WO2015184335A1

WO2015184335A1 - Real-time audience segment behavior prediction

Info

Publication number: WO2015184335A1
Application number: PCT/US2015/033301
Authority: WO
Inventors: Abhishek Chatterjee
Original assignee: Tootitaki Holdings Pte Ltd; AHMANN, William F.
Priority date: 2014-05-30
Filing date: 2015-05-29
Publication date: 2015-12-03

Abstract

An audience persona graph can be generated by semantically enhancing market segmentation data. A trend graph can be generated by semantically enhancing viral content from a data stream. A campaign asset graph can be generated by semantically enhancing a campaign asset description. Conceptually, trend graphs are combined with an audience persona graph to create a trend applicability graph and campaign asset graphs are combined with the audience persona graph to create a campaign asset applicability graph; the trend applicability graph and campaign asset applicability graph can then be overlaid to identify an interest graph useful in the distribution of the campaign asset. Using the interest graph, the system described herein can use trends to reach an appropriate audience for a semantic proxy of the campaign asset.

Description

REAL-TIME AUDIENCE SEGMENT BEHAVIOR PREDICTION

BACKGROUND

[0001] An area of ongoing research and development is distribution of content to a receptive audience. There are many problems associated with identifying an appropriate or ideal target for the delivery of ads. For example, audiences may not want to share personal information to improve the effectiveness of ad distribution. This problem is exacerbated by the increasingly prevalent cookie-less environments associated with mobile computing.

SUMMARY

[0002] The following implementations and aspects thereof are described and illustrated in conjunction with systems, tools, and methods that are meant to be exemplary and illustrative, not necessarily limiting in scope. In various implementations one or more of the above- described problems have been addressed, while other implementations are directed to other improvements.

[0003] Real-time audience segment behavior prediction is described in this paper. An audience persona graph can be generated by semantically enhancing market segmentation data. A trend graph can be generated by semantically enhancing viral content from a data stream. A campaign asset graph can be generated by semantically enhancing a campaign asset description. Conceptually, trend graphs are combined with an audience persona graph to create a trend applicability graph and campaign asset graphs are combined with the audience persona graphs to create a campaign asset applicability graph; the trend applicability graph and campaign asset applicability graph can then be overlaid to identify an interest graph useful in the distribution of the campaign asset. Using the interest graph, the system described herein can use trends to reach an appropriate audience for a semantic proxy of the campaign asset. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] FIG. 1 depicts a diagram of an example of a system for distributing a semantic proxy of a campaign asset using real-time audience segment behavior prediction.

[0005] FIG. 2 depicts a flowchart of an example of a method for distributing a semantic proxy of a campaign asset using real-time audience segment behavior prediction.

[0006] FIG. 3 depicts a diagram of an example of a system for asset class graph generation.

[0007] FIG. 4 depicts a diagram of an example of a portion of an audience persona graph.

[0008] FIG. 5 depicts a diagram of an example of a system for audience persona graph generation.

[0009] FIG. 6 depicts a diagram of an example of a system for trend graph generation.

[0010] FIG. 7 depicts a diagram of an example of a system for real-time audience segment behavior prediction and semantic proxy generation.

[0011] FIG. 8 depicts a diagram of an example of a system for real-time audience segment behavior prediction with feedback from a semantic proxy distributor.

[0012] FIG. 9 depicts a flowchart of an example of a method for combining audience persona themes with a campaign assets graph.

[0013] FIG. 10 depicts a diagram an audience persona graph mapped onto campaign asset subgraphs.

[0014] FIG. 11 depicts a flowchart of an example of a method for combining audience persona themes with a trends graph.

[0015] FIG. 12 depicts a flowchart of an example of a method for building an interest graph.

[0016] FIG. 13 depicts a diagram illustrating generation of a semantic proxy for distribution by one or more distribution networks.

[0017] FIG. 14 depicts a flowchart of an example of a method for real-time audience segment behavior prediction and semantic proxy generation. [0018] FIG. 15 depicts a flowchart of an example of a method for distribution of at least a portion of a semantic proxy to consumers of thematically compatible content.

DETAILED DESCRIPTION

[0019] FIG. 1 depicts a diagram 100 of an example of a system for distributing a semantic proxy of a campaign asset using real-time audience segment behavior prediction. The diagram 100 includes a computer-readable medium 102, a market segment parameters provisioning system 104, a campaign asset parameters provisioning system 106, a trend parameters provisioning system 108, a semantic proxy distribution instructions provisioning system 110, a real-time audience segment behavior prediction and semantic proxy generation system 112, semantic proxy distribution systems 114 (comprising a semantic proxy distribution system 114-1 to a semantic proxy distribution system 114-n), and semantic proxy presentation systems 116 (comprising semantic proxy presentation systems 116-1 to semantic proxy presentation systems 116-n). Each of the semantic proxy distribution systems 116-1 to 116-n have a set of semantic proxy distribution systems, which are designated by adding an additional reference numeral (e.g. semantic proxy presentation system 116-1-n refers to the n^th semantic proxy presentation system of the presentation systems for the (1^st) semantic proxy distribution system 114-1).

[0020] The computer-readable medium 102 is intended to represent a variety of potentially applicable technologies. For example, the computer-readable medium 102 can be used to form a network or part of a network. Where two components are co-located on a device, the computer-readable medium 102 can include a bus or other data conduit or plane. Where a first component is co-located on one device and a second component is located on a different device, the computer-readable medium 102 can include a wireless or wired back-end network or LAN. The computer-readable medium 102 can also encompass a relevant portion of a WAN or other network, if applicable.

[0021] As used in this paper, a "computer-readable medium" is intended to include all mediums that are statutory (e.g., in the United States, under 35 U.S.C. 101), and to specifically exclude all mediums that are non-statutory in nature to the extent that the exclusion is necessary for a claim that includes the computer-readable medium to be valid. Known statutory computer- readable mediums include hardware (e.g., registers, random access memory (RAM), nonvolatile (NV) storage, to name a few), but may or may not be limited to hardware.

[0022] The computer-readable medium 102 or portions thereof, as well as other systems, interfaces, engines, datastores, and other devices described in this paper, can be implemented as a computer system, a plurality of computer systems, or a part of a computer system or a plurality of computer systems. In general, a computer system will include a processor, memory, non-volatile storage, and an interface. A typical computer system will usually include at least a processor, memory, and a device (e.g., a bus) coupling the memory to the processor. The processor can be, for example, a general-purpose central processing unit (CPU), such as a microprocessor, or a special-purpose processor, such as a microcontroller.

[0023] The memory can include, by way of example but not limitation, random access memory (RAM), such as dynamic RAM (DRAM) and static RAM (SRAM). The memory can be local, remote, or distributed. The bus can also couple the processor to non-volatile storage. The non-volatile storage is often a magnetic floppy or hard disk, a magnetic-optical disk, an optical disk, a read-only memory (ROM), such as a CD-ROM, EPROM, or EEPROM, a magnetic or optical card, or another form of storage for large amounts of data. Some of this data is often written, by a direct memory access process, into memory during execution of software on the computer system. The non-volatile storage can be local, remote, or distributed. The nonvolatile storage is optional because systems can be created with all applicable data available in memory.

[0024] Software is typically stored in non-volatile storage. Indeed, for large programs, it may not even be possible to store the entire program in memory. Nevertheless, it should be understood that for software to run, if necessary, it is moved to a computer-readable location appropriate for processing, and for illustrative purposes, that location is referred to as the memory in this paper. Even when software is moved to the memory for execution, the processor will typically make use of hardware registers to store values associated with the software, and local cache that, ideally, serves to speed up execution. As used herein, a software program is assumed to be stored at an applicable known or convenient location (from non-volatile storage to hardware registers) when the software program is referred to as "implemented in a computer- readable storage medium." A processor is considered to be "configured to execute a program" when at least one value associated with the program is stored in a register readable by the processor.

[0025] In one example of operation, a computer system can be controlled by operating system software, which is a software program that includes a file management system, such as a disk operating system. One example of operating system software with associated file management system software is the family of operating systems known as Windows® from Microsoft Corporation of Redmond, Washington, and their associated file management systems. Another example of operating system software with its associated file management system software is the Linux operating system and its associated file management system. The file management system causes the processor to execute the various acts required by the operating system to input and output data and to store data in the memory, including storing files on the non-volatile storage.

[0026] The bus can also couple the processor to the interface. The interface can include one or more input and/or output (I/O) devices. The I/O devices can include, by way of example but not limitation, a keyboard, a mouse or other pointing device, disk drives, printers, a scanner, and other I/O devices, including a display device. The display device can include, by way of example but not limitation, a cathode ray tube (CRT), liquid crystal display (LCD), or some other applicable known or convenient display device. The interface can include one or more of a modem or network interface. It will be appreciated that a modem or network interface can be considered to be part of the computer system. The interface can include an analog modem, isdn modem, cable modem, token ring interface, satellite transmission interface (e.g. "direct PC"), or other interfaces for coupling a computer system to other computer systems. Interfaces enable computer systems and other devices to be coupled together in a network.

[0027] The computer systems can be compatible with or implemented as part of or through a cloud-based computing system. As used in this paper, a cloud-based computing system is a system that provides virtualized computing resources, software and/or information to client devices. The computing resources, software and/or information can be virtualized by maintaining centralized services and resources that the edge devices can access over a communication interface, such as a network. "Cloud" may be a marketing term and for the purposes of this paper can include any of the networks described herein. The cloud-based computing system can involve a subscription for services or use a utility pricing model. Users can access the protocols of the cloud-based computing system through a web browser or other container application located on their client device.

[0028] A computer system can be implemented as an engine, as part of an engine, or through multiple engines. As used in this paper, an engine includes at least two components: 1) a dedicated or shared processor and 2) hardware, firmware, and/or software modules that are executed by the processor. Depending upon implementation- specific or other considerations, an engine can be centralized or its functionality distributed. An engine can include special purpose hardware, firmware, or software embodied in a computer-readable medium for execution by the processor. The processor transforms data into new data using implemented data structures and methods, such as is described with reference to the FIGs. in this paper.

[0029] The engines described in this paper, or the engines through which the systems and devices described in this paper can be implemented, can be cloud-based engines. As used in this paper, a cloud-based engine is an engine that can run applications and/or functionalities using a cloud-based computing system. All or portions of the applications and/or functionalities can be distributed across multiple computing devices, and need not be restricted to only one computing device. In some embodiments, the cloud-based engines can execute functionalities and/or modules that end users access through a web browser or container application without having the functionalities and/or modules installed locally on the end-users' computing devices.

[0030] As used in this paper, datastores are intended to include repositories having any applicable organization of data, including tables, comma-separated values (CSV) files, traditional databases (e.g., SQL), or other applicable known or convenient organizational formats. Datastores can be implemented, for example, as software embodied in a physical computer-readable medium on a general- or specific-purpose machine, in firmware, in hardware, in a combination thereof, or in an applicable known or convenient device or system. Datastore-associated components, such as database interfaces, can be considered "part of" a datastore, part of some other system component, or a combination thereof, though the physical location and other characteristics of datastore-associated components is not critical for an understanding of the techniques described in this paper.

[0031] Datastores can include data structures. As used in this paper, a data structure is associated with a particular way of storing and organizing data in a computer so that it can be used efficiently within a given context. Data structures are generally based on the ability of a computer to fetch and store data at any place in its memory, specified by an address, a bit string that can be itself stored in memory and manipulated by the program. Thus, some data structures are based on computing the addresses of data items with arithmetic operations; while other data structures are based on storing addresses of data items within the structure itself. Many data structures use both principles, sometimes combined in non-trivial ways. The implementation of a data structure usually entails writing a set of procedures that create and manipulate instances of that structure. The datastores, described in this paper, can be cloud-based datastores. A cloud-based datastore is a datastore that is compatible with cloud-based computing systems and engines. [0032] Referring once again to the example of FIG. 1, the market segment parameters provisioning system 104 is coupled to the computer-readable medium 102. The market segment parameters provisioning system 104 is intended to represent an applicable system controlled by an entity responsible for providing information about market segmentation applicable to a campaign asset, such as a product or service. The market segment parameters provisioning system 104 may or may not be controlled by an entity that creates or markets the asset. For example, the market segment parameters provisioning system 104 could be controlled by an internal marketing division of a product or service provider or an external marketing entity acting on behalf of the product or service provider.

[0033] Market segmentation parameters can include geographic, demographic, psychographic, behavioristic, and/or other data about one or more market segments predicted or accepted to be of use in carrying out a campaign or presenting a semantic proxy of a campaign asset, such as a product, to an audience most likely to want to obtain or want to be aware of the campaign asset. In this context, the term "market" is intended to be construed broadly. As used in this paper, a semantic proxy is a data structure suitable for display to a content consumer or for interpretation by a distributor to display relevant portions of the semantic proxy or content identifiable therefrom. For example, in an advertising implementation, the semantic proxy could be referred to as an ad unit, such as a banner advertisement, provided to a suitable platform for display.

