Patent Application
20060095852
This invention relates to information retrieval apparatus and methods.
There are many established systems for locating information (e.g. documents, images, emails, patents, internet content or media content such as audio/video content) by searching under keywords. Examples include internet search “engines” such as those provided by “Google”™ m or “Yahoo”™ m where a search carried out by keyword leads to a list of results which are ranked by the search engine in order of perceived relevance.
However, in a system encompassing a large amount of content, often referred to as a massive content collection, it can be difficult to formulate effective search queries to give a relatively short list of search “hits”. For example, at the time of preparing the present application, a Google search on the keywords “massive document collection” drew 243000 hits. This number of hits would be expected to grow if the search were repeated later, as the amount of content stored across the internet generally increases with time. Reviewing such a list of hits can be prohibitively time-consuming.
In general, some reasons why massive content collections are not well utilised are:
The paper “Self Organisation of a Massive Document Collection”, Kohonen et al, IEEE Transactions on Neural Networks, Vol 11, No. 3, May 2000, pages 574-585 discloses a technique using so-called “self-organising maps” (SOMs). These make use of so-called unsupervised self-learning neural network algorithms in which “feature vectors” representing properties of each document are mapped onto nodes of a SOM.
In the Kohonen et al paper, a first step is to pre-process the document text, and then a feature vector is derived from each pre-processed document. In one form, this may be a histogram showing the frequencies of occurrence of each of a large dictionary of words. Each data value (i.e. each frequency of occurrence of a respective dictionary word) in the histogram becomes a value in an n-value vector, where n is the total number of candidate words in the dictionary (43222 in the example described in this paper). Weighting may be applied to the n vector values, perhaps to stress the increased relevance or improved differentiation of certain words.
The n-value vectors are then mapped onto smaller dimensional vectors (i.e. vectors having a number of values m (500 in the example in the paper) which is substantially less than n. This is achieved by multiplying the vector by an (n×m) “projection matrix” formed of an array of random numbers. This technique has been shown to generate vectors of smaller dimension where any two reduced-dimension vectors have much the same vector dot product as the two respective input vectors. This vector mapping process is described in the paper “Dimensionality Reduction by Random Mapping: Fast Similarity Computation for Clustering”, Kaski, Proc IJCNN, pages 413-418, 1998.
The reduced dimension vectors are then mapped onto nodes (otherwise called neurons) on the SOM by a process of multiplying each vector by a “model” (another vector). The models are produced by a learning process which automatically orders them by mutual similarity onto the SOM, which is generally represented as a two-dimensional grid of nodes. This is a non-trivial process which took Kohonen et al six weeks on a six-processor computer having 800 MB of memory, for a document database of just under seven million documents. Finally the grid of nodes forming the SOM is displayed, with the user being able to zoom into regions of the map and select a node, which causes the user interface to offer a link to an internet page containing the document linked to that node.
Various aspects and features of the present invention are defined in the appended claims.
According to one aspect of the present invention these is provided an information retrieval apparatus for searching a set of information items and displaying the results of the search, the information items each having a set of characterising information features. The apparatus comprises a search processor operable to search the information items in accordance with a user defined characterising information feature and to identify information items having characterising information features corresponding to that user defined characterising information feature. A mapping processor operable to generate data representative of a map of information items from a set of information items identified in the search. The map provides the identified information items with respect to positions in an array in accordance with a mutual similarity of the information items, similar information items mapping to similar positions in the array. The apparatus includes a graphical user interface for displaying a representation of at least some of the identified items, and a user control for selecting an identified information item. The search processor is operable to refine the search to identify information items relating to the selected information item. As such the user is provide with a facility for refining a search by identifying desired information items with respect to items revealed as positions within the array which are mutually similar. In combination with the arrangement of the information items with respect to the positions in the array, which are displayable to a user, a facility for searching and refining the results of the search is provided. Furthermore, navigation of information items is facilitated, which is particularly advantageous when the amount of information items is large.
The present invention addresses a technical problem of defining a search query for search information items and for refining a search for information items, which particularly advantageous for searching a navigating large amounts of data.
The characterising information features may include metadata describing the content or attributes of the information items, video images or audio signals or audio metadata, or a combination of these types of characterising information features. As such, in order to refine the search of information items, the user control may be operable to select the identified information item in accordance with metadata, video images or audio metadata associated with the identified item.
