing a cache in a memory of a device for storing information of items of the metadata graph.

18. The method according to claim 17 comprising using the relevance value of an item to determine whether the information of the item is stored in the cache.

19. The method according to any of the claims 17 to 18 comprising providing a time to live value for items stored in the cache, wherein the time to live value, the relevance value or both the time to live value and the relevance value are used to determine whether the item will be replaced by another item or maintained in the cache.

20. The method according to any of the claims 1 to 19 comprising storing the metadata graph in at least one of the following apparatuses:

- a user device;

- a server.

21 . An apparatus comprising:

- nodes representing the items; and

22. The apparatus according to claim 21 comprising:

- means for collecting information regarding items the user has selected for viewing by a device; and

- means for using the collected information in determining the relevance value.

23. The apparatus according to claim 21 or 22 comprising:

- means for examining the relevance values;

- means for selecting from the metadata graph one or more items having the highest relevance values; and

- means for providing information on the selected items to the user.

24. The apparatus according to claim 23 comprising means for clustering items of which information is provided to the user on the basis of the relevance values.

25. The apparatus according to claim 24, said means for clustering being configured for:

- clustering items of the highest relevance values into separate clusters; and

26. The apparatus according to any of the claims 21 to 25 comprising means for calculating the relevance value by using the equation

in which di is a first probability,

02 is a second probability,

p is a node index,

m(p) is a value indicative of the previous access frequency of the node P.

N is the total number of nodes in the metadata graph

P ,P2,—,PN are the metadata items,

LinkTo(pi) is the set of nodes that link to p,, and

L(pj) is the number of outbound links on node p ,

27. The apparatus according to claim 26 configured for:

28. The apparatus according to any of the claims 21 to 27 configured for recalculating at least one relevance value when there is a change in the metadata graph.

29. The apparatus according to claim 28 wherein the change in the metadata graph includes at least one of the following:

- information of previous transitions between nodes has changed, or

- an item is added, removed or modified.

30. The apparatus according to claim 28 or 29 configured for performing the recalculation by using the equation

_{D D / \} , *_{( Λ} l - d_l - d₂ ^ ( r. relevance PR (cn) = d₂ * m(cn) H - + d_l ^

N r in cn. related -tOtdlC J in which

di is a first probability,

d2 is a second probability,

cn is the changed node,

m(cn) is a value indicative of the previous access frequency of the node cn,

N is the total number of nodes in the metadata graph r.relevance is the current relevance value of the node in the metadata graph,

r.linkcount is the count of the nodes in the metadata graph,

r.totalC = r.linkcount + r linksToChangedCount

31 . The apparatus according to any of the claims 21 to 30 comprising:

- means for storing a copy of information of at least one item of the metadata graph in a device;

- means for determining, whether said information needs to be synchronized with the original information of the at least one item; and

- means for using the relevance value to select the items to be synchronized.

32. The apparatus according to any of the claims 21 to 31 comprising:

- a threshold value; and

- means for using the threshold value to select items to be provided to a user interface process.

33. The apparatus according to claim 32 comprising:

- means for calculating the number of items in the metadata graph; and

- means for using the threshold value as a percentage value, wherein the number of items to be provided to the user interface process of all items in the metadata graph is determined by the threshold value; and

- means for selecting those items which have the highest relevance value to be provided to the user interface process.

34. The apparatus according to claim 32 comprising:

- means for comparing the threshold value with the relevance value of an item in the metadata graph; and

- means for providing the item to the user interface process if the comparison indicates that the relevance value exceeds the threshold value.

35. The apparatus according to any of the claims 21 to 34 comprising means for using the relevance value in determining how much information of an item is provided to a user interface process.

36. The apparatus according to claim 35 comprising:

- means for providing all information of items having the highest relevance values to the user interface process; and

- means for providing only a part of the information of items having lower relevance values to the user interface process.

37. The apparatus according to any of the claims 21 to 36 comprising a cache in a memory of a device for storing information of items of the metadata graph.

38. The apparatus according to claim 37 comprising means for using the relevance value of an item to determine whether the information of the item is stored in the cache.

39. The apparatus according to any of the claims 37 to 38 comprising a time to live value for items stored in the cache, wherein the apparatus is configured to use the time to live value, the relevance value or both the time to live value and the relevance value to determine whether the item will be replaced by another item or maintained in the cache.

40. The apparatus according to any of the claims 21 to 39, wherein it is a part of a user device or a server.

41 . A computer program product stored on a storage medium comprising a computer program code configured to, with at least one processor, cause an apparatus to:

- nodes representing the items; and

- connections between nodes representing links between the items; - determine a relevance value for a node of the metadata graph regarding a probability of the user activating that node after a number of activations;

42. The computer program according to claim 41 comprising computer instructions for:

- collecting information regarding items the user has selected for viewing by the apparatus; and

- using the collected information in determining the relevance value.