[0034] In the example of FIG. 1, the campaign asset parameters provisioning system

106 is coupled to the computer-readable medium 102. The campaign asset parameters provisioning system 106 is intended to represent an applicable system controlled by an entity responsible for providing information about a campaign asset. The campaign asset parameters provisioning system 106 may or may not be controlled by the entity that creates, wholesales, or retails the campaign asset. For example, the campaign asset parameters provisioning system 106 could be controlled by an internal marketing division of a product or service provider or an external marketing entity acting on behalf of the product or service provider. The campaign asset parameters provisioning system 106 may or may not be controlled by the same entity as that which controls the market segment parameters provisioning system 104. Indeed, marketers of products or services often have access to market segmentation data, making concurrent control of both systems synergistic in some respects. [0035] Campaign asset parameters can include images (or media content), a message associated with the campaign asset, a description of the campaign asset, specifications and characteristics of products, identities of related products, specifications and characteristics of related products, the manufacturers or developers of products, or other content predicted to be of use in carrying out the campaign or more effectively presenting a semantic proxy of the campaign asset. A campaign asset can include, for example, a product or service being advertised, a recommendation or review, or some other asset designed to inform a market segment about a campaign. In this paper, an example of an advertising campaign is frequently described by way of example, but not necessarily by limitation.

[0036] In the example of FIG. 1, the trend parameters provisioning system 108 is coupled to the computer-readable medium 102. The trend parameters provisioning system 108 is intended to represent an applicable system controlled by an entity responsible for providing information about trends that may be applicable to a campaign asset. The trend parameters provisioning system 108 may or may not be controlled by the entity that creates or markets an asset. For example, the trend parameters provisioning system 108 could be controlled by an internal marketing division of a product or service provider or an external marketing entity acting on behalf of the product or service provider. The trend parameters provisioning system 108 may or may not be controlled by the same entity as that which controls the market segment parameters provisioning system 104 and/or the campaign asset parameters provisioning system 106. However, the provisioning of trend parameters is not necessarily as synergistic as the provisioning of market segment and campaign asset parameters are with one another, at least in some respects. So it is expected the entity controlling the trend parameters provisioning system 108 is less likely to be the same as the entity controlling the market segment parameters provisioning system 104 and/or the campaign asset parameters provisioning system 106.

[0037] Trend parameters can include images (or media content), news articles, people, events, products, music, phrases (spoken or written), or practically any other form of content being consumed on a particular channel or medium. In a specific implementation, the channel or medium is itself included in the trend parameters. For example, the trend parameters can be characterized as defining what content is currently being consumed on TWITTER® or FACEBOOK®.

[0038] In the example of FIG. 1, the semantic proxy distribution instructions provisioning system 110 is coupled to the computer-readable medium 102. The semantic proxy distribution instructions provisioning system 110 is intended to represent an applicable system controlled by an entity responsible for providing information about which semantic proxy distribution networks may be applicable to a campaign asset. The semantic proxy distribution instructions provisioning system 110 may or may not be controlled by the entity that creates or markets an asset. For example, the semantic proxy distribution instructions provisioning system 110 could be controlled by an internal marketing division of a product or service provider or an external marketing entity acting on behalf of the product or service provider. The semantic proxy distribution instructions provisioning system 110 may or may not be controlled by the same entity as that which controls the market segment parameters provisioning system 104 and/or the campaign asset parameters provisioning system 106, and may be considered synergistic with these systems. In an alternative, the semantic proxy distribution instructions provisioning system 110 is instead or in addition controlled by the same entity as controls the trend parameters provisioning system 108.

[0039] Proxy distribution instructions can include instructions to provide semantic proxies to a specific distribution network (e.g. FACEBOOK®) or multiple distribution networks (e.g. FACEBOOK®, ADMOB®, and YOUTUBE®). The instructions can include a frequency requirement, a cost limitation, a range of preferences, an instruction to provide the semantic proxies in the most effective way possible (e.g. to distribution networks providing the greatest results relative to cost), or other instructions suitable for guiding a semantic proxy generation system to push semantic proxies with greater or lesser flexibility depending upon the instruction set.

[0040] In the example of FIG. 1, the real-time audience segment behavior prediction and semantic proxy generation system 112 is coupled to the computer-readable medium 102. The real-time audience segment behavior prediction and semantic proxy generation system 112 is intended to represent an applicable system controlled by an entity responsible for identifying a real-time audience for a campaign asset using trends. The entity may or may not be the same entity as that which controls the market segment parameters provisioning system 104, the campaign asset parameters provisioning system 106, the trend parameters provisioning system 108, and/or the proxy distribution instructions provisioning system 110. In a specific implementation, the trend parameters provisioning system 108 and the real-time audience segment behavior prediction and semantic proxy generation system 112 are controlled by the same entity. [0041] In a specific implementation, the real-time audience segment behavior prediction and semantic proxy generation system 112 takes advantage of knowledge about content a market segment is currently consuming (e.g. the system looks for content being consumed by an identifiable audience) and ties the trending content to a campaign asset for presentation in association therewith. The more trends the real-time audience segment behavior prediction and semantic proxy generation system 112 can identify, the higher the probability the system can identify the trends that motivate a market segment.

[0042] In a specific implementation, the real-time audience segment behavior prediction and semantic proxy generation system 112 is capable of semantically enhancing market segments to create audience personas, semantically enhancing campaign assets to create campaign asset graphs, and semantically enhancing trending content to create trend graphs. These techniques are described in more detail later in this paper. Advantageously, semantic enhancement enables the real-time audience segment behavior prediction and semantic proxy generation system 112 to operate on potentially anonymized market segments, in potentially cookie-less environments, or in other environments in which the real-time audience segment behavior prediction and semantic proxy generation system 112 has no or limited knowledge specific to a particular user. User-specific knowledge can also be used, if desired, but cookie- less environments appear to be growing more prevalent, making technologies capable of identifying audiences without user- specific knowledge increasingly advantageous.

[0043] Advantageously, the real-time audience segment behavior prediction and semantic proxy generation system 112 can predict in real-time what a particular market segment "likes" to enable more targeted distribution of semantic proxies for the campaign asset. The semantic proxy will typically be in a format the real-time audience segment behavior prediction and semantic proxy generation system 112 was able to determine was appropriate. The determination can be by way of observation or, more likely, by finding the format requirements in a publication by the semantic proxy distribution systems 114 or agents thereof; it is in the interests of the semantic proxy distribution systems 114 to make their semantic proxy format requirements known.

[0044] In the example of FIG. 1, the semantic proxy distribution systems 114 are coupled to the computer-readable medium 102. The semantic proxy distribution systems 114 are intended to represent applicable systems controlled by an entity with a platform suitable for distribution of semantic proxies of campaign assets. One of the entities may or may not be the same entity as that which controls the market segment parameters provisioning system 104, the campaign asset parameters provisioning system 106, the trend parameters provisioning system 108, the proxy distribution instructions provisioning system 110, or the real-time audience segment behavior prediction and semantic proxy generation system 112.

[0045] Appropriate platforms include, by way of example but not limitation, web pages

(e.g. the semantic proxy of the campaign asset could be distributed as a banner advertisement on a web page, such as a FACEBOOK® web page), mobile advertising platforms (e.g. the semantic proxy of the campaign asset could be distributed as a mobile app advertisement, such as by ADMOB®), video distribution platforms (e.g., the semantic proxy of the campaign asset could be distributed as a video advertisement, such as by YOUTUBE®), electronic messages (e.g. messages such as TWITTER® tweets; email is less likely to be useful in systems utilizing anonymized consumer data, though it is theoretically possible in such systems to send email to a distribution address without knowledge of the recipients of the distribution address), electronic billboards, radio broadcasts (particularly live broadcasts to enable timeliness), print media (e.g. flyers; books are less likely to be useful because of the real-time nature of trends, though it is in theory possible to include print media such as books), and other platforms suitable for providing content to a content consumer.

[0046] Given the nature of a semantic proxy distribution platform, the semantic proxy distribution systems 114 can provide trending data to the trend parameters provisioning system 108 and/or act as a subsystem of the trend parameters provisioning system 108 in particular. For example, TWTTTER® can act as a semantic proxy distribution systems 114, with a semantic proxy being distributed as a tweet. Other tweets on the TWITTER® network comprise a data stream, which can be monitored by a third party to identify trends, or by TWITTER® if identifying trends is determined to be desirable. Thus, TWITTER® could itself identify trends and provide trend parameters as a trend parameters provisioning subsystem of the trend parameters provisioning system 108, if it so desired and were appropriately configured to do so. In a similar way, the semantic proxy distribution systems 114 could make data associated with users of the semantic proxy distribution systems 114 available (inherently, intentionally, or unintentionally) for use by a market segmentation parameters provisioning system 104 or act as a market segment parameters provisioning subsystem of the market segment parameters provisioning system 104. Also, the entity controlling the semantic proxy distribution systems 114 could have a marketing division responsible for receiving information from a party interested in marketing a campaign asset, and the campaign asset parameters provisioning system 106 could be associated with the marketing division. Thus, at least theoretically, a single entity could control all of the systems 104-114.

[0047] In the example system shown in FIG. 1, the semantic proxy presentation systems

116 are coupled to the computer-readable medium 102. The semantic proxy presentation systems 116 are intended to represent applicable systems controlled by content consumers (or downstream broadcasters, such as message board or billboard owners) capable of displaying an applicable semantic proxy. The semantic proxy presentation systems 116 should have the ability to display the semantic proxy in a manner appropriate for the semantic proxy, such as a screen for visual content or a speaker for audio content. An entity of the entities may or may not be the same entity as that which controls the semantic proxy distribution systems 114, but if all of the entities are the same entity as that which controls the semantic proxy distribution systems 114, the intended functionality of distributing semantic proxies to an appropriate audience would appear to be defeated. Accordingly, it is generally assumed the semantic proxy presentation systems 116 are controlled by entities other than the entity that controls the semantic proxy distribution systems 114. Moreover, in most intended implementations, the computer-readable medium 102 will take the form of a network between the semantic proxy presentation systems 116 and the other illustrated components (as opposed to a bus or other connection suitable for coupling components on a single machine).

[0048] In an alternative, the semantic proxy distribution systems 114 can obviate the need for the semantic proxy presentation systems 116 (or a subset thereof). Specifically, the semantic proxy distribution systems 114 can display the semantic proxy in a manner in which humans can consume the content without the use of a device. For example, the semantic proxy distribution systems 114 can display the semantic proxy on flyers or on an electronic display under the control of the entity that controls the semantic proxy distribution systems 114 (e.g. on an electronic billboard).

[0049] A specific example of operation for an advertising campaign using a system such as is illustrated in the example of FIG. 1 is now described. The market segment parameters provisioning system 104 provides geographic, demographic, psychographic, and/or behavioristic data, which may or may not be anonymized, to the real-time audience segment behavior prediction and semantic proxy generation system 112. The campaign asset parameters provisioning system 106 provides a campaign asset image, campaign asset message, and campaign asset description to the trend-based audience targeting system 106. The trend parameters provisioning system 108 provides identified trends and distribution channels associated with the trends to the real-time audience segment behavior prediction and semantic proxy generation system 112. The real-time audience segment behavior prediction and semantic proxy generation system 112 semantically enhances the market segment data to generate an audience persona graph, semantically enhances the campaign asset using the campaign asset description to generate a campaign asset graph, and semantically enhances the identified trends to generate a trends graph; using these graphs, the real-time audience segment behavior prediction and semantic proxy generation system 112 generates a semantic proxy of a campaign asset, such as an ad unit, which is provided to the semantic proxy distribution systems 114, which can be referred to as an ad network in applicable implementations. The semantic proxy distribution systems 114 distributes the semantic proxies to the semantic proxy presentation systems 116 (or intervening nodes that forward the semantic proxies to the semantic proxy presentation systems 116) for presentation to the target audience in an appropriate manner, such as a banner advertisement on a web page displayed on a screen of one or more of the semantic proxy presentation systems 116.

[0050] FIG. 2 depicts a flowchart 200 of an example of a method for distributing a semantic proxy of a campaign asset using real-time audience segment behavior prediction. The flowchart 200 and other flowcharts in this paper are illustrated as a sequence of modules. It should be understood the sequence of the modules can be changed and the modules can be rearranged for serial or parallel processing, if appropriate.

