Patent Application
20130268664
The embodiments of the present invention relate to an operator network node and a method thereof for enabling distribution of content, e.g. personalized content.
Today internet traffic in mobile networks is increasing fast. The internet traffic can be analyzed to derive information about subscriber's internet behavior and dynamic subscriber profiles can be created. A challenge for the mobile networks operators is to efficiently analyze all this internet traffic information and leverage on the analysis results.
An example of how the analysis results can be used is transmission of content such as advertisements and other personalized content exemplified by mobile apps, games, ring tones, movies, music, etc. This is a growing area and has a huge potential to enable mobile operators to support advertisers and content providers of information such that the advertisers and content providers can personalize content and advertisements to the subscribers. In order to optimize the revenue for the advertisers and content providers, it is important to optimize the distribution of the content by sending each content to the subscribers that best fit the profile of the content and at a suitable time. This is referred to as dynamic subscriber profiling.
Thus, an object is to achieve a solution for providing dynamic subscriber profile information to be used for personalizing content transmission.
By using embodiments of the present invention it is possible to provide a dynamic subscriber profile based on a weighted accumulation of the subscriber historic and current internet traffic data considering recency and/or frequency and/or browsing duration of the visits to the different content category types. The dynamic subscriber profile characterises the content which a subscriber wants to consume. “Dynamic” implies that the subscriber profile is updated as content is consumed.
The trend analysis function provided by an embodiment is able to detect variations in the web browsing content category patterns e.g. that a subscriber is changing his normal browsing behaviour e.g. suddenly starting to browse pages related to music.
According to a first aspect of embodiments of the present invention a method in a network node in a communication network is provided. The communication network is controlled by an operator and is configured to carry internet traffic data. In the method, information of internet traffic data associated with subscribers within the communication network controlled by the operator is received and the received information is analysed. A dynamic subscriber profile is created or updated for a respective subscriber by using said analysis.
According to a second aspect of embodiments of the present invention a network node in a communication network is provided. The communication network is controlled by an operator and the communication network is configured to carry internet traffic data. The network node comprises an input section configured to receive information of internet traffic data associated with subscribers within the communication network controlled by the operator. The network node further comprises a processor configured to analyze the received information and to create or update a dynamic subscriber profile for the respective subscriber by using said analysis and a memory configured to store said dynamic subscriber profile.
An advantage with embodiments of the present invention is that an enhanced profiling of subscribers is provided. The enhanced profiling is based on the content types of the subscribers' internet traffic behaviour within the mobile network. The profiling information can be used by application to adapt applications according to the subscriber's interest. An example is to provide the users with the most appropriate advertisement at a given time and according to the subscriber's interest.
Moreover, those skilled in the art will appreciate that the means and functions explained herein below may be implemented using software functioning in conjunction with a programmed microprocessor or general purpose computer, and/or using an application specific integrated circuit (ASIC). It will also be appreciated that while the current embodiments are primarily described in the form of methods and devices, the embodiments may also be embodied in a computer program product as well as a system comprising a computer processor and a memory coupled to the processor, wherein the memory is encoded with one or more programs that may perform the functions disclosed herein.
In the embodiments of the present invention, a network node in a communication network controlled by an operator is introduced. The network node is an analysis component configured to analyze the subscriber behaviour based on the internet traffic data within mobile and fixed network. The network node is configured to provide a dynamic profile of the subscribers based on the current and past internet traffic. The dynamic profile of the subscribers also referred to as dynamic subscriber profile may be used by other applications in the operator network or third parties. For example, a content provider can take a decision on what content to provide to a certain subscriber, based on the dynamic subscriber profile information of this subscriber received from the network node according to the embodiments of the present invention. Another example is that an operator can use the dynamic subscriber profile when selecting commercial offers to his own subscribers e.g. when a subscriber has a new music mobile when visiting music sites. Although personalized content is the main scenario, the embodiments of the present invention are also applicable to other scenarios. The other scenarios comprise scenarios when the surfing behaviour of the subscriber is used as input. The operator may also use this information as one parameter in general subscriber profiling. Further, examples of personalized content are applications, content, services and user profiling.
According to one aspect of the embodiments of the present invention a method in a network node in a communication network controlled by an operator is provided as illustrated in
The network node is referred to as Dynamic Subscriber Profiling Node (DSPN) and interacts with other nodes. This is illustrated in
However, any type of inspection functionality e.g. a Deep Packet Inspection engine (DPI) or a packet inspection functionality in a network node e.g. a gateway or a radio network controller may be used to obtain the same information as the proxy server.
Hence, the URLs of the monitored internet traffic data is sent in step 1 to the network node 100. The network node may then send in step 2 URL information to a content categorization engine (CCE) 120 and the CCE 120 is configured to categorize the URLs that the mobile terminal has visited. The result of the categorization is returned in step 3 to the network node and the dynamic subscriber profile is updated or created in step 4 accordingly. The network node comprises a memory 210 for storing these dynamic subscriber profiles 160. Hence, the network node 100 is configured to analyze the total stream of events derived from the subscribers HTTP traffic within the network.