The search processor may be operable to search the set of information items for information items including the same and/or similar metadata, the same and/or a similar video image, or the same and/or similar audio metadata. For example, the search processor may compare a feature vector formed from the metadata associated with a selected information item, to find the position in the array which is closest to this feature vector. The search processor is then operable to search the set of information items in relation to the metadata associated with the user selected identified information item. The search processor identifies information items from the set, which are within a predetermined number of positions within the array from the position in the array which is closest to the feature vector.
The user control may be operable to provide a user with a facility for selecting a plurality of identified information items and to specify a search relationship between the identified items in accordance with Boolean logic. As a result, a plurality of charactering features may be combined to form a search query in accordance with the Boolean operators specified the user. As such a more focused search can be performed, which is directed to the information item, which is of interest to a user.
According to another aspect of the invention there is provided a graphical user interface, comprising rows of fields for selecting information items from a set of information items. Each row defines a search condition for forming a search query, in accordance with a different type of characterising information feature associated wioth each row. If more than one information feature is provided in each row, then the conditions for the search are specified by Boolean operators. Accordingly a user may specify a search query in accordance with the information items elected in different rows of the interface.
Further respective aspects and features of the invention are defined in the appended claims.
Embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings in which:
a and 3b schematically illustrate term frequency histograms;
a schematically illustrates a raw feature vector;
b schematically illustrates a reduced feature vector;
FIGS. 7 to 9 schematically illustrate display screens providing a user interface to access information represented by the SOM;
FIGS. 18(a) and 18(b) provide illustrative bar graphs providing two examples of colour histograms for two video images forming a search query, and
The storage system operates in two general modes of operation. In a first mode, a set of information items (e.g. textual information items) is assembled on the disk storage 30 or on a network disk drive connected via the network 50 and is sorted and indexed ready for a searching operation. The second mode of operation is the actual searching against the indexed and sorted data.
The embodiments are applicable to many types of information items. A non-exhaustive list of appropriate types of information includes patents, video material, emails, presentations, internet content, broadcast content, business reports, audio material, graphics and clipart, photographs and the like, or combinations or mixtures of any of these. In the present description, reference will be made to textual information items. The textual information items may be associated with, or linked to, non-textual items. So, for example, audio and/or video material may be associated with “MetaData” which is a textual information item defining that material in textual terms.
The information items are loaded onto the disk storage 30 in a conventional manner. Preferably, they are stored as part of a database structure which allows for easier retrieval and indexing of the items, but this is not essential. Once the information and items have been so stored, the process used to arrange them for searching is shown schematically in
It will be appreciated that the indexed information items need not be stored on the local disk drive 30. The information items could be stored on a remote drive connected to the system 10 via the network 50. Alternatively, information may be stored in a distributed manner, for example at various sites across the internet. If the information is stored at different internet or network sites, a second level of information storage could be used to store locally a “link” (e.g. a Universal Resource Identifier URI) to the remote information, perhaps with an associated summary, abstract or metadata associated with that link. So, the remotely held information would not be accessed unless the user selected the relevant link (e.g. from the results list 260 to be described below), although for the purposes of the technical description which follows, the remotely held information, or the abstract/summary/metadata, or the link/URI could be considered as the “information item”.
In other words, a formal definition of the “information item” is an item from which a feature vector is derived and processed (see below) to provide a mapping to the SOM. The data shown in the results list 260 (see below) may be the actual information item which a user seeks (if it is held locally and is short enough for convenient display) or may be data representing and/or pointing to the information item, such as one or more of metadata, a URI, an abstract, a set of key words, a representative key stamp image or the like. This is inherent in the operation “list” which often, though not always, involves listing data representing a set of items. The data representing information items may include different types of information. The types of information of each information item and the data representing each type will be referred to as characterising information features.
In a further example, the information items could be stored across a networked work group, such as a research team or a legal firm. A hybrid approach might involve some information items stored locally and/or some information items stored across a local area network and/or some information items stored across a wide area network. In this case, the system could be useful in locating similar work by others, for example in a large multi-national research and development organisation, similar research work would tend to be mapped to similar output nodes in the SOM (see below). Or, if a new television programme is being planned, the present technique could be used to check for its originality by detecting previous programmes having similar content.
It will also be appreciated that the system 10 of
The processes are not limited to particular numbers of information items.