[0051] In the example of FIG. 2, the flowchart 200 starts at modules 202, 204, 206, and

208 with obtaining distribution instructions, obtaining market segment parameters, obtaining campaign asset parameters, and obtaining trend parameters, respectively. Distribution instructions include flexible or inflexible parameters to be followed by an agent acting on behalf of the sender of the instructions when providing semantic proxies for campaign assets to distribution networks. Market segment parameters include characteristics of potential consumers of a campaign asset. The market segment parameters can be provided as generalized data, data specific to particular markets, or data expected to be particularly relevant to a specific campaign asset. Semantically enhanced market segment parameters can be referred to as an audience persona or audience persona graph. Campaign asset parameters include at least a representation of a particular campaign asset, which can include an audio clip, image, or multimedia clip, a message, and/or a description. Trend parameters include at least an identification of a potential trend, which can include trending media or memes and the channels on which the media or memes are trending. Identifying trends can include determining whether media or a meme is sufficiently viral to be characterized as a trend.

[0052] In the example of FIG. 2, the flowchart 200 continues to module 210 with generating a semantic proxy of the campaign asset utilizing trends applicable to an audience persona. The technique for combining the market segment, campaign asset, and trend parameters to obtain an interest graph that is used to generate the semantic proxy are provided in more detail later in this paper.

[0053] In the example of FIG. 2, the flowchart 200 continues to module 212 with providing the semantic proxy to a semantic proxy distribution system for distribution to consumers of content on the semantic proxy distribution system. The semantic proxy distribution system can be referred to as an ad network in appropriate circumstances. One example of an ad network is FACEBOOK®, which distributes semantic proxies (ad units) on web pages viewed by FACEBOOK® users. Distribution of semantic proxies to one or more applicable distribution networks should follow distribution instructions and/or other decisionmaking constants, variables, or algorithms suitable for maximizing semantic proxy presentation to receptive audiences within the parameters of a given instruction set.

[0054] In the example of FIG. 2, the flowchart 200 continues to module 214 with obtaining feedback to improve semantic enrichment. The feedback can be obtained directly from the semantic proxy distribution system, from an entity paying for distribution of semantic proxies on the semantic proxy distribution system, or from a third party analyst. In a specific implementation, feedback is used to determine whether semantic models that are effective in general are effective when applied to audiences with particular characteristics or to campaign assets with particular characteristics and/or how trends presented in association with campaign assets or campaign assets having particular characteristics are received by a target audience or target audiences with particular characteristics.

[0055] In the example of FIG. 2, the flowchart 200 returns to module 208 to obtain trend parameters as described previously. In this way, for as long as a campaign continues, the real-time behavior of an audience persona can be accounted for, with a corresponding shift in characteristics of semantic proxies of campaign assets. The flowchart 200 is not intended to suggest distribution instructions, market segment parameters, and campaign asset parameters could not be changed over time, but any deviation could be considered to be a new campaign, ending the flow of the flowchart 200, and starting a new flow. It may further be noted the feedback obtained as described in module 214 can serve to improve upon distribution network selection (depending upon the flexibility of the distribution instruction set), improve the audience persona generated at least in part by semantically enhancing the market segment parameters, and/or improve the campaign asset graph generated at least in part by semantically enhancing the campaign asset parameters.

[0056] In the example of FIG. 2, the flowchart 200 ends upon the completion of a campaign. It may be noted systems can learn from each campaign. That is, after a campaign ends, the system may have improved its semantic models based upon what was learned over the course of the campaign.

[0057] FIG. 3 depicts a diagram 300 of an example of a system for asset class graph generation. The diagram 300 includes a campaign asset parameters provisioning subsystem 302-1 to campaign asset parameters provisioning subsystem 302-n (collectively, the campaign asset class parameters system 302), a campaign asset parameters datastore 304, a campaign asset class parameters provider interface engine 306, a campaign asset descriptions datastore 308, a semantic enrichment engine 310, an asset ontology datastore 312, a campaign asset class graph datastore 314, and a campaign asset content datastore 316.

[0058] In the example of FIG. 3, the campaign asset class parameters provisioning system 302 is intended to represent a potentially distributed collection of subsystems that make available (directly or indirectly, intentionally or inherently) campaign asset parameters for a campaign asset class. The campaign asset parameters provisioning subsystem 302-1 need not be under the control of the same entity that controls a campaign asset parameters provisioning subsystem 302-2 (not shown). It may be noted the use of subsystems is optional if a single entity is responsible for making available all relevant campaign asset parameters for a campaign asset class in a particular instance. For example, a single entity may provide all of the campaign asset parameters for a class of campaign assets (e.g., TOYOTA® could provide campaign asset parameters for each model of automobile). As another example, multiple entities may provide campaign asset parameters for different campaign assets within a class (e.g., AMAZON® could instruct vendors to provide campaign asset parameters for each campaign asset that is part of Amazon's campaign).

[0059] In the example of FIG. 3, the campaign asset parameters datastore 304 is coupled to the campaign asset class parameters provisioning system 302. The campaign asset parameters datastore 304 is intended to represent campaign asset parameters made available by the campaign asset class parameters provisioning system 302 in whatever format or on whatever channel they are made available.

[0060] In the example of FIG. 3, the campaign asset class parameters provider interface engine 306 is coupled to the campaign asset parameters datastore 304. The campaign asset class parameters provider interface engine 306 is intended to represent whatever interface is needed to make campaign asset parameters available for semantic enrichment and/or use in building a semantic proxy. For example, in an implementation in which the campaign asset class parameters provisioning system 302 is under the control of an entity different from the entity that will perform semantic enrichment, the campaign asset class parameters provider interface engine 306 can include a network interface. If the campaign asset class parameters provisioning system 302 is under the control of the same entity to perform semantic enrichment, the campaign asset parameters datastore 304 and the campaign asset class parameters provider interface engine 306 can be omitted.

[0061] In the example of FIG. 3, the campaign asset descriptions datastore 308 is coupled to the campaign asset class parameters provider interface engine 306 (or, in an alternative in which the campaign asset parameters datastore 304 and the campaign asset class parameters provider interface engine 306 are omitted, to the campaign asset class parameters provisioning system 302). Campaign asset descriptions are a part of campaign asset parameters that are useful for semantic enrichment. In a specific implementation, a description includes words associated with a campaign asset (e.g. fuel-efficient, green technology, automobile, and hybrid for a TOYOTA® PRIUS®), which may or may not be locally constant across the entire campaign asset class (e.g. if the campaign asset class is every model of TOYOTA®, the word "automobile" might be a class constant). In an alternative, the description can include other media that can be semantically enhanced through interpretation of the media and application of an ontology (e.g. the word "baby" could be derived from a picture of a baby or a sound datastore could determine an audio clip is "Stairway to Heaven," the terms of which could be semantically enhanced).

[0062] In the example of FIG. 3, the semantic enrichment engine 310 is coupled to the asset descriptions datastore 308. The semantic enrichment engine 310 is intended to represent an engine capable of combining campaign asset descriptions in the campaign asset descriptions datastore 308 with semantic data sets stored in the asset ontology datastore 312 to generate a campaign asset class graph for storage in the campaign asset graph datastore 314. The asset ontology datastore 312 includes ontological nodes associated with people, places, objects, events, activities, or the like that can be associated with words (explicit or derived) of campaign asset descriptions to generate a semantically enriched campaign asset class graph including a mapping for each of the campaign assets of the class. Advantageously, particularly when semantically enriched trend graphs and an audience persona graph are available, the semantically enriched campaign asset class graph is more effective at identifying an appropriate target audience segment than is possible with conventional market segment data.

[0063] In the example of FIG. 3, the campaign asset content datastore 316 is coupled to the campaign asset class parameters provider interface engine 306. The campaign asset content datastore 316 is intended to represent a datastore of content for campaign asset parameters that are not incorporated into the asset class graph or, if they are incorporated, are also for presentation to a target audience. For example, the campaign asset content datastore 316 can include an image associated with a campaign asset, an image associated with each campaign asset, or multiple images associated with each campaign asset. As another example, the campaign asset content datastore 316 can include a message associated with the campaign asset, such as a message for display in association with an image associated with the campaign asset.

[0064] FIG. 4 depicts a diagram 400 of an example of a portion of a campaign asset class graph. The diagram 400 is intended to illustrate an example of how a semantically enhanced campaign asset description is presented as a graph, and how the graph can be augmented. It should be understood the simplistic nature of the diagram 400 is to aid in an understanding of the concepts. The graph includes a root node 402, an Energy Efficiency node 404, an Automobiles node 406, a Green Technology node 412, a Hybrids node 416, a Toyota node 420, and a Prius node 424. Other nodes are also illustrated, but are not explicitly referenced (so no reference numeral is necessary). Certain of the edges between nodes have reference numerals: edges 408, 410, 414, 418, 422, 426, 428, and 430. Edges 428 and 430 are represented as dashed lines, which is intended to illustrate a change, as is described shortly.

[0065] In the example of FIG. 4, the root node 402, which represents the top-most node of an asset ontology, is coupled to the Energy Efficiency node 404 and the Automobiles node 406 via respective edges 408 and 410. The root node 402 is, of course, likely to be connected to other nodes in the ontology, as well. The edges 408 and 410 have weights that correspond to the relative tightness of the nodes to the root. For illustrative purposes, the weights are represented in this paper as a value from 0 to 1. For illustrative purposes, graph traversal is accomplished using a multiplication function. Using these assumptions, graph traversal from the Energy Efficiency node 404 to the Automobiles node 406 would be accomplished by multiplying the weight of the edge 408 by the weight of the edge 410, and the product would represent the relative tightness of the Energy Efficiency node 404 to the Automobiles node 406.

[0066] In the example of FIG. 4, the Energy Efficiency node 404 is coupled to the

Green Technology node 412 via edge 414 and the Automobiles node 406 is coupled to the Hybrids node 416 via the edge 418, which is coupled to the Toyota node 420 via the edge 422, which is coupled to the Prius node 424 via the edge 426. Continuing the example from the preceding paragraph, multiplying the weight of the edges 414, 408, 410, 418, 422, and 426 yields the relative tightness of the Green Technology node 412 to the Prius node 424.

[0067] For a given campaign asset, the relative tightness between geographic characteristics can be different than that indicated by the asset ontology. Consider, for example, a campaign asset description for a "2014 Toyota Prius hybrid automobile." Establishing the relatedness of Green Technology and Prius could be associated with a graph traversal up the hierarchical levels of the graph from the Green Technology node 412 and back down the hierarchical levels of the graph to the Prius node 424 (at least until the graph is modified based upon a heretofore unidentified tightening connection between Green Technology and Prius). With the exception of edge weights of 1 (one), each time a traversal is made from one node to the next, the computed tightness of the relationship decreases and if an edge weight is 0 (zero), the nodes are treated as having no relationship at all (because the product of any number and 0 is 0).

[0068] It could be determined that, at least in some instances, the Prius node 424 is more tightly associated with the Green Technology node 412 than the graph traversal indicates. Accordingly, the edge 428 can be added to the graph to establish a direct connection between the Green Technology node 412 and the Prius node 424, providing a "shortcut" between the nodes and eliminating the need to traverse the graph up to the root node 402 and back down again. Conceptually, the edge 428 establishes Green Technology and Prius have a single degree of separation. Graph traversal from Biofuels (a child node of the Green Technology node 412) to 2015 (a child node of the Prius node 424 representing the year) would still be accomplished by multiplying all intervening edges, but traversal higher than the Green Technology node 428 and the Prius node 424 is unnecessary in this case. [0069] As another example, say, for example, it is determined over time that the graph traversal from the Prius node 424 through the Toyota node 420 to the Hybrids node 416 is not as representative as a single degree of separation by edge 430 between the Prius node 424 and the Hybrids node 416. Maybe this is caused by determining audiences are more receptive to Prius when hybrids are trending than they are to hybrids and Toyota. So, while the edge 422 may not be adjusted, the edge 430 can bypass the Toyota node 420 to represent the tighter relationship between Prius and hybrid than there is between Toyota and hybrid. Conceptually, the edge 430 establishes the Prius is one degree of separation from hybrid, making the location of Prius hierarchically beneath Toyota irrelevant. Graph traversal from 2015 (a Prius year) would still be accomplished by multiplying the edge to the Prius node 424, the edge 430, and any edges between the Hybrids node 416 and some other applicable node.

[0070] As another example, say, for example, the weight between nodes might be adjusted for particular audience personas (not shown). For example, it may be the case hybrids become more popular automobile choices. So the weight of the edge 418 can be increased to more tightly associate hybrids with automobiles.

[0071] The dotted line 432 in the example of FIG. 4 is intended to illustrate how a campaign asset graph might be mapped onto the asset ontology. In this example, the nodes 406, 416, 420, and 424 are explicitly mentioned in a campaign asset description and the Green Technology node 412 is added to the campaign asset graph using semantic enrichment. For a campaign asset class graph, there are one or more campaign asset graphs mapped onto the ontology, which may or may not overlap in whole or in part.