In accordance with embodiments of the present invention, a dynamic subscriber profile may be created or updated if there already exists a subscriber profile for this subscriber by the network node based on the type of browsing content. As illustrated in the flowchart of
The dynamic subscriber profiles may be weighted which implies that a weight may be used to influence the subscriber profiles based on a number of parameters e.g. time, duration etc. The network node is able to dynamically inform other nodes, via different interfaces, about the current weighted profile of a subscriber derived from the kind of content the subscriber is consuming and has consumed.
According to a further embodiment, the network node is able to analyze 302c trends and to take 302d this analysis into account when creating the dynamic subscriber profiles as illustrated in the flowchart of
Further the network node may comprise interfaces, referred to input and output sections, towards a content categorization engine, a proxy server and application nodes. The interface towards the categorization engine is used to send Uniform Resource Locators (URLs) and to receive categorization information, the interface towards the proxy server is used to receive URLs, including time stamp and duration, and the interface towards the application nodes is used to receive a request for subscriber profiling information and to send profiling information.
The received information of the internet traffic data associated with subscribers of the operator within the communication network comprises information of the URLs browsed during a period of time. According to an embodiment, the URLs browsed are detected by the proxy analyzing the internet traffic i.e. the interactions between the subscriber and the content servers. A number of attributes received from the content analyzer and by analyzing the web traffic is used as input when creating or updating the dynamic subscriber profiles.
The analysis of the received information of the internet traffic data associated with subscribers within the communication network controlled by the operator may comprise:
Determining categories and confidence values of all URLs browsed with time stamps.
Calculating recency values by e.g. sorting the interactions between the mobile terminal and the content providers according to the time stamps in a chronological order and weighting them on an exponentially decaying scale computed over their ordinal rank.
Detecting the browsing duration of a page is detected by analyzing the HTTP traffic i.e. the time between the HTTP messages indicating request for and reception of a webpage.
As explained above, the network node 100 also referred to DSPN is configured to analyse the HTTP traffic for each subscriber. The URLs are analyzed to detect what type of content the subscriber is currently and historically consuming. The DSPN may be equipped with or interworking with a content categorization engine 120 that supports the analysis of the URLs and map them to content types. The content categorization engine 120 classifies the URL with different methods. One classification method comprises analysis of the URL by analyzing text and checking the number of occasions certain words that are related to certain categories appears.
There are usually a number of predefined main content classes with subclasses. An example of a main class is sport and the corresponding subclasses football, basket, all sports etc.
An example of a category tree is shown in
The content categorization engine 120 responds with a number of categories for each URL in a descending order that best fits the content that is reached by the URL. One URL usually corresponds to several categories. The different categories that match the URL are marked with confidence values. The confidence value corresponds to the probability that the content of a certain URL belong to a certain category.
Example: URL=newspaper XYZ may generate the following response with the main/sub class marked. The confidence value is usually normalized between 0 and 1.
According to an embodiment of the present invention, the DSPN comprises a memory 210 for temporarily storing all received categories and confidence values in a vector database per subscriber. The stored categories and confidence values may be deleted when they are no longer interesting to be used in the calculation, e.g. after a predetermined time.
The subscriber profile analysis is performed to get one accumulated weighted value for the subscriber's profile. The profile may be weighted based on the following criteria's:
Confidence & Categories: The categories of the web pages and the confidence values are used as a weight to decide the order of the impact of the subscriber profile. A category with high confidence has a stronger impact on the accumulated category.
Frequency: URLs visited more frequently gives more impact to the subscriber profile.
Recency: The more recently visited URLs are weighted higher. Weights decay exponentially over time with the half-life as a configurable parameter. This means that a browsing for a certain content that occurred in the current time has a contribution of 1 to whereas an interaction from a browsing that occurred one half-life ago contributes ½ and so on.
Below is a table with an example of the impact of the profile for an example subscriber from two browsing events. The categories and confidence values for two different browsing sessions i.e. URLs with the prediction decay included.
The half time(λ) is set to 30 hours.
The URL xyz was accessed 5 hours ago. This means that the decay factor due to aging is 0.5(5/30)=0.89
URL=abc, the URL was accessed 50 hours ago. This means that the decay factor due to aging (recency) is 0.5(50/30)=0.31
The aggregated category and confidence values (which are the sum of the Confidence with decay factors) for the example subscriber from the two browsing events is then:
The table shows that the impact of the browsing to the dynamic subscriber profile of the example subscriber is more impacted from the more recent browsing events.
As an additional weight factor, the browsing duration may be used. A longer duration of the visit to a web page may indicate a higher interest in the content category.
