GB2549835A

GB2549835A - Method and system for influencing digital content or access to content

Info

Publication number: GB2549835A
Application number: GB1703435.6A
Authority: GB
Inventors: Smith Tom; Carlson David; Hackett Christopher; Warford Ian; Degroot Nelius
Original assignee: Mporium Group PLC
Current assignee: Mporium Group PLC
Priority date: 2016-03-03
Filing date: 2017-03-03
Publication date: 2017-11-01
Also published as: WO2017149321A1; GB201703435D0; EP3424223A1; GB201603685D0

Abstract

A method of influencing access to website, social media or mobile application content is described comprising identifying the occurrence of an item of live or previously broadcast media content, identifying consumer Internet activity at the time of the occurrence of the item of media content, and identifying, from said detected consumer internet activity, a strategy to influence an Internet search function, social media site or mobile application to display selected information. Then when a subsequent broadcast of the item of media content is identified, and the identified strategy is triggered at the point the item occurs during the time of the identified subsequent broadcast. The identification of the occurrence of an item of media content may comprise interrogating a scheduling database. The step of identifying the trigger term associated with the media content may comprise extracting a list of entities from the program information.

Description

Method and System for Influencing Digital Content or Access to Content

The present invention relates to a method and apparatus for influencing content or access to content. Embodiments of the present invention relate to influencing access to website, social media or mobile application content, and to influencing website, social media or mobile application content.

It is common for consumers of media content to be using multiple devices at once. For example, a user may be watching a television programme while using a smartphone or tablet computer to surf the internet or interact with social media platforms such as Facebook. Sometimes a user may be motivated by television content (for example) to interact with their smartphone or tablet computer to find out more about something they saw on the television.

One aspect of the invention provides a method of influencing website, social media or application content or access to content, comprising: identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content substantially at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the item of live or previously broadcast media content trigger term; identifying, from said detected variation in consumer internet activity and the correlation data, a strategy to modify the content or access to the content of a web site, web page, social media site or application; triggering the identified modification substantially at the point the trigger term occurs during a subsequent media broadcast.

In certain embodiments, the method comprises receiving and processing by a computer processor an input signal comprising broadcast media metadata.

Optionally, the method comprises receiving a request for content from a user device;

In certain embodiments, the method comprises the step of in response to the request for content automatically generating an output to execute the strategy and trigger the modification according to the correlation data;

The method may comprise transmitting the content to the user device over a network.

The user device may comprise an unknown user device.

Optionally, identifying the variation in the level of consumer Internet activity comprises analysing aggregate Internet user data.

Optionally, identifying the variation in the level of consumer Internet activity comprises analysing one or more sources of time series data.

The user or user may be anonymous. The data may comprise anonymised data.

The user device broadcast media content may be external to the user device.

The user device may be outwith a network connection to the media broadcast.

In certain embodiments, the level of consumer Internet activity is from a predetermined geographical area.

Optionally, the item of live or previously broadcast media content is broadcast to substantially the same predetermined geographical area.

In certain embodiments, the step of identifying a variation in a level of consumer Internet activity comprises analysing the time series data to generate an anomaly event indicative of a relevant variation in the level of consumer Internet activity in relation to a keyword. A relevant variation may be calculated using an algorithm.

In certain embodiments, the step of identifying a variation in a level of consumer Internet activity comprises analysing time series data to generate an anomaly event indicative of a relevant variation in the level of consumer Internet activity in relation to a keyword within a defined time period.

The step of generating and storing correlation data may be based on the analysis.

In certain embodiments, the step of identifying the occurrence of an item of media content and/or identifying the trigger term comprises receiving and processing by a computer processor a signal feed input comprising media content metadata.

In certain embodiments, the step of identifying the occurrence of an item of media content and/or identifying the trigger term comprises substantially continuously receiving and processing by a computer processor a signal feed input comprising media content metadata. This may be an RSS feed.

Influencing website, social media or application content or access to content may comprise, for example, influencing online search results.

One aspect of the invention provides a method of influencing access to website, social media or application content, comprising: identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the trigger term; identifying, from said detected variation in consumer internet activity and the correlation data, a strategy to influence an Internet search function, social media site or application to display selected information; triggering the identified strategy at the point the trigger term occurs during a subsequent media broadcast.

Influencing website, social media or application content may comprise influencing access to content.

This may, for example, comprise influencing online search results.

Optionally, identifying the occurrence of an item of media content comprises interrogating a scheduling database to determine a broadcast time for the media item.

In certain embodiments, identifying the occurrence of an item of media content comprises detecting the item of media content in a live media stream.

The step of identifying the trigger term associated with the media content may comprise interrogating a scheduling database to determine programme information.

The step of identifying the trigger term associated with the media content may comprise extracting a list of entities from the programme information.

Advantageously entities such as celebrity names, descriptions, locations etc may be extracted and classified.

Statistical analysis may be performed using an algorithm to assign probability of an association between a predetermined keyword and a trigger term and/or of a trigger term in a media stream influencing internet search in relation to the keyword.

The statistical analysis may compare consumer Internet activity in relation to keywords and trigger terms at a plurality of time points and/or different TV channels, different programmes and/or periods of time.

The probability that an increase in consumer Internet activity in relation to a predetermined keyword is related to an entity and/or trigger term in media content may be assigned.

Trigger terms may be assigned a weighting based on correlation with variation or level of variation in internet search activity in relation to predetermined keywords.