[0072] As should be clear from the preceding examples, for particular campaign assets, the hierarchical campaign asset class graphs can be modified with campaign asset-specific edges, potentially flattening the hierarchy. The same principles can be applied to an audience persona graph, with market segment characteristics of a given audience persona being mapped to an audience ontology. The same principles can be applied to trends, with individual trend graphs being mapped to a trend ontology. Conceptually, the relevant ontologies are hierarchically arranged and are all connected to a single root node (not shown in FIG. 4), but it may be understood a connection by an edge with weight 0 and no connection are logically equivalent; so in some implementations, graphs can be disjointed.

[0073] FIG. 5 depicts a diagram 500 of an example of a system for audience persona graph generation. The diagram 500 includes a market segmentation parameters provider interface engine 502, a market segmentation datastore 504, a semantic enrichment engine 506, an audience ontology 508, and a persona graph datastore 510.

[0074] In the example of FIG. 5, the market segmentation parameters provider interface engine 502 obtains market segmentation parameters from a market segmentation parameters provider. The market segment parameters provider interface engine 502 is intended to represent whatever interface is needed to make market segment parameters available for semantic enrichment, such as, e.g., a network interface.

[0075] In the example of FIG. 5, the market segment datastore 504 is coupled to the market segmentation parameters provider interface engine 502. The market segment datastore 504 is intended to represent a market segmentation description.

[0076] In the example of FIG. 5, the semantic enrichment engine 506 is coupled to the market segment datastore 504. The semantic enrichment engine 506 is intended to represent an engine capable of combining the market segment data in the market segment datastore 504 with semantic data sets stored in the audience ontology datastore 508 to generate an audience persona graph for storage in the persona graph datastore 510. The audience ontology datastore 508 includes ontological nodes associated with market segmentation data that can be applied to conventional market segment data to generate a semantically enriched audience persona graph. Advantageously, particularly when semantically enriched campaign asset graphs and trend graphs are available, the semantically enriched audience persona graph is more effective at identifying an appropriate target audience segment than is possible with conventional market segment data.

[0077] FIG. 6 depicts a diagram 600 of an example of a system for trends graph generation. The diagram 600 includes a data stream 602, a data stream monitoring engine 604, a content datastore 606, a virality determination engine 608-1 to virality determination engine 608-n (collectively, the virality determination engine 608), a viral content datastore 610, a trends ontology 612, a semantic enrichment engine 614, and a trends graph datastore 616.

[0078] In the example of FIG. 6, the data stream 602 is intended to represent a multi- source content stream. The data stream 602 can include by way of example but not limitation, RSS feeds, news feeds, Facebook® posts, Twitter® tweets, YouTube® videos and/or comments, search engine results, or any other data indicative of content consumption. [0079] In the example of FIG. 6, the data stream monitoring engine 604 is coupled to the data stream 602. The data stream monitoring engine 604 is intended to represent an applicable engine for receiving or observing a data stream from multiple data sources. For example, the data stream monitoring engine 604 can observe the most popular searched items on a search engine over a specific period of time.

[0080] In the example of FIG. 6, the content datastore 606 is coupled to the data stream monitoring engine 604 and stores the content observed by the data stream monitoring engine 604 for at least a non-transitory time span. The content stored in the content datastore 606 can include redundancy, which is useful for determining whether content is popular and/or going viral.

[0081] In the example of FIG. 6, the content datastore 606 is coupled to the virality determination engine 608. The virality determination engine 608 is intended to represent an engine that is configured to sort through content or content-related statistics to find the subset of content that exceeds a virality threshold. The virality threshold can be adjusted to ensure at least some content is identified as viral (or "sufficiently viral"). For example, the virality determination engine 608 may have a first virality threshold that results in adequate content being flagged as viral for a first time-span. For some reason, the amount of viral content decreases; so the virality threshold is reduced to increase the amount of "sufficiently viral" content passed through the system. It may also be desirable to adjust the virality threshold in the opposite direction to reduce the amount of content flagged as viral. In a specific implementation, the degree of virality can also be measured.

[0082] In a specific implementation, the virality determination engine 608 identifies viral content in a multi-source data stream received by the data stream monitoring engine 604 and stored in the content datastore 606. For example, the virality determination engine 608 can determine that data in a multi-source data stream is viral by the number of times words included in the content are used. Depending upon implementation- specific or other considerations, the virality determination engine 608 can determine that content is viral if words included in the content are used a certain number of times over a threshold value during a specific amount of time. For example if a name of a celebrity is used 100 times within a minute and a threshold value is set for 99 times in a minute, then the virality determination engine 608 can determine content that includes the name of the celebrity is viral. In identifying virality from content, the virality determination engine 608 can extract one or a plurality of words from the content that semantically identifies the content or characteristics related to the trend. Depending upon implementation- specific or other considerations, the viraltiy determination engine 608 can extract one or more words that are repeated in content. For example, if a name of a celebrity appears multiple times in content, then the virality determination engine 608 can extract the name of the celebrity to semantically identify the viral content.

[0083] In a specific implementation, the virality determination engine 608 also determines whether viral content is gaining momentum. As used in this paper, viral content is gaining momentum when the content is detected more frequently over time. In determining whether viral content is gaining momentum, the virality determination engine 608 can monitor a multi- source data stream to see how often either or both content, or one or more words that identify the content appear in the multi-source data stream. For example, if content is identified or associated with a celebrity's name, then the virality determination engine 608 can monitor a multi- source data stream to determine if the content is gaining momentum. Depending upon implementation- specific or other considerations, the virality determination engine 608 can determine that content is gaining momentum if content, or one or more words that identify the content appear in a multi-source data stream at an increasing frequency. It may be considered desirable to weight more heavily viral content that has gained momentum relative to viral content that is relatively new.

[0084] In the example of FIG. 6, the virality determination engine 608 is intentionally represented as a set of virality determination engines 608-1 to 608-n. The purpose of this representation is to suggest multiple instances of viral content can be identified, which ultimately are each ultimately stored as a trend graph in a trends graph (of all of the trends, potentially with pruning).

[0085] In the example of FIG. 6, the viral content datastore 610 is coupled to the virality determination engine 608 and stores the content, or a subset, description, or identifier thereof, that the virality determination engine 608 flagged as sufficiently viral. In a specific implementation, the viral content datastore 610 can also store the degree of virality of viral content.

[0086] In the example of FIG. 6, the trends ontology datastore 612 includes an ontology of content comprising objects arranged in a hierarchical or otherwise linked manner and connected with weighted edges. Nodes of the ontological graph can include words that semantically identify content and characteristics related to the content. [0087] In the example of FIG. 6, the semantic enrichment engine 614 is coupled to the viral content datastore 610 and the trends ontology datastore 612. The semantic enrichment engine 614 is intended to represent an engine configured to semantically enhance viral content from the viral content datastore 610 for mapping onto the trends ontology in the trends ontology datastore 612 to generate a trends graph for storage in the trends graph datastore 616. Advantageously, relatively unstructured viral content can thereby be translated into structured and up-to-date trends.

[0088] For example, viral content may be associated with a specific celebrity with a node in the trends ontology (under a "People" node). The ontology can be modified for the specific celebrity to create edges (or change the weights of edges) between the specific celebrity and relatives or people with associated names. The edges can also extend directly from the specific celebrity node to brands of products (under an "Object" node) the specific celebrity is known to wear or endorse or to events (under an "Events" node) the specific celebrity has attended or intends to attend.

[0089] See, e.g., FIG. 4 for an example of how edges can be modified for campaign assets. A similar graph can be drawn for a universal semantic set, with edges being drawn for nodes as the graph grows. As an additional layer, the nature of a specific instance of a trend can modify the edges in association with that specific instance of the trend. For example, the birth of Prince George might establish a more heavily- weighted connection between Prince William and baby diapers, which may diminish as the trend subsides. The trends graph stored in the trends graph datastore 616 is intended to represent the current ("real time") relationships as they are related to a particular trend.

[0090] FIG. 7 depicts a diagram 700 of an example of a system for real-time audience segment behavior prediction and semantic proxy generation. The diagram 700 includes a distribution instructions datastore 702, a campaign asset class graph datastore 704, a persona graph datastore 706, a trends graph datastore 708, a campaign asset applicability designation engine 710, an applicable campaign assets datastore 712, an applicable campaign assets graph pruning engine 714, a pruned applicable campaign assets datastore 716, a trend applicability designation engine 718, an applicable trends graph datastore 720, an applicable trends graph pruning engine 722, a pruned trend applicability graphs datastore 724, an interest graph building engine 726, an interest graph 728, a network preference selection engine 730, a network preferences datastore 732, a semantic proxy set generation engine 734, a semantic proxy set datastore 736, a distributor interface engine 738.

[0091] In the example of FIG. 7, the distribution instructions datastore 702 is intended to store a set of instructions regarding the distribution of content associated with a campaign asset class. In a specific implementation, the market segment (though not necessarily the semantically-enhanced market segment referred to in this paper as an audience persona) to which the content is to be distributed is known and provided in association with the distribution instructions. In a specific implementation, the set of campaign assets (sometimes referred to as a campaign asset class in this paper) is known and provided in association with the distribution instructions. In a specific implementation, distribution networks are known, identities of which are provided in association with the distribution instructions. For example, an entity can provide distribution instructions that identify a market segment to which campaign assets associated with a campaign assets class and a set of distribution networks to which content associated with the campaign assets are to be distributed. Instructions can be explicit (e.g. 50% of content must be sent to a first distribution network and 50% of content must be sent to a second distribution network; 25% of content must be associated with a first campaign asset and 75% of content must be associated with a second campaign asset; etc.), conditional (e.g. if it is determined a first distribution network will be more effective, send more content to the first distribution network; if a first campaign asset is trending, send more content associated with the first campaign asset, etc.), or implicit/unconditional (e.g. choose the best distribution network for distribution of content; choose the most trendy campaign asset; etc.).

[0092] In the example of FIG. 7, the campaign asset class graph datastore 704 is intended to store a campaign asset graph including semantically enhanced campaign asset descriptions. In a specific implementation, a universal asset ontology can be combined with a specific campaign asset description to create at least a portion of the semantically enhanced campaign asset class graph, as described previously with reference to FIG. 3. The campaign asset class graph datastore 704 can include one or more campaign assets, any of which can be "activated" for applicability considerations, depending upon distribution instructions, feedback from distribution networks, configurations, or other considerations. The campaign asset class graph datastore 704 may or may not include campaign assets with similar characteristics or elements that are constant within the class. For example, the campaign asset class may include dissimilar campaign assets that are provided by a single entity interested in marketing the campaign assets (with potentially the only similarity between a first campaign asset and a second campaign asset being they are marketed by the same entity). As another example, campaign assets of a class can be similar, enabling the class to be semantically enhanced and applied to all campaign assets of the class. Campaign asset classes are discussed in more detail later.

[0093] In the example of FIG. 7, the audience persona graph datastore 706 is intended to store an audience persona graph of a semantically-enriched market segment. In a specific implementation, a universal audience ontology can be combined with a specific market segment to create the semantically enhanced audience persona graph, as described previously with reference to FIG. 5. The audience persona graph datastore 706 may or may not include multiple audience personas, any of which can be "activated" for applicability considerations. The audience persona graph datastore 706 may or may not include audience persona groups with similar characteristics or elements that are constant within the group. An audience persona group can be treated the same as an audience persona for a set of audience personas, and a specific audience persona can be considered if greater detail is called for. Audience persona classes are predicted to be of lesser importance than campaign asset classes due to the way market segment data and asset descriptions are traditionally used in marketing, and are therefore not discussed in any greater detail in this paper. However, it should be understood from the discussion of the campaign asset classes that audience persona classes could be treated in a similar fashion.

[0094] In the example of FIG. 7, the trends graph datastore 708 is intended to store a trend graph of semantically enhanced viral content (as defined by the content being of sufficient virality to, in the aggregate, exceed a virality threshold). The trends graph datastore 708 can include such information as where a trend was found, which can be useful when selecting a distribution network. For example, if a trend is trending on TWTTTER®, the TWITTER® distribution network might be considered an appropriate distribution network for content associated with the trend. In a specific implementation, a universal trends ontology can be combined with a specific trend to create at least a portion of the semantically enriched trends graph, as described previously with reference to FIG. 6. In a typical implementation, the trends graph datastore 708 will store multiple trend subgraphs, any of which can be "activated" for applicability considerations, depending upon distribution instructions, feedback from distribution networks, configurations, and other considerations. In a specific implementation, different trends are treated discretely, though similar trends could be consolidated into a single trend (potentially resulting in two sets of not sufficiently viral content, in the aggregate, being considered sufficiently viral to qualify as a trend).