The duration of the visit to a certain content (e.g. web page) is possible to detect from e.g. the proxy server by analyzing the web traffic. The duration of each browsing session is detected by e.g. the proxy server and sent to the DSPN for each browsing session.
The algorithm to be used should give a weight between 0.5 for very short durations and 1 for longer durations.
A modified sigmoid function may be used to get the requested output.
If for example the duration of the session to URL xyz x was 1 minute the weight factor y is 0.6. It should be noted that also a very short browsing session will get at least 0.5 weight.
In some cases it is more important to detect patterns in the behavior and sudden changes in the patterns than the accumulated weighted value. One example could be a subscriber called Bob who very seldom browses pages with content related to entertainment/music but suddenly starts to heavily browse pages related to entertainment/music. Another subscriber called Bill may have a more constant rate or decreasing rate of browsing content related to entertainment/music. This information could be interesting for a directed marketing campaign for music i.e. Bob may be regarded as an inexperienced music consumer and Bill as a more experienced consumer and should thereby get different offers.
Accordingly, the DSPN may comprise a trend analyzer 220. The trend analyzer 220 may be configured from an external node via an API. Examples of attributes to be used for the trend analyzer are content Category, Increasing/decreasing, and speed=moderate/medium/high.
Several categories may be traced simultaneously e.g. multiple categories per marketing campaign. An example may be an advertising campaign for music with different offers depending on experience level of the listener. The advertisement provider may provide the following trend analysis request targeting the inexperienced with the following parameters: Category=(entertainment/music), direction=increasing, speed=high.
The algorithm used to detect the trends may compare the weighted moving mean value for a short period with weighted moving average for a longer period. The values from the weighted recency algorithm as described above may be used to estimate a moving weighted average for the classification of the subscriber. A configurable parameter called “classification deviation” is used to define thresholds for the different levels. Examples of the “classification deviation” parameter follow:
ClassDevModerate=the minimal deviation that is required to classify the subscribers as Moderate increase/decrease. E.g. 20%
ClassDevMedium=the minimal deviation that is required to classify the subscribers as Medium increase/decrease. E.g. 35%
ClassDevHigh=the minimal deviation that is required to classify the subscribers as High increase/decrease. E.g. 50%
The recency algorithm described above is used to estimate a Weighted Moving Average (WMA) for the period.
E.g. a subscriber called David has been browsing more on web pages related to sports the last week than normally e.g. during last year. The WBeventShort for sport is 0.456 and WBeventsLong for sport is 0.280. This means that the deviation is +62% and is classified as a high deviation (CLassDevHigh).
An example of equation to be used for estimating the short term deviation is Deviation=(WBeventsShort−WBeventLong)/WBeventLong
Embodiments of the present invention are described by the following example of an advertisement campaign to subscribers.
An advertisement engine is going to launch an advertisement for sports gear for a sport retailer. The target group for the advertisement is persons with an interest in sport.
Referring again to
In step 2, the DSPN forwards the URL information to the content categorization engine and requests a categorization of the URL. The content categorization engine returns a content categorization in step 3. The received content categorization of the URL is used to create or update the dynamic subscriber profiles for the concerned subscribers.
Accordingly, these steps 1-4 are performed continuously to keep an updated profile of all subscribers.
In step 5, an advertisement engine or another engine that distributes personalized content requests dynamic subscriber profile information to be able to distribute content to a certain group of subscribers. In this example, the advertisement engine requests a profiling of subscribers with interest in sport and health. The profiling request sent may contain the following attributes:
Profile(Sport/all sports, Personal/health)
In step 6, the DSPN analyzes the request, identifies the subscribers having the requested profile and sends 7 the requested subscriber information to the advertisement engine or another engine configured to distribute personalized content. In the advertisement scenario, steps 5-7 are performed for each new campaign.
In this example, the advertisement engine distributes the advertisement to the identified subscribers.
With reference to
According to embodiments, the network node 100 comprises a first input/output section 240 configured to send the information of the internet traffic data to a content categorization engine, and to receive a categorization of the internet traffic data for a certain subscriber.
Moreover, the processor 220 may further be configured to analyze the received internet traffic data to detect trends associated with behaviour of the respective subscriber, and to take the detected trends in account when creating the dynamic subscriber profile.
In addition, the network node 100 may further comprise a second input/output section 230 configured to receive a request from a content provider to identify subscribers with a certain dynamic subscriber profile. In this embodiment, the processor 220 is configured to identify subscribers with the certain dynamic subscriber profile, and the second input/output section 230 is configured to inform the content provider of the identified subscribers.
The input and output sections may be interfaces such as APIs.
Modifications and other embodiments of the disclosed invention will come to mind to one skilled in the art having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Therefore, it is to be understood that the embodiments of the invention are not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of this disclosure. Although specific terms may be employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/SE2010/051384 | 12/15/2010 | WO | 00 | 6/12/2013 |