The process of generating a self-organising map (SOM) representation of the information items will now be described with reference to FIGS. 2 to 6.
Feature extraction is the process of transforming raw data into an abstract representation. These abstract representations can then be used for processes such as pattern classification, clustering and recognition. In this process, a so-called “feature vector” is generated, which is an abstract representation of the frequency of terms used within a document.
The process of forming the visualisation through creating feature vectors includes:
Considering these steps in more detail, each document (information item) 100 is opened in turn. At a step 110, all “stop words” are removed from the document. Stop-words are extremely common words on a pre-prepared list, such as “a”, “the”, “however”, “about”, “and”, and “the”. Because these words are extremely common they are likely, on average, to appear with similar frequency in all documents of a sufficient length. For this reason they serve little purpose in trying to characterise the content of a particular document and should therefore be removed.
After removing stop-words, the remaining words are stemmed at a step 120, which involves finding the common stem of a word's variants. For example the words “thrower”, “throws”, and “throwing” have the common stem of “throw”.
A “dictionary” of stemmed words appearing in the documents (excluding the “stop” words) is maintained. As a word is newly encountered, it is added to the dictionary, and running count of the number of times the word has appeared in the whole document collection (set of information items) is also recorded.
The result is a list of terms used in all the documents in the set, along with the frequency with which those terms occur. Words that occur with too high or too low a frequency are discounted, which is to say that they are removed from the dictionary and do not take part in the analysis which follows. Words with too low a frequency may be misspellings, made up, or not relevant to the domain represented by the document set. Words that occur with too high a frequency are less appropriate for distinguishing documents within the set. For example, the term “News” is used in about one third of all documents in a test set of broadcast-related documents, whereas the word “football” is used in only about 2% of documents in the test set. Therefore “football” can be assumed to be a better term for characterising the content of a document than “News”. Conversely, the word “fottball” (a misspelling of “football”) appears only once in the entire set of documents, and so is discarded for having too low an occurrence. Such words may be defined as those having a frequency of occurrence which is lower than two standard deviations less than the mean frequency of occurrence, or which is higher than two standard deviations above the mean frequency of occurrence.
A feature vector is then generated at a step 130.
To do this, a term frequency histogram is generated for each document in the set. A term frequency histogram is constructed by counting the number of times words present in the dictionary (pertaining to that document set) occur within an individual document. The majority of the terms in the dictionary will not be present in a single document, and so these terms will have a frequency of zero. Schematic examples of term frequency histograms for two different documents are shown in
It can be seen from this example how the histograms characterise the content of the documents. By inspecting the examples it is seen that document 1 has more occurrences of the terms “MPEG” and “Video” than document 2, which itself has more occurrences of the term “MetaData”. Many of the entries in the histogram are zero as the corresponding words are not present in the document.
In a real example, the actual term frequency histograms have a very much larger number of terms in them than the example. Typically a histogram may plot the frequency of over 50000 different terms, giving the histogram a dimension of over 50000. The dimension of this histogram needs to be reduced considerably if it is to be of use in building an SOM information space.
Each entry in the term frequency histogram is used as a corresponding value in a feature vector representing that document. The result of this process is a (50000×1) vector containing the frequency of all terms specified by the dictionary for each document in the document collection. The vector may be referred to as “sparse” since most of the values will typically be zero, with most of the others typically being a very low number such as 1.
The size of the feature vector, and so the dimension of the term frequency histogram, is reduced at a step 140. Two methods are proposed for the process of reducing the dimension of the histogram.
The method selected for reducing the dimension of the term frequency histogram in the present embodiment is “random mapping”, as explained in detail in the Kaski paper referred to above. Random mapping succeeds in reducing the dimension of the histogram by multiplying it by a matrix of random numbers.
As mentioned above, the “raw” feature vector (shown schematically in
In fact as the dimension of the vector increases any given set of randomly generated vectors are nearly orthogonal to each other. This property means that the relative direction of vectors multiplied by this a matrix of random numbers will be preserved. This can be demonstrated by showing the similarity of vectors before and after random mapping by looking at their dot product.
It can be shown experimentally that by reducing a sparse vector from 50000 values to 200 values preserves their relative similarities. However, this mapping is not perfect, but suffices for the purposes of characterising the content of a document in a compact way.