[0095] In a specific implementation, trends may have an associated lifespan in the trends graph datastore 708, causing trends to subside if they are not refreshed by continuously finding similar viral content, refreshed by feedback indicating the trends are still effective, or refreshed in some other manner. The lifespan may or may not be the same upon introduction of a new trend. For example, trends with momentum may be given a longer lifespan that trends that pop up without any momentum, or trends on one type of distribution network might have a longer lifespan than another type of distribution network. In general, it is an objective of the trends graph datastore 708 to store trends that are trending in real-time (i.e. right now on a given distribution network), with the understanding the trends graph could conceivably lag somewhat.

[0096] In the example of FIG. 7, the campaign asset applicability designation engine 710 is coupled to the campaign asset class graph datastore 704 and the audience persona graph datastore 706. The campaign asset applicability designation engine 706 is intended to represent an engine that finds and ranks a subset of the plurality of campaign asset graphs and/or subgraphs in the campaign asset class graph datastore 704 by how effectively an audience persona in the audience persona graph datastore 706 maps to each of the subsets of the plurality of campaign asset graphs and/or subgraphs. The subsets of campaign asset graphs and/or subgraphs that score above a campaign asset applicability threshold are at least conceptually combined with the persona graph to form a campaign asset applicability graph for the audience persona. The mapping of an audience persona onto a campaign asset class graph is described in more detail later with reference to FIG. 9.

[0097] In the example of FIG. 7, the applicable campaign assets graph datastore 712 is coupled to the campaign asset applicability designation engine 710. The applicable campaign assets graph datastore 712 is intended to at least store subsets of campaign asset graphs and/or subgraphs determined by the campaign asset applicability designation engine 710 to be relevant to an active audience persona. Depending upon implementation- specific or other considerations, the applicable campaign assets graph datastore 712 can store a campaign asset applicability graph for an audience persons that includes campaign assets graphs and/or subgraphs determined by the campaign asset applicability designation engine 710 and including campaign assets to which an audience person can be mapped. [0098] In the example of FIG. 7, the applicable campaign assets graph pruning engine

714 is coupled to the applicable campaign assets graph datastore 712. In a specific implementation, the campaign asset applicability graph includes all campaign assets because at some future time, it may be determined a persona audience that was not previously found to be associated with a given campaign asset is actually more closely associated than previously thought. So it may be desirable to maintain the applicable campaign assets graph datastore 712 with more campaign asset graphs and/or subgraphs than are currently seen as applicable to a given audience persona. A purpose of the applicable campaign assets graph pruning engine 714 is to prune some of the campaign assets from a campaign asset applicability graph to generate a pruned campaign asset applicability subgraph. This can be accomplished by requiring only a set number of subgraphs included in the campaign asset applicability graph to be considered, pruning all subgraphs with lower relative association between a campaign asset and the audience persona; requiring only subgraphs that exceed a relevance threshold, pruning all subgraphs with weights that are lower than the relevance threshold; a combination; introducing a pseudo-randomized, partially pseudo-randomized, or non-random element to cause some subgraphs that do not have sufficient weight to be considered anyway for the purpose of seeing if a relevance is not represented in a current semantic model; or some other pruning consideration.

[0099] In the example of FIG. 7, the pruned applicable campaign asset graphs datastore

716 is coupled to the applicable campaign assets graph pruning engine 714. The pruned applicable campaign asset graphs datastore 716 is intended to include a pruned campaign asset applicability subgraph that includes only those campaign asset graphs and/or subgraphs that are to be considered when building an interest graph.

[00100] In the example of FIG. 7, the trend applicability designation engine 718 is coupled to the audience persona graph datastore 706 and the trends graph datastore 708. The trend applicability designation engine 718 is intended to represent an engine that finds and ranks a subset of the trends graphs and/or subsets of trend graphs in the trends graph datastore 708 by how effectively an audience persona in the audience persona graph datastore 706 maps to each of the set of trends graphs and/or subsets of trend graphs. The subsets of trend graphs and/or trend subgraphs that score above a trend applicability threshold are at least conceptually combined with the relevant trend graph to form a trend applicability graph for the audience persona. The mapping of an audience persona onto a trends graph is described in more detail later with reference to FIG. 10. [00101] In the example of FIG. 7, the applicable trends graph datastore 720 is coupled to the trend applicability designation engine 718. The applicable trends graph datastore 720 is intended to at least store a subset of trend graphs and/or subgraphs determined by the trend applicability designation engine 718 to be relevant to an active audience persona. Depending upon implementation- specific or other considerations, the applicable trends datastore 720 stores a trend applicability graph which includes an audience persona mapped onto a subset of trend graphs and/or subgraphs determined to be compatible with the audience persona. An example of an a campaign asset applicability graph is described with reference to FIG. 9; an a trend applicability graph is conceptually similar and there may be many such graphs (potentially as many as the number of trends times the number of audience personas).

[00102] In the example of FIG. 7, the applicable trends graph pruning engine 722 is coupled to the applicable trends graphs datastore 720. The applicable trends graph pruning engine 722 is similar to the applicable campaign assets graph pruning engine 714 described previously, but operates on a trend applicability graph.

[00103] In the example of FIG. 7, the pruned applicable trends graphs datastore 724 is coupled to the applicable trends graph pruning engine 722. The pruned applicable trends graphs datastore 724 is similar to the pruned applicable campaign assets datastore 716 described previously, but operates to store pruned trend applicability subgraphs, as generated by the applicable trends graph pruning engine 722.

[00104] In the example of FIG. 7, the interest graph building engine 726 is coupled to the pruned campaign asset applicability graph datastore 716 and the pruned trend applicability graph datastore 724. The interest graph building engine 726 functions to build an interest graph based on a pruned campaign asset applicability subgraph and a pruned trend applicability subgraph. The building of the interest graph is described in more detail with reference to FIG. 12.

[00105] In the example of FIG. 7, the interest graph datastore 728 is coupled to the interest graph building engine 726. An interest graph can be characterized as an audience persona plus a theme. The edges of an interest graph can be associated with two weights, one for the audience persona and one for a theme. For an audience persona, if it can be determined what the audience persona is interested in (trending) and it can be determined what the audience persona prefers from among the campaign assets of a campaign asset class, the interest graph can represent the relationships between trends and campaign assets with the audience persona market segment characteristics binding the two (if possible). That is, if there is a trend in which an audience persona seems to be interested and there is a campaign asset in which an audience persona is predicted to be interested, the trend and the campaign should be combined to find the appropriate audience members who are consuming trending content and who might be interested in the campaign asset.

[00106] In the example of FIG. 7, the network preference selection engine 730 is coupled to the distribution instructions datastore 702, the audience persona graph 706, and the trends graph 708. The network preference selection engine 730 is intended to represent an engine configured to select appropriate network preferences for a given audience persona, taking into account trends, on a particular semantic proxy distribution network. The audience persona graph 706 may or may not include information about a potential audience segment that is useful for picking an appropriate distribution network. For example, the audience persona graph 706 may include a market segment characteristic defined as "mobile user," which would cause (or weigh in favor of) the network preference selection engine 730 to select from a distribution network with network channels appropriate for mobile users. Some market segment characteristics can be sufficiently detailed that a specific distribution network is singled out. For example, the audience persona graph 706 may include a market segment characteristic defined as "Facebook member," which suggests the FACEBOOK® distribution network might be most appropriate, while other market segment characteristics might be a bit more general, such as "social network member," which suggests FACEBOOK®, YAHOO®, GOOGLE®, or any of a number of distribution networks with social networking platforms might be appropriate. The network preference selection engine 730 can also take into account trends. In a specific implementation, the trends graph 708 includes information about where a trend was found. So, for example, if a trend is trending on TWITTER®, the network preference selection engine 730 could choose to weigh the TWEETER® distribution network more heavily when choosing which distribution network or networks to select for distribution of a semantic proxy set or subset.

[00107] In the example of FIG. 7, the network preferences datastore 732 is coupled to the network preference selection engine 730. The network preferences datastore 732 is intended to represent the portions of the data available to the system illustrated in the diagram 700 that serve as network preference indicators. The network preferences datastore 732 may or may not be built on an as-needed basis. For example, the network preferences datastore 732 could include network preferences for the (active) audience persona graph in the audience persona graph datastore 706 or for audience personas that are not currently part of an active campaign.

[00108] In the example of FIG. 7, the semantic proxy generation engine 734 is coupled to the network preferences datastore 732 and the interest graph datastore 728. The semantic proxy generation engine 734 generates a set of semantic proxies appropriate for an audience persona. Generation of a set of semantic proxies is described in more detail with reference to FIG. 13.

[00109] In the example of FIG. 7, the semantic proxy set datastore 736 is coupled to the semantic proxy generation engine 734. The semantic proxy set datastore 736 is intended to represent a datastore for storing (e.g. buffering) semantic proxies for transmission to one or more distribution networks.

[00110] In the example of FIG. 7, the distributor interface engine 738 is coupled to the semantic proxy set datastore 736. The distributor interface engine 738 is intended to represent an interface (e.g. a network interface) to a relatively remote distribution network and all hardware and software under the control of the system for real-time audience segment behavior prediction and semantic proxy generation necessary for its operation. It should be understood the distributor interface engine 738 is likely coupled to the one or more distribution networks through a computer-readable medium that is not under the control of either party.

[00111] In the example of FIG. 7, in operation, distribution instructions are stored in the distribution instructions datastore 702, a campaign asset class graph is stored in the campaign asset class graph datastore 704, an audience persona graph is stored in the audience persona graph datastore 706, and a trends graph is stored in the trends graph datastore 708.

[00112] In this example of operation, the campaign asset applicability designation engine

710 designates campaign asset applicability by mapping the audience persona graph onto the campaign asset class graph, yielding an applicable campaign assets graph for storage in the applicable campaign assets graph datastore 712. The campaign asset applicability designation engine 710 can augment the applicable campaign assets graph by finding semantic-based connections between nodes of a campaign asset and the audience persona, which can become part of one or more campaign asset applicability graphs that comprise the applicable campaign assets graph. The determination as to semantic -based node connections can be augmented in its effectiveness by considering feedback from a current or past campaigns that establish, in a relevant context, a closer relationship between nodes of an ontology than would exist otherwise. The campaign asset applicability designation engine 710 can also inject a bias against edge weights if the indicated weights are not borne out in practice, essentially introducing a time- based degradation of edge weights if current or past campaigns do not suggest the indicated weight. Specifically, if a relationship is not shown for a period of time, the edge weight between relevant nodes can be reduced over time, while a recurring indication of closer relationship can increase the edge weight (though the weigh can again decrease over time if the relationship does not continue to hold).

[00113] In this example of operation, the applicable campaign assets graph pruning engine 714 removes campaign subgraphs from the campaign asset applicability graphs to yield a pruned campaign assets applicability subgraph for storage in the pruned campaign assets graph datastore 716. Pruning can be done to eliminate campaign assets that have no or insufficient weighted relationship between a campaign asset and the audience persona. The applicable campaign assets graph pruning engine 714 can take into account the highest weight, total number of non-zero (or above applicability threshold) weights, or both. The subgraphs that survive the pruning process may or may not survive a next pruning process depending upon the results of feedback to the campaign asset applicability designation engine 710.

[00114] In this example of operation, the trends applicability designation engine 718 designates trend applicability by mapping the audience persona graph onto the trends graphs and/or subgraphs, yielding a trend applicability graph for storage in the applicable trends graph datastore 720. The trend applicability designation engine 718 can augment the trend applicability graph by finding semantic-based connections between nodes of a trend and the audience persona, which can become part of one or more applicable trend subgraphs that comprise the applicable trends graphs. The determination as to semantic-based node connections can be augmented in its effectiveness by considering feedback from a current or past campaigns that establish, in a relevant context, a closer relationship between nodes of an ontology than would exist otherwise. The trends applicability designation engine 718 can also inject a bias against edge weights if the indicated weights are not borne out in practice, essentially introducing a time-based degradation of edge weights if current or past campaigns do not suggest the indicated weight. Specifically, if a relationship is not shown for a period of time, the edge weight between relevant nodes can be reduced over time, while a recurring indication of closer relationship can increase the edge weight (though the weigh can again decrease over time if the relationship does not continue to hold). [00115] In this example of operation, the applicable trends graph pruning engine 722 removes subgraphs from the trend applicability graph to yield a pruned trend applicability subgraph for storage in the pruned trends graph datastore 724. Pruning can be done to eliminate trends that have no or insufficient weighted relationship between a trend and the audience persona. The applicable trends graph pruning engine 722 can take into account the highest weight, total number of non-zero (or above applicability threshold) weights, or both. The subgraphs that survive the pruning process may or may not survive a next pruning process depending upon the results of feedback to the trend applicability designation engine 718.