43. The computer program according to claim 41 or 42 comprising computer instructions for:

- examining the relevance values;

- providing information on the selected items to the user.

44. The computer program according to claim 43 comprising computer instructions for clustering items of which information is provided to the user on the basis of the relevance values.

45. The computer program according to claim 44 comprising computer instructions for:

- clustering items of the highest relevance values into separate clusters; and

46. The computer program according to any of the claims 41 to 45 comprising computer instructions for calculating the relevance value by using the equation

in which

di is a first probability,

62 is a second probability,

p is a node index,

m(p) is a value indicative of the previous access frequency of the node

P.

N is the total number of nodes in the metadata graph

P ,P2,—,PN are the metadata items,

LinkTo(pi) is the set of nodes that link to p,, and

L(pj) is the number of outbound links on node p ,

47. The computer program according to claim 46 comprising computer instructions for:

48. The computer program according to any of the claims 41 to 47 comprising computer instructions for recalculating at least one relevance value when there is a change in the metadata graph.

49. The computer program according to claim 48 comprising computer instructions for performing the recalculation when:

- information of previous transitions between nodes has changed, or

- an item is added, removed or modified.

50. The computer program according to claim 48 or 49 comprising computer instructions for performing the recalculation by using the equation

_{DD /} \ , *_{( Λ} l - d_l - d₂ ^ ( r. relevance PR (cn) = d₂ * m(cn) H ^ + α '

N_r i_n a .related V .tOtCllC in which

di is a first probability, 02 is a second probability,

cn is the changed node,

rn(cn) is a value indicative of the previous access frequency of the node cn,

N is the total number of nodes in the metadata graph

r.relevance is the current relevance value of the node in the metadata graph,

r.linkcount is the count of the nodes in the metadata graph,

r.totalC = r.linkcount + r linksToChangedCount

51 . The computer program according to any of the claims 41 to 50 comprising computer instructions for:

- using the relevance value to select the items to be synchronized.

52. The computer program according to any of the claims 41 to 51 comprising:

- a threshold value; and

- computer instructions for using the threshold value to select items to be provided to a user interface process.

53. The computer program according to claim 52 comprising:

- computer instructions for calculating the number of items in the metadata graph; and

- computer instructions for using the threshold value as a percentage value, wherein the number of items to be provided to the user interface process of all items in the metadata graph is determined by the threshold value; and

- computer instructions for selecting those items which have the highest relevance value to be provided to the user interface process.

54. The computer program according to claim 52 comprising: - computer instructions for comparing the threshold value with the relevance value of an item in the metadata graph; and

- computer instructions for providing the item to the user interface process if the comparison indicates that the relevance value exceeds the threshold value.

55. The computer program according to any of the claims 41 to 54 comprising computer instructions for using the relevance value in determining how much information of an item is provided to a user interface process.

56. The computer program according to claim 55 comprising:

- computer instructions for providing all information of items having the highest relevance values to the user interface process; and

- computer instructions for providing only a part of the information of items having lower relevance values to the user interface process.

57. The computer program according to any of the claims 41 to 56 comprising a cache in a memory of a device for storing information of items of the metadata graph.

58. The computer program according to claim 57 comprising computer instructions for using the relevance value of an item to determine whether the information of the item is stored in the cache.

59. The computer program according to any of the claims 57 to 58 comprising a time to live value for items stored in the cache, wherein the computer program is configured to use the time to live value, the relevance value or both the time to live value and the relevance value to determine whether the item will be replaced by another item or maintained in the cache.

60. The computer program according to any of the claims 41 to 59, wherein the computer program is stored in at least one of the following:

- a user device;

- a server;

- a memory;

- a data carrier.

. An apparatus comprising:

- nodes representing the items; and

- connections between nodes representing links between the items;

- determine a relevance value for a node of the metadata graph regarding a probability of the user activating that node after a number of activations; and

- use information of previous activations of nodes in the determination of the relevance value.

62. A data structure comprising:

- nodes representing the items; and

- connections between nodes representing links between the items;

63. A network service comprising:

- nodes representing the items; and

- connections between nodes representing links between the items; deternnining a relevance value for a node of the metadata graph regarding a probability of the user activating that node after a lot of activations;

said determining a relevance value comprising using information of previous activations of nodes. wireless terminal comprising:

means for providing a metadata graph of interconnected items in a network relating to a user, the metadata graph comprising

- nodes representing the items; and

- connections between nodes representing links between the items;

means for determining a relevance value for a node of the metadata graph regarding a probability of the user activating that node after a number of activations;

said means for determining a relevance value are configured for using information of previous activations of nodes in the determination of the relevance value.