Once feature vectors have been generated for the document collection, thus defining the collection's information space, they are projected into a two-dimensional SOM at a step 150 to create a semantic map. The following section explains the process of mapping to 2-D by clustering the feature vectors using a Kohonen self-organising map. Reference is also made to
A Kohonen Self-Organising map is used to cluster and organise the feature vectors that have been generated for each of the documents.
A self-organising map consists of input nodes 170 and output nodes 180 in a two-dimensional array or grid of nodes illustrated as a two-dimensional plane 185. There are as many input nodes as there are values in the feature vectors being used to train the map. Each of the output nodes on the map is connected to the input nodes by weighted connections 190 (one weight per connection).
Initially each of these weights is set to a random value, and then, through an iterative process, the weights are “trained”. The map is trained by presenting each feature vector to the input nodes of the map. The “closest” output node is calculated by computing the Euclidean distance between the input vector and weights associated with each of the output nodes.
The closest node, identified by the smallest Euclidean distance between the input vector and the weights associated with that node is designated the “winner” and the weights of this node are trained by slightly changing the values of the weights so that they move “closer” to the input vector. In addition to the winning node, the nodes in the neighbourhood of the winning node are also trained, and moved slightly closer to the input vector.
It is this process of training not just the weights of a single node, but the weights of a region of nodes on the map, that allow the map, once trained, to preserve much of the topology of the input space in the 2-D map of nodes.
Once the map is trained, each of the documents can be presented to the map to see which of the output nodes is closest to the input feature vector for that document. It is unlikely that the weights will be identical to the feature vector, and the Euclidean distance between a feature vector and its nearest node on the map is known as its “quantisation error”.
By presenting the feature vector for each document to the map to see where it lies yields an x, y map position for each document. These x, y positions when put in a look up table along with a document ID can be used to visualise the relationship between documents.
Finally, a dither component is added at a step 160, which will be described with reference to
A potential problem with the process described above is that two identical, or substantially identical, information items may be mapped to the same node in the array of nodes of the SOM. This does not cause a difficulty in the handling of the data, but does not help with the visualisation of the data on display screen (to be described below). In particular, when the data is visualised on a display screen, it has been recognised that it would be useful for multiple very similar items to be distinguishable over a single item at a particular node. Therefore, a “dither” component is added to the node position to which each information item is mapped. The dither component is a random addition of ±½ of the node separation. So, referring to
So, the information items can be considered to map to positions on the plane of
At any time, a new information item can be added to the SOM by following the steps outlined above (i.e. steps 110 to 140) and then applying the resulting reduced feature vector to the “pre-trained” SOM models, that is to say, the set of SOM models which resulted from the self-organising preparation of the map. So, for the newly added information item, the map is not generally “retrained”; instead steps 150 and 160 are used with all of the SOM models not being amended. To retrain the SOM every time a new information item is to be added is computationally expensive and is also somewhat unfriendly to the user, who might grow used to the relative positions of commonly accessed information items in the map.
However, there may well come a point at which a retraining process is appropriate. For example, if new terms (perhaps new items of news, or a new technical field) have entered into the dictionary since the SOM was first generated, they may not map particularly well to the existing set of output nodes. This can be detected as an increase in a so-called “quantisation error” detected during the mapping of newly received information item to the existing SOM. In the present embodiments, the quantisation error is compared to a threshold error amount. If it is greater than the threshold amount then either (a) the SOM is automatically retrained, using all of its original information items and any items added since its creation; or (b) the user is prompted to initiate a retraining process at a convenient time. The retraining process uses the feature vectors of all of the relevant information items and reapplies the steps 150 and 160 in full.
In operation, initially, the display area 270 is blank. The user types a key word search enquiry into the enquiry area 250. The user then initiates the search, for example by pressing enter on the keyboard 70 or by using the mouse 80 to select a screen “button” to start the search. The key words in the search enquiry area 250 are then compared with the information items in the database using a standard keyword search technique. This generates a list of results, each of which is shown as a respective entry 280 in the list area 260. Then the display area 270 displays display points corresponding to each of the result items.
Because the sorting process used to generate the SOM representation tends to group mutually similar information items together in the SOM, the results for the search enquiry generally tend to fall in clusters such as a cluster 290. Here, it is noted that each point on the area 270 corresponds to the respective entry in the SOM associated with one of the results in the result list 260; and the positions at which the points are displayed within the area 270 correspond to the array positions of those nodes within the node array.