[00116] In this example of operation, the interest graph building engine 726 combines the pruned trend applicability subgraph and pruned campaign asset applicability subgraph to generate an interest graph for storage in the interest graph datastore 728. The interest graph includes a set of subgraphs that can be characterized as themes. Themes can be further semantically enriched to add additional connecting nodes between trends and campaign assets. In a specific implementation, the themes include a bag of words associated with a trend, a campaign asset, and the audience persona.

[00117] In this example of operation, the network preference selection engine 730 takes into account distribution instructions in the distribution instructions datastore 702 to determine network preferences for storage in the network preferences datastore 732. The network preference selection engine 730 can also take into account feedback establishing the degree of success of various themes on various distribution networks. Depending upon the distribution instructions, the feedback can influence network preferences. For example, if feedback indicates a particular theme is more effective on a first network than a second network, the network preferences can indicate the first network is preferable over the second network in the relevant context.

[00118] In this example of operation, the semantic proxy set generation engine 734 uses the interest graph datastore 728 and network preferences datastore 732 to generate a semantic proxy set for storage in the semantic proxy set datastore 736. The semantic proxy sets can include, for example, an image associated with the relevant campaign asset, a message associated with the relevant campaign asset, and a subset of the bag of words in the interest graph datastore 728. Thus, in this example of operation, a semantic proxy can include an image, a message, and a bag of words. A set of semantic proxies may be desirable over a single semantic proxy because semantic models can be improved (with feedback) if different combinations of thematic words are used with different semantic proxies for the same campaign asset on the same or different distribution networks. For example, it may be determined that a semantic proxy generates more interest in an audience segment when it includes a combination of a first and second word than when it includes a combination of a third and fourth word. The themes are not necessarily optimized because it may be possible to determine through feedback whether words are connected by impacting edge weights based upon how effective a theme with a combination of words is. Even if optimization is attempted, it may be desirable from time to time to introduce a small sample of experimental themes to see if thematic combinations of words are effective despite their lack of a high-weight relationship; such experimentation can serve to improve semantic models by identifying previously unrecognized relationships. For large campaigns, the small number of experimental themes can easily be incorporated without having a significant impact on performance even in the event many of the experimental themes fail to yield useful results.

[00119] In this example of operation, the distributor interface engine 738 sends the set of semantic proxies to one or more distribution networks. As was suggested throughout this example, it may be desirable to receive feedback from the distribution networks in order to improve semantic modeling.

[00120] FIG. 8 depicts a diagram 800 of an example of a system for real-time audience segment behavior prediction with feedback from a semantic proxy distributor. The diagram 800 includes a campaign asset class graph datastore 802, a persona graph datastore 804, a trends graph datastore 806, a campaign asset applicability designation engine 808, an applicable campaign assets datastore 810, a trend applicability designation engine 812, an applicable trends graph datastore 814, a network preference selection engine 816, a network preferences datastore 818, a semantic proxy set generation engine 820, a semantic proxy set datastore 822, a distributor interface engine 824, a feedback datastore 826, a feedback interpretation and incorporation engine 828, a theme effectiveness datastore 830, an edge effectiveness datastore 832, and an edge effectiveness datastore 834. The campaign asset class graph datastore 802, persona graph datastore 804, trends graph datastore 806, campaign asset applicability designation engine 808, applicable campaign assets datastore 810, trend applicability designation engine 812, applicable trends graph datastore 814, network preference selection engine 816, network preferences datastore 818, semantic proxy set generation engine 820, semantic proxy set datastore 822, and distributor interface engine 824 can be implemented in a manner similar to, respectively, the campaign asset class graph datastore 704, persona graph datastore 706, trends graph datastore 708, campaign asset applicability designation engine 710, applicable campaign assets datastore 712, trend applicability designation engine 718, applicable trends graph datastore 720, network preference selection engine 730, network preferences datastore 732, semantic proxy set generation engine 734, semantic proxy set datastore 736, and distributor interface engine 738 described with reference to FIG. 7.

[00121] In the example of FIG. 8, the feedback datastore 826 is coupled to the distributor interface engine 824. The feedback datastore 826 is intended to represent any feedback received from the distribution network to which the semantic proxies were sent. The feedback can include data sufficient to determine the effectiveness of a campaign. For example, if an aspect of a trend (e.g. the title of Lady Gaga's recent album) is effective in increasing interest from members of an audience persona when Lady Gaga is trending, then the name of the album could be seen as more effective in view of the trend than the artist herself. The effectiveness of words associated with trends in drawing the interest of members of an audience persona is useful for improving semantic models. In an alternative, instead of or in addition to the distribution network feedback, feedback can be received from any entity with data associated with a relevant campaign.

[00122] In the example of FIG. 8, the feedback interpretation and incorporation engine

828 is coupled to the feedback datastore 826. The feedback interpretation and incorporation engine 828 is intended to represent an engine capable of analyzing the feedback in the feedback datastore 826 and creating useful data structures from the feedback for storage in various datastores to be discussed in the next few paragraphs.

[00123] In the example of FIG. 8, the theme effectiveness datastore 830 is coupled to the feedback interpretation and incorporation engine 828. The theme effectiveness datastore 830 is intended to represent data the network preference selection engine 816 can make use of when generating network preferences that would be more effective given the effectiveness of themes. Themes are the combinations of words associated with trend and campaign asset graphs onto which an audience persona is mapped. Some themes may play better on certain distribution networks, for example. It may also be desirable to save money by not spending money on an ineffective distribution network.

[00124] In the example of FIG. 8, the edge effectiveness datastore 832 is coupled to the feedback interpretation and incorporation engine 828. The edge effectiveness datastore 832 is intended to represent data useful to the campaign asset applicability designation engine 808 when updating edge weights of an campaign asset applicability graph in the applicable campaign assets graph datastore 810. For example, if it is determined for a particular campaign that the relative weight between two nodes of an ontology is less than usual, the relevant edge can be modified to effect the reality of the relationship for the particular campaign. In a more specific example, if it is determined "green technology" and "automobile" are more closely related for a TOYOTA® PRIUS® campaign asset, an edge between a green technology node and an automobile node can be formed with an appropriate weight in lieu of the path or edge between the two nodes in a more universal ontology. Because the weight is different in the specific context, the edge can be added to the campaign asset applicability graph as opposed to a more general asset, persona, or trend ontology.

[00125] In the example of FIG. 8, the edge effectiveness datastore 834 is coupled to the feedback interpretation and incorporation engine 828. The edge effectiveness datastore 834 is intended to represent data useful to the trend applicability designation engine 812 when updating edge weights of a trend applicability graph the applicable trends graph datastore 814. For example, if it is determined for a particular campaign that the relative weight between two nodes of an ontology is less than usual, the relevant edge can be modified to effect the reality of the relationship for the particular campaign. In a more specific example, if it is determined "green technology" and "automobile" are more closely related for a TOYOTA® PRIUS® trend, an edge between a green technology node and an automobile node can be formed with an appropriate weight in lieu of the path or edge between the two nodes in a more universal ontology. Because the weight is different in the specific context, the edge can be added to the trend applicability graph as opposed to a more general asset, persona, or trend ontology.

[00126] In a specific implementation, the campaign asset applicability designation engine

808 and/or the trend applicability designation engine 812 can incorporate time-based edge degradation instead of or in addition to the feedback-based edge effectiveness or ineffectiveness.

[00127] FIG. 9 depicts a flowchart 900 of an example of a method for combining audience persona themes with a campaign assets graph. Depending upon implementation- specific or other considerations the flowchart can be used to generate a campaign asset applicability graph. The flowchart 900 starts at module 902 with traversing an ontological graph to connect a granular campaign asset characteristic to a granular market segment characteristic. [00128] In the example of FIG. 9, the flowchart 900 continues to decision point 904 where it is determined what nodes exist between a first node in a persona graph and one or more nodes in at least a subset of campaign asset subgraphs. If it is determined there are additional graph traversals to do (904- Y), the flowchart 900 returns to module 902 where the ontological graph is traversed between two more nodes. This loop between module 902 and decision point 904 can at least in theory continue until each intervening node between each node of an audience persona graph and each node of a campaign asset graph is identified.

[00129] If, on the other hand, it is determined there are no additional nodes for which graph traversal is necessary (904-N), the flowchart 900 continues to module 906 where granular characteristic node relationship weight is computed. As was discussed above with reference to FIG. 4, weights can be computed in a number of ways, a simple example being computing a product of edge weights for each edge between a first node and a second node. In an example that includes an edge weight of 0 to 1, an edge weight of 0 would essentially result in "no relationship." At this point, each market segment characteristic of an audience persona can, at least in theory, have an associated weight vis-a-vis each node in a campaign assets graph.

[00130] In the example of FIG. 9, the flowchart 900 continues to decision point 908 where it is determined whether to accept a theme for an audience persona. The theme for the audience persona can include a set of bags of words obtained through ontological graph traversal comprising a granular characteristic of the original audience persona graph; the granular characteristics of intervening nodes, if any; and the granular characteristic of one of the original campaign asset graphs. For example, if a granular characteristic of the original audience persona graph is "mobile user," an intervening node is "gamer," and the granular characteristic of one of the original campaign asset graphs is "MINECRAFT®," the associated bag of words is {Mobile User, Gamer, MINECRAFT®}. Each audience persona granular characteristic can have an associated bag of words with respect to each campaign asset granular characteristic, the combination of which can be characterized as a "theme." It may be noted when a semantic proxy is generated, the theme can be pared down to a set of themes, each with a subset of potentially overlapping word sets.

[00131] If it is determined an identified theme is not to be accepted (908-N), the flowchart 900 returns to module 906 to consider a next identified theme. The granular characteristics that are considered inadequate for the purpose of generating a bag of words generally do not have a sufficient weight. For example, for weights that do not exceed the relationship weight threshold, the bag of words are not added to the super-theme. It may be noted it is still possible to provide semantic proxies for campaign assets for which there is no bag of words connecting granular characteristics of the audience persona and granular characteristics of the campaign asset; it is possible to simply rely upon trends to guess an audience persona member will be interested in the semantic proxy of the campaign asset (and perhaps edge weights can be updated if a relationship is discovered).

[00132] If, on the other hand, it is determined an identified theme is to be accepted (908-

Y), the flowchart 900 continues to module 910 where the theme is added to a super-theme. The granular characteristics that are considered adequate for the purpose of generating a bag of words generally must have a sufficient weight. For example, if the weight of a product of edges between an audience persona granular characteristic and a campaign asset granular characteristic exceeds a relationship weight threshold, the bag of words can be added to the super-theme. To distinguish, the acceptable themes identified in the modules 902 and 906 could be characterized as a "super-theme" of the themes that can be provided in association with semantic proxies. Moreover, the themes can be reorganized, dropping some words from a first theme in favor of words from a second theme to create a third theme. Indeed the themes used by a semantic proxy can for practical purposes be treated as any applicable permutation of words of the super-theme, which may or may not be the same as the themes identified in the modules 902 and 906 (and may even include words not identified in the modules 902 and 906).

[00133] In the example of FIG. 9, the flowchart 900 continues to decision point 912 where it is determined whether there are more themes to consider. If it is determined there are more themes to consider (912-Y), then the flowchart 900 returns to module 906 where another theme is computed and the flowchart continues as described previously. If, on the other hand, it is determined there are no more themes to consider (912-N), then the flowchart 900 ends. At this point, a super-theme has been identified for an audience persona to which content associated with campaign assets is to be targeted.

[00134] FIG. 10 depicts a diagram 1000 an audience persona graph mapped onto campaign asset subgraphs. FIG. 10 is intended to conceptually illustrate how an audience persona can be mapped to a campaign assets graph. The diagram 1000 includes nodes 1002 to 1028; an audience persona graph comprising nodes 1006, 1012, 1014, 1022, 1026, and 1028; a campaign asset subgraph 1032 comprising nodes 1002, 1006, and 1008; and a campaign asset subgraph 1034 comprising nodes 1010 and 1018. Themes that can be identified include the campaign asset nodes and any nodes in the campaign asset subgraphs. For example, a first theme includes the nodes 1002, 1006, 1008, 1012, 1014, 1022, 1024, 1026, and 1028; that is, the union of the audience persona graph 1030 and the campaign asset subgraph 1032. The campaign asset subgraph 1032 has nodes that overlap with the audience persona, suggesting this theme is an acceptable one.