In certain embodiments, the step of identifying a trigger term associated with the media content comprises extracting caption information and/or subtitle information associated with a television programme broadcast or video on demand stream.

Optionally, the step of identifying a trigger term associated with the media content comprises extracting caption information and/or subtitle information associated a television programme broadcast or video on demand stream at the time of the occurrence of the variation in consumer Internet activity;

The subtitle information may be embedded within image data within the television broadcast signal or the video on demand stream, and wherein the subtitle information is converted into text using optical character recognition.

Identifying the variation in the level of consumer Internet activity in relation to the predetermined keyword may comprise identifying an increase in Internet searches conducted using the predetermined keyword.

Optionally, the variation in the level of consumer Internet activity comprises an increase in the level of consumer Internet search activity above a threshold value.

In certain embodiments, the step of identifying the strategy comprises specifying an Internet search function to be influenced based on trigger terms and/or keywords determined to result in a level of consumer search activity which is greater than a threshold value.

The method may comprise identifying a correlation between the variation in consumer Internet activity in relation to the predetermined keyword and a variation in consumer Internet activity in relation to the trigger term.

Optionally, identifying the variation in the level of consumer Internet activity in relation to the trigger term comprises identifying an increase in Internet searches conducted using the trigger term.

The keyword and/or trigger term may each comprise a single word or a plurality of words.

The keyword may be conceptually unrelated to the trigger term.

The Internet activity may be an Internet search carried out by a consumer. The Internet activity may be consumer access to website, social media or application content received over the Internet.

The selected information may be predetermined or dynamically generated.

In certain embodiments, the selected information comprises a link to a web page or images and text.

Optionally, the step of generating and storing correlation data indicative of an association between the keyword and the trigger term comprises identifying a correlation between the occurrence of a trigger term associated with media content and an increase in Internet searches conducted using the predetermined keyword.

Optionally, the step of generating and storing correlation data indicative of an association between the keyword and the trigger term comprises storing trigger terms in a trigger term database.

The method may comprise maintaining a database of stored trigger terms associated with websites.

The method may comprise the step of generating a trigger term trie from the stored trigger terms. Optionally, the trigger term trie comprises a dynamic trie.

Internet activity may comprise Internet search activity.

The method may comprise generating and storing correlation data indicative of the impact of the item of media content on Internet searches conducted using the trigger words, or using keywords associated with the trigger terms.

The method may comprise generating and storing correlation data indicative of the impact of the trigger terms in media content on Internet searches.

Optionally, the method comprises selecting information to be displayed as part of an Internet search function, social media site or application in dependence on the correlation information.

The method may comprises combining the predetermined keyword and trigger term to create a further trigger term.

The method may include the step of identifying a variation in a level of consumer Internet activity in relation to the trigger term.

Trigger terms generated may be further investigated to analyse search activity around these terms.

Optionally, the method comprises an initial step of selecting the predetermined keyword by interrogating or scraping a retail website.

Optionally, the method may comprise identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword by receiving first time series data in relation to the predetermined keyword from a first advertising platform; receiving second time series data in relation to the predetermined keyword from a second advertising platform; identifying at least one anomaly in the first time series data and at least one anomaly in the second time series data; generating and storing correlation data indicative of an association between the at least one anomaly in the first time series data and the at least one anomaly in the second time series data; and based on the correlation, generating an anomaly event indicative of a variation in the level of consumer Internet activity in relation to the keyword within a defined time period;

One aspect of the invention provides apparatus for influencing website, social media or application content or access to the content, comprising: an Internet activity identifier, for identifying a variation in level of consumer Internet activity in relation to a particular keyword; a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in level of consumer Internet activity and identifying a trigger term associated with the media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or application; a subsequent broadcaster identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified modification to the content of the web site, web page, social media site or application at the point the trigger term occurs during the time of the identified subsequent broadcast.

One aspect of the invention provides a non-transitoiy computer readable storage medium providing instructions for influencing website, social media or application content and/or access to the content; the instructions when executed by a computing device, performing a method comprising identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content substantially at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the item of live or previously broadcast media content trigger term; identifying, from said detected variation in consumer internet activity and the correlation data, a strategy to modify the content or access to the content of a web site, web page, social media site or application; triggering the identified modification substantially at the point the trigger term occurs during a subsequent media broadcast.

Another aspect of the invention provides a system for influencing website, social media or application content and/or access to the content comprising processing apparatus and a non-transitory computer readable storage medium providing instructions for influencing website, social media or application content and/or access to content; wherein the processing apparatus comprises an Internet activity identifier, for identifying a variation in level of consumer Internet activity in relation to a particular keyword; a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in level of consumer Internet activity and identifying a trigger term associated with the media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or application; a subsequent broadcaster identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified modification to the content of the web site, web page, social media site or application at the point the trigger term occurs during the time of the identified subsequent broadcast.

Another aspect of the invention provides a method of influencing access to website, social media or application content, comprising: detecting an occurrence of a trigger term in a real time media source by matching data extracted from the real time media source against a trigger term trie; and influencing an Internet search function or social media site or application to display predetermined information in response to the detection.

Preferably, the trigger term trie comprises a dynamic trie. This provides the advantage that the trie may be modified any number of times. For example, when new trigger terms are added or trigger terms are removed.

The method may comprise sequencing the data extracted from