It is noted that the results area 260 displays list entries for those results for which display points are displayed within the boundary 300 and which satisfied the search criteria in the word search area 250. The boundary 300 may encompass other display positions corresponding to populated nodes in the node array, but if these did not satisfy the search criteria they will not be displayed and so will not form part of the subset of results shown in the list 260.
Cross-Cluster Association/Assisted Keyword Search
An example embodiment of the present invention will now be described with reference to
In
The data repository 400 may be separately located to the mapping processor 412. Correspondingly the search processor may be separately located from the data repository 400, mapping processor 412 and those parts shown in
The embodiment shown in
In an alternative embodiment, the search processor 404 may be arranged to search the information items and to generate search results, which identify information items which correspond to a search query. The mapping processor 412 may then receive data representing the results of the search identifying information items corresponding to the search query. The mapping processor then generates the x, y co-ordinates of the positions in the array corresponding to the identified information items.
The mapping processor 412 is operable to identify clusters of information items at a first global level by conducting a k-means clustering algorithm. The k-means clustering algorithm identifies the clusters and position of the clusters within the array. The k-means clustering algorithm is disclosed in book entitled “Neural Networks for Pattern Recognition,” by Christopher M. Bishop, pp 187-188, Oxford University Press. A further dislosure of the k-means clustering algorithm is disclosed in the web address:
As illustrated in
In some embodiments a further operation may be performed by the mapping processor 412 to refine the identification of clusters using the k-means algorithm. The further operation is known as “k-means clustering and pruning”. The known k-means clustering process identifies groups of array positions for information items identified in the search results which denote similar information items. A further pruning process of determining whether adjacent sub-clusters of x, y positions of result items are part of the same main cluster is then performed. If a distance between the centres of two sub-clusters is less than a threshold value, then the two sub-clusters are deemed to be part of the same main cluster. The pruning is performed iteratively in known manner until the clustering is stable.
The mapping processor 412 operates to perform a further analysis of each of the clusters of information items identified at the first hierarchical level h_level1. In order to provide a user with a facility for examining the clusters of information items individually and identifying further clusters within those information items the mapping processor 412 forms a further hierarchical level. Accordingly, for each cluster of information items the k-means clustering algorithm is performed for that cluster to identify further clusters within that first hierarchical level of information items. So for example, as illustrated in
As illustrated for the first hierarchical level each cluster is labelled in accordance with a keyword. The keyword is identified by finding the most common word which each of the information items within the cluster have present in the metadata associated with that information item. So for example in the first hierarchical level three clusters are identified by the words “quiz”, “game” and “DIY”.
In a corresponding manner to the labelling of the clusters of the first hierarchical level h_level1 a keyword is identified for each of the clusters in the second hierarchical level h_level2. Accordingly, the three clusters are labelled “the chair”, “wipeout” and “enemy within”. Each of these three clusters comprises different episodes of a quiz show.
As will be appreciated a further iteration of the analysis of each cluster can be performed. This is achieved by performing the k-means algorithm on each of the clusters identified at the second hierarchical level h_level2. As illustrated in
The mapping processor 412 is therefore operable to identify clusters of information items at different hierarchical levels. Data representing each of the hierarchical levels is fed to the display processor 416. Accordingly, in combination with the graphical user interface 418 a second area may be displayed on the display 420 which may for example correspond to the second hierarchical level h_level2. Thus, using the zoom control a user may zoom into the clusters displayed in the first hierarchical level h_level1. The zoom control may be operated using the user control 414. Accordingly, zooming into a particular cluster can have an effect of revealing the second hierarchical level of information items h_level2. Alternatively, the user control 414 may be used to select a “current view” area within the first area Accordingly, the second display is illustrated with respect to the clusters identified within the “quiz” cluster identified at the first hierarchical level shown in the first display h_level1.