[00135] The campaign asset subgraph 1034 has no nodes that overlap with the audience persona. However, the nodes 1010 (in the campaign asset subgraph 1034) and 1012 (in the audience persona graph 1030) have the same parent node 1004. Graph traversal yields a new theme that includes the union of the campaign asset subgraph 1034, the audience persona graph 1030, and the node 1004, which is an intervening node. For illustrative purposes, assume the relationship between nodes is estimated by taking the product of intervening edge weights. In this illustrative example, the relationship between the nodes 1010 and 1012 is the product of the weights of the edges between nodes 1010 and 1004 and between nodes 1004 and 1012. If the product is higher than an acceptable relationship weight threshold, the theme comprising the union of the campaign asset subgraph 1034, the audience persona graph 1030, and the node 1004 can be added to a super-theme. If the product is lower than an acceptable relationship weight threshold, the theme can be discarded.

[00136] Using the principles described in association with FIG. 10, it should be possible to build a graph of practically any size, subject to technological restraints, and determine a super-theme for an audience persona mapped to a campaign assets graph. It may be noted, combinations that work particularly well can be characterized as "successful themes," which may or may not be the same as the themes identified using this process. For example, it could be discovered that a successful theme for a campaign asset associated with the campaign asset subgraph 1032 includes a bag of words associated with nodes 1012 and 1026 (which could be represented by adding an edge between the two nodes with an appropriate weight), but is less successful if any of the words associated with the originally-identified theme are included.

[00137] FIG. 11 depicts a flowchart 1100 of an example of a method for combining audience persona themes with a trends graph. The flowchart 1100 starts at module 1102 with traversing an ontological graph to connect a granular trend characteristic to a granular market segment characteristic.

[00138] In the example of FIG. 11, the flowchart 1100 continues to decision point 1104 where it is determined what nodes exist between a first node in a persona graph and one or more nodes in at least a subset of trend subgraphs. If it is determined there are additional graph traversals to do (1104-Y), the flowchart 1100 returns to module 1102 where the ontological graph is traversed between two more nodes. This loop between module 1102 and decision point 1104 can at least in theory continue until each intervening node between each node of an audience persona graph and each node of a trends graph is identified.

[00139] If, on the other hand, it is determined there are no additional nodes for which graph traversal is necessary (1104-N), the flowchart 1100 continues to module 1106 where granular characteristic node relationship weight is computed. As was discussed above with reference to FIG. 4, weights can be computed in a number of ways, a simple example being computing a product of edge weights for each edge between a first node and a second node. In an example that includes an edge weight of 0 to 1, an edge weight of 0 would essentially result in "no relationship." At this point, each market segment characteristic of an audience persona can, at least in theory, have an associated weight vis-a-vis each node in a trends graph.

[00140] In the example of FIG. 11, the flowchart 1100 continues to decision point 1108 where it is determined whether to accept a theme for an audience persona. The theme for the audience persona can include a set of bags of words obtained through ontological graph traversal comprising a granular characteristic of the original audience persona graph; the granular characteristics of intervening nodes, if any; and the granular characteristic of one of the original trend subgraphs. For example, if a granular characteristic of the original audience persona graph is "mobile user," an intervening node is "gamer," and the granular characteristic of one of the original trend subgraphs is "MINECRAFT®," the associated bag of words is {Mobile User, Gamer, MINECRAFT®}. Each audience persona granular characteristic can have an associated bag of words with respect to each trend granular characteristic, the combination of which can be characterized as a "theme." It may be noted when a semantic proxy is generated, the theme can be pared down to a set of themes, each with a subset of potentially overlapping word sets.

[00141] If it is determined an identified theme is not to be accepted (1108-N), the flowchart 1100 returns to module 1106 to consider a next identified theme. The granular characteristics that are considered inadequate for the purpose of generating a bag of words generally do not have a sufficient weight. For example, for weights that do not exceed the relationship weight threshold, the bag of words are not added to the super-theme. It may be noted it is still possible to provide semantic proxies for trends for which there is no bag of words connecting granular characteristics of the audience persona and granular characteristics of the trend; it is possible to simply rely upon campaign asset characteristics to guess an audience persona member will be interested in the semantic proxy of the campaign asset (and perhaps edge weights can be updated if a relationship is discovered).

[00142] If, on the other hand, it is determined an identified theme is to be accepted

(1108-Y), the flowchart 1100 continues to module 1110 where the theme is added to a super- theme. The granular characteristics that are considered adequate for the purpose of generating a bag of words generally must have a sufficient weight. For example, if the weight of a product of edges between an audience persona granular characteristic and a trend granular characteristic exceeds a relationship weight threshold, the bag of words can be added to the super-theme. To distinguish, the acceptable themes identified in the modules 1102 and 1106 could be characterized as a "super-theme" of the themes that can be provided in association with semantic proxies. Moreover, the themes can be reorganized, dropping some words from a first theme in favor of words from a second theme to create a third theme. Indeed the themes used by a semantic proxy can for practical purposes be treated as any applicable permutation of words of the super-theme, which may or may not be the same as the themes identified in the modules 1102 and 1106 (and may even include words not identified in the modules 1102 and 1106).

[00143] In the example of FIG. 11, the flowchart 1100 continues to decision point 1112 where it is determined whether there are more themes to consider. If it is determined there are more themes to consider (1112-Y), then the flowchart 1100 returns to module 1106 where another theme is computed and the flowchart continues as described previously. If, on the other hand, it is determined there are no more themes to consider (1112-N), then the flowchart 1100 ends. At this point, a super- theme has been identified for an audience persona to which content associated with trends is to be targeted.

[00144] FIG. 12 depicts a flowchart 1200 of an example of a method for building an interest graph. In the example of FIG. 12, the flowchart 1200 starts at module 1202 with merging a campaign assets super-theme and trends super-theme. The merging of the super- themes can involve graph traversal to link nodes associated with campaign assets with nodes associated with trends.

[00145] In the example of FIG. 12, the flowchart 1200 continues to module 1204 with identifying a relationship weight between trends and campaign assets for an audience persona. The themes identified by mapping the audience persona to the trends and by mapping the audience persona to the campaign assets are not necessarily the strongest themes that can be identified because trends and campaign assets might be more tightly related than the trends-to- persona and campaign asset-to-persona.

[00146] In the example of FIG. 12, the flowchart 1200 continues to module 1206 with identifying successful themes for semantic proxy generation, if any. Successful themes can be identified by observing feedback from past or current campaigns. If a theme is determined to have elicited interest in semantic proxy content, edges associated with the nodes of the theme can be more heavily weighted, causing the successful themes to be identifiable through their relationship weights. Alternatively, the themes can be identified explicitly and reused without, or in addition to, considering the relationship weight of the theme's bag of words.

[00147] In the example of FIG. 12, the flowchart 1200 continues to module 1208 with defining new most-relevant themes for semantic proxy generation, if needed. If there are no successful themes yet identified (assuming an implementation of a relevant system is capable of identifying successful themes), the themes that are predicted to be most successful are the ones that most effectively link trends to the audience segment and/or link trends to the relevant campaign asset and/or link the audience segment to the relevant campaign asset.

[00148] In the example of FIG. 12, the flowchart 1200 continues to module 1210 with defining experimental themes for semantic proxy generation. The module 1210 is optional. It may or may not be considered valuable to experiment with small sample sizes of semantic proxies by creating themes that do not necessarily have the tightest relationship between component words. This approach can serve to identify undetected desirable themes by determining whether the experimental themes are successful. In a specific implementation, if the relationship between words of a most-relevant theme are below a confidence threshold, experimental themes can be generated to attempt to find more relevant themes using feedback. In an alternative implementation, the experimental themes could be generated for a sample size sufficiently small as to not significantly impact the overall cost of distributing the semantic proxies, and with the potential benefit of improving the cost-effectiveness of semantic proxy distribution by identifying new successful themes.

[00149] FIG. 13 depicts a diagram 1300 illustrating generation of a semantic proxy for distribution by one or more distribution networks. The diagram 1300 includes a semantic proxy set 1302 (illustrated as dashed boxes 1302-1 to 1302-n) and a distributor interface engine 1304. [00150] Each of the semantic proxies of the semantic proxy set 1302 includes a format block 1306, a network ID block 1308, an image block 1310, a message block 1312, and a theme block 1314.

[00151] The format block 1306 is intended to represent a data structure with structural characteristics suitable for properly presenting the semantic proxy on the relevant distribution network. The format block 1306 may or may not also include a data structure suitable for getting the semantic proxy to the distribution network. The format block 1306 receives input associated with a persona graph. Specifically by way of example, the persona graph can include segment characteristics such as "mobile user," which can have an impact on the format associated with a semantic proxy. The format block 1306 receives input associated with a trends graph. Specifically by way of example, the trends graph can include trend characteristics such as where content is trending such as "FACEBOOK®," which can have an impact on the format because the semantic proxy may be most effective where the trend is trending and should therefore be sent there and the format must be a format that is acceptable to the applicable distribution network.

[00152] The network ID block 1308 is intended to represent a data structure sufficient to identify a distribution network to which the semantic proxy is being sent for distribution. Different distribution networks may receive semantic proxies in different ways, such as email, web page submissions, ftp, etc. The network ID block 1308 receives input associated with a trends graph. Specifically by way of example, the trends graph can include trend characteristics such as where content is trending, which is where it may be determined the semantic proxy should be sent.

[00153] The image block 1310 is intended to represent a typical manner of providing content (i.e. in the form of an image). Other content is also possible. The image block 1310 is associated with data from an asset content datastore. One example of a semantic proxy image is an image suitable for display in a banner ad.

[00154] The message block 1312 is intended to represent a typical manner of providing content (i.e. in the form of a message for display in association with the image). Other content is also possible. The message block 1312 is associated with data from an asset content datastore. One example of a semantic proxy message is a message suitable for display in association with an image in a banner ad. [00155] The theme block 1314 is intended to represent a bag of words that have been identified as effective in association with the campaign asset and/or audience persona associated with the semantic proxy. When the distribution network identifies content that includes a word or words from the theme, the distribution network knows to display at least a portion of the semantic proxy to a consumer of the content including the word or words.

[00156] In the example of FIG. 13, the distributor interface engine 1304 is coupled to the semantic proxy sets 1302. The distributor interface engine 1304 is responsible for sending the semantic proxy sets to the appropriate distribution networks.

[00157] FIG. 14 depicts a flowchart 1400 of an example of a method for real-time audience segment behavior prediction and semantic proxy generation. The flowchart 1400 starts at module 1402 with identifying an applicable trend for an audience persona. A trend can be characterized as applicable if words associated with the trend are represented as nodes in a graph and words associated with the audience persona are mapped onto the graph. Overlapping words may establish applicability, but if there is no overlap of words, graph traversal of an ontology can establish a path from nodes associated with trends to nodes associated with the audience persona. The closer the connection, the higher the applicability. If a weight associated with the connection exceeds an applicability threshold, a trend can be characterized as applicable for the audience persona.

[00158] In the example of FIG. 14, the flowchart 1400 also starts at module 1404 with identifying an applicable campaign asset for the audience persona. A campaign asset is typically selected from a campaign asset class provided by an entity interested in drawing interest to the campaign assets. A campaign asset can be characterized as applicable if words associated with the campaign asset are represented as nodes in a graph and words associated with the audience persona are mapped onto the graph. Overlapping words may establish applicability, but if there is no overlap of words, graph traversal of an ontology can establish a path from nodes associated with campaign assets to nodes associated with the audience persona. The closer the connection, the higher the applicability. If a weight associated with the connection exceeds an applicability threshold, a campaign asset can be characterized as applicable for the audience persona.

[00159] In the example of FIG. 14, the flowchart 1400 continues to module 1406 with building an interest graph including themes associated with the audience persona. Depending upon implementation- specific or other considerations, an interest graph can be built at module 1406 from either or both a trend applicability graph and a campaign asset applicability graph. A trend applicability graph that can be used to build an interest graph can be generated by combining one or a plurality of trend graphs or subgraphs with one or a plurality of audience persona graphs or subgraphs. A campaign asset applicability graph that can be used to build an interest graph can be generated by combining one or a plurality of asset graphs or subgraphs with one or a plurality of persona graphs or subgraphs. Further depending upon implementation- specific or other considerations, an interest graph can be built using either or both a pruned trend applicability subgraph and a pruned campaign asset applicability subgraph. A pruned trend applicability subgraph that is used to build an interest graph can be generated by removing nodes from a trend applicability graph. A pruned campaign asset applicability subgraph that is used to build an interest graph can be generated by removing nodes from a campaign asset applicability graph.

[00160] In the example of FIG. 14, the flowchart 1400 continues to module 1408 with generating a set of semantic proxies in accordance with themes in the interest graph and distribution network preferences. In generating a set of semantic proxies, themes, including words associated with a campaign asset, can be extracted from the interest graph. Words associated with a campaign asset can be used to determine whether to present content for a campaign asset that is the subject of a semantic proxy. Semantic proxies generated at module 1408 can also include content for a campaign asset that is presented to a consumer. Content can include images and/or messages used in advertising a campaign asset.

[00161] In the example of FIG. 14, the flowchart 1400 continues to module 1410 with distributing the set of semantic proxies to distribution networks. Distribution networks can include applicable networks for utilizing the set of semantic proxies to display content associated with campaign assets that are the subject of the set of semantic proxies. Example of distribution networks include but are not limited to FACEBOOK® , YOUTUBE®, and TWEETER®.

[00162] In the example of FIG. 14, the flowchart 1400 continues to module 1412 with receiving feedback sufficient to determine whether a theme was effective. Depending upon implementation- specific or other considerations, feedback sufficient to determine whether a theme was effective can be received from distribution network to which the set of semantic proxies are distributed. Further depending upon implementation-specific or other considerations, feedback sufficient to determine whether a theme was effective can be received from a distributor of campaign assets or a creator of campaign assets. For example, feedback can be generated from a manufacturer of a device based on the number of times the device sold.

[00163] In the example of FIG. 14, the flowchart 1400 continues to module 1414 with improving semantic models for trend applicability and campaign asset applicability using feedback. Depending upon implementation- specific or other considerations, in improving semantic models based on feedback, edge weights of an campaign asset applicability graph can be changed based on the effectiveness of the themes. For example, if it is determined for a particular campaign that the relative weight between two nodes of an ontology is less than usual, the relevant edge can be modified to effect the reality of the relationship for the particular campaign. Further depending upon implementation- specific or other considerations, network preferences included in the set of semantic proxies can be modified based on feedback. The flowchart 1400 then returns to modules 1402 and 1404 and continues as described previously.

[00164] FIG. 15 depicts a flowchart 1500 of an example of a method for distribution of at least a portion of a semantic proxy to consumers of thematically compatible content. The flowchart 1500 starts at module 1502 with receiving a semantic proxy. A semantic proxy received at module 1502 can be created using an interest graph. A semantic proxy received at module 1502 can be used to determine themes that are used to determine when to present content associated with campaign assets associated with the semantic proxy.

[00165] In the example of FIG. 15, the flowchart 1500 continues to module 1504 with serving at least a portion of the semantic proxy on a consumer of content identifiable as associated with a theme of the semantic proxy. In serving content to a consumer, the content can be displayed on a semantic proxy presentation system. Depending upon implementation- specific or other considerations, the semantic proxy presentation system can include a applicable systems controlled by content consumers or downstream broadcasters.

[00166] In the example of FIG. 15, the flowchart 1500 continues to module 1506 with collecting data associated with the presentation of the at least a portion of the semantic proxy to the consumer of the thematically compatible content. Data associated with the presentation of the at least a portion of the semantic proxy can include feedback as to the success of a semantic proxy based on a theme extracted from the semantic proxy.

[00167] In the example of FIG. 15, the flowchart 1500 ends at module 1508 with providing the data as feedback to a semantic proxy generation system. Of course, the flowchart 1500 could continue in the sense the semantic proxy can continue to be distributed on the distribution network. Also, the flowchart could be restarted by receiving a new semantic proxy.

[00168] These and other examples provided in this paper are intended to illustrate but not necessarily to limit the described implementation. As used herein, the term "implementation" means an implementation that serves to illustrate by way of example but not limitation. The techniques described in the preceding text and figures can be mixed and matched as circumstances demand to produce alternative implementations.

Claims

CLAIMS We claim:

1. A method comprising:

obtaining market segmentation parameters;

obtaining campaign asset parameters;

obtaining trend parameters;

generating a semantic proxy for a campaign asset based on trends applicable to audience personas using the market segmentation parameters, the campaign asset parameters, and the trend parameters;

providing the semantic proxy to a proxy distribution system for use in determining whether to present content associated with the campaign asset to a consumer.

2. The method of claim 1, further comprising:

determining a campaign asset description of the campaign asset from the campaign asset parameters;

associating words included in the campaign asset description with asset ontological nodes included in an asset ontology to generate a campaign asset graph including the campaign asset;

generating the semantic proxy for the campaign asset using the campaign asset graph.

3. The method of claim 1, further comprising:

determining a market segmentation description for a market segment from the market segmentation parameters;

associated words included in the market segmentation description with audience ontological nodes included in an audience ontology to generate an audience persona graph including the market segment;

generating the semantic proxy for the campaign asset using the audience persona graph.

4. The method of claim 1, further comprising:

determining at least one word that semantically identifies characteristics of a trend from the trend parameters; associating the at least one word that semantically identifies the characteristics of the trend ontological nodes in a trend ontology to generate a trend graph including the trend; generating the semantic proxy for the campaign asset using the trend graph.

5. The method of claim 4, further comprising:

receiving content associated with the trend, the content associated with the trend included as part of the trend parameters;

determining whether the content is viral if the content exceeds a virality threshold; if it is determined that the content is viral, determining the at least one word that semantically identifies the characteristics of the trend from the content associated with the trend.

6. The method of claim 1, further comprising:

generating a campaign asset graph using the campaign asset parameters;

generating an audience persona graph using the market segmentation parameters; generating a trend graph using the trend parameters;

mapping the audience persona graph to the campaign asset graph to generate a campaign asset applicability graph;

mapping the audience persona graph to the trend graph to generate a trend applicability graph;

generating an interest graph using the campaign asset applicability graph and the trend applicability graph;

generating the semantic proxy for the campaign asset using the interest graph.

7. The method of claim 6, further comprising:

determining a theme for the campaign asset using the interest graph, the theme including a plurality of words associated with the campaign asset used in determining whether to present the content associated with the campaign asset;

generating the semantic proxy to include the at least one word associated with the campaign asset.

8. The method of claim 7 further comprising:

determining the effectiveness of the theme; modifying edge weights between nodes in the campaign asset applicability graph according to the effectiveness of the theme.

9. The method of claim 7 further comprising:

generating network preferences regarding distribution of content associated with the campaign asset;

distributing the semantic proxy according to the network preferences;

determining the effectiveness of the theme;

modifying the network preferences according to the effectives of the theme to generate modified network preferences;

distributing the semantic proxy according to the modified network preferences.

10. The method of claim 6, further comprising;

traversing the campaign asset applicability graph to connect granular campaign asset characteristics of the campaign asset to granular market segment characteristics;

determining campaign asset node relationship weights between the connected granular campaign asset characteristics and the granular market segment characteristics;

determining campaign asset super- themes based on the campaign asset node

relationship weights;

traversing the trend applicability graph to connect granular trend characteristics to the granular market segment characteristics;

determining trend node relationship weights between the connected granular trend characteristics and the market segment characteristics;

determining trend super- themes based on the trend node relationship weights;

generating the semantic proxy based on the campaign asset super-themes and the trend super- themes.

11. The method of claim 10, further comprising:

merging the asset super-themes and the trend-super themes to create the identity graph; traversing the identity graph to link nodes associated with trends in the identity graph to nodes associated with campaign assets, including the campaign asset, in the identity graph; determining relationship weights of the linked nodes associated with trends and the nodes associated with campaign assets in the identity graph; determining a theme from the relationship weights of the linked nodes associated with trends and the nodes associated with campaign assets in the identity graph, the theme including a plurality of words associated with the campaign asset used in determining whether to present the content associated with the campaign asset;

generating the semantic proxy for the campaign asset using the theme.

12. A system comprising:

a market segmentation parameters provider interface engine configured to obtain market segmentation parameters;

a campaign asset class parameters provider interface engine configured to obtain campaign asset parameters;

a data stream monitoring engine configured to obtain trend parameters;

a semantic proxy set generation engine configured to generate a semantic proxy for a campaign asset based on trends applicable to audience personas using the market segmentation parameters, the campaign asset parameters, and the trend parameters;

a distributor interface engine configured to provide the semantic proxy to a proxy distribution system for use in determining whether to present content associated with the campaign asset to a consumer.

13. The system of claim 12, further comprising:

the campaign asset class parameters providers interface engine further configured to determine a campaign asset description of the campaign asset from the campaign asset parameters;

a semantic enrichment engine configured to associate words included in the campaign asset description with asset ontological nodes included in an asset ontology to generate a campaign asset graph including the campaign asset;

the semantic proxy set generation engine further configured to generate the semantic proxy for the campaign asset using the campaign asset graph.

14. The system of claim 12, further comprising:

the market segmentation parameters provider interface engine further configured to determine a market segmentation description for a market segment from the market segmentation parameters; a semantic enrichment engine configured to associate words included in the market segmentation description with audience ontological nodes included in an audience ontology to generate an audience persona graph including the market segment;

the semantic proxy set generation engine further configured to generate the semantic proxy for the campaign asset using the audience persona graph.

15. The system of claim 12, further comprising:

a virality determination engine configured to determine at least one word that semantically identifies characteristics of a trend from the trend parameters;

a semantic enrichment engine configured to associate the at least one word that semantically identifies the characteristics of the trend ontological nodes in a trend ontology to generate a trend graph including the trend;

the semantic proxy set generation engine further configured to generate the semantic proxy for the campaign asset using the trend graph.

16. The system of claim 15, wherein the virality determination engine is further configured to:

receive content associated with the trend, the content associated with the trend included as part of the trend parameters;

determine whether the content is viral if the content exceeds a virality threshold;

if it is determined that the content is viral, determine the at least one word that semantically identifies the characteristics of the trend from the content associated with the trend.

17. The system of claim 12, further comprising:

a semantic enrichment engine configured to:

generate a campaign asset graph using the campaign asset parameters;

generate an audience persona graph using the market segmentation parameters; generate a trend graph using the trend parameters;

a campaign asset applicability designation engine configured to map the audience persona graph to the campaign asset graph to generate a campaign asset applicability graph; a trend applicability designation engine configured to map the audience persona graph to the trend graph to generate a trend applicability graph; an interest graph building engine configured to generate an interest graph using the campaign asset applicability graph and the trend applicability graph;

the semantic proxy set generation engine further configured to generate the semantic proxy for the campaign asset using the interest graph.

18. The system of claim 17, wherein the semantic proxy set generation engine is further configured to:

determine a theme for the campaign asset using the interest graph, the theme including a plurality of words associated with the campaign asset used in determining whether to present the content associated with the campaign asset;

generate the semantic proxy to include the at least one word associated with the campaign asset.

19. The system of claim 18, further comprising:

a feedback interpretation and incorporation engine configured to determine the effectiveness of the theme;

the campaign asset applicability designation engine further configured to modify edge weights between nodes in the campaign applicability graph according to the effectiveness of the theme.

20. The system of claim 18, further comprising:

a network preferences selection engine configured to:

generate network preferences regarding distribution of content associated with the campaign asset, the semantic proxy distributed according to the network preferences;

modify the network preferences according to the effectiveness of the theme to generate modified network preferences;

the distributor interface engine further configured to distribute the semantic proxy according to the modified network preferences.

21. The system of claim 17, wherein:

the campaign asset applicability designation engine further configured to: traverse the campaign asset applicability graph to connect granular campaign asset characteristics of the campaign asset to granular market segment characteristics;

determine campaign asset node relationship weights between the connected granular campaign asset characteristics and the granular market segment characteristics;

determine campaign asset super-themes based on the campaign asset node relationship weights;

the trend applicability designation engine further configured to:

traverse the trend applicability graph to connect granular trend characteristics to the granular market segment characteristics;

determine trend node relationship weights between the connected granular trend characteristics and the market segment characteristics;

determine trend super- themes based on the trend node relationship weights; the semantic proxy set generation engine further configured to generate the semantic proxy based on the campaign asset super-themes and the trend super-themes.

22. The system of claim 21, wherein:

the interest graph building engine is further configured to:

merge the asset super-themes and the trend-super themes to create the identity graph;

traverse the identity graph to link nodes associated with trends in the identity graph to nodes associated with campaign assets, including the campaign asset, in the identity graph;

determine relationship weights of the linked nodes associated with trends and the nodes associated with campaign assets in the identity graph;

the semantic proxy set generation engine further configured to:

determine a theme from the relationship weights of the linked nodes associated with trends and the nodes associated with campaign assets in the identity graph, the theme including a plurality of words associated with the campaign asset used in determining whether to present the content associated with the campaign asset;

generate the semantic proxy for the campaign asset using the theme.

A system comprising:

means for obtaining market segmentation parameters; means for obtaining campaign asset parameters;

means for obtaining trend parameters;

means for generating a semantic proxy for a campaign asset based on trends applicable to audience personas using the market segmentation parameters, the campaign asset parameters, and the trend parameters;

means for providing the semantic proxy to a proxy distribution system for use in determining whether to present content associated with the campaign asset to a consumer.