A further advantage provided by embodiments of the present invention is an arrangement in which the second or a subsequent level, which is displayed in a second or subsequent area of the display, may be provided with indicators of other clusters. The indicators direct the user to alternative clusters to the keyword associated with the cluster being viewed at a lower hierarchical level. Thus the clusters which are being illustrated at a lower hierarchical level within the second display area 440, will have alternative clusters to the cluster being viewed. For example, in
A further advantageous feature of some embodiments is to provide a list of additional keywords, that is to say the keywords associated with second level clusters within first level clusters. As illustrated in
Multi-Modal Refined Search
Another example embodiment of the present invention will now be described with reference to
The representative key stamp RKS is generated by forming a colour histogram of each of the frames of video data. The colour histogram for all or selected video frames are combined and then normalised to produce a composite colour histogram, which is illustrated in representative form as a bar graph 466 in
In other embodiments an RKS may be generated for each information item from the video frames, by any of the following methods:
Embodiments of the present invention can provide a facility for producing a refined search based upon selected characterising information features. In one embodiment the search processor 142 is operable to search those information items which were identified in a first search in accordance with either an item of metadata, a video image or audio data. In alternative embodiments the search may be conducted just on metadata or just video data or only audio data or any combination thereof. To facilitate the formation of a search query, the display device 420 shown in
In
A user may also select an information item to be searched in accordance with the audio data within that information item. Accordingly, the third row within the display area 476 provides a facility for a user to introduce a representative image of that information item to identify within the row for audio data that the search query is to include audio data corresponding to that information item within the search query.
In addition to selecting information items to be searched in accordance with the type of the characterising information features, embodiments of the present invention also provide a facility for searching in accordance with Boolean operators between the selected information items. As illustrated in
Having built the search query, the search processor 404 is operable to search the information items identified from a keyword search in accordance with the search query built from the selection made by the user and illustrated in
For the example of searching for characterising information features such as metadata, then for any information item the feature vector for that information item generated from the metadata can be used to identify a point in the two-dimensional array corresponding to that feature vector. Accordingly, information items within a predetermined distance of that identified position in the array can be returned as a result of the search query. However, if more than one information item has been selected within the metadata search row then a search query must be built in a way which searches both of these items in accordance with the Boolean operator selected.
For the example of the “AND” Boolean operator then the feature vector for each information item is combined to form a composite feature vector as illustrated in
For the example of the Boolean “OR” operator for a corresponding metadata search then for the first feature vector A and the second feature B the corresponding position in the array for those feature vectors are identified. As such, the result of the search query is to return all the information items within a predetermined number of positions of each of those identified points in the array. This is illustrated in
For the second line in the search query corresponding to video image data being the characterising feature of the search, then the search processor is operable to search the video data for representative key stamps corresponding to the selected user video image. To this end, the colour histogram associated with the user selected video image is compared with the colour histogram for each of the representative key stamps associated with the information items. A distance is calculated between the colour histogram of the representative key stamp of each of the information items and the colour histogram of the user specified video image. This is effected by calculating a distance between each of the columns representing the colour components of that image and summing these distances for each column. The array position corresponding to the information item having the least distance between the colour histogram of the user selected video image and that of the representative key stamp corresponding to that array position is identified. Again the results of the query would be to return information items having array positions within a predetermined number of positions from the identified array position.
For the case of Boolean operators then again a colour histogram can be formed by combining the colour histograms for two images selected and specified for the Boolean “AND” operator. The process of forming a composite colour histogram is illustrated in
For the example of audio data then the search processor may form a feature vector from the audio metadata associated with the selected information item. For example, the audio metadata may identify harmonics present in the audio signal, speech data or whether there is music present within the audio signals represented by the audio metadata. In addition, the metadata may identify whether a particular speaker is present on the audio signal such as Tony Blair or a particular commentator, such as John Motson. Accordingly, again a feature vector may be generated from the selected audio data which may be searched with respect to other feature vectors associated in particular with audio data. In a corresponding way to that explained above, the Boolean operators may be used to combine a search for more than one audio metadata type. For the example of the “AND” operator the audio metadata items may be combined to produce a composite metadata item. Searching for a corresponding information item which has a feature vector which is closest to this composite item will identify an information item. The search processor may then recover information items within a predetermined number of positions within the array for both metadata items when an “OR” operator is specified. Again the “NOT” Boolean operator will serve to exclude information items returned having matching audio data from the results of the search query.
The embodiments of the present invention have been provided for refining a search from identified information items. However it will be appreciated that in other embodiments the search query formed by the display illustrated in
Various modifications may be made to the embodiments described above without departing from the scope of the present invention. Various aspects and features of the present invention are defined in the appended claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 0227659.0 | Nov 2002 | GB | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/GB03/04758 | 11/4/2003 | WO | 5/26/2005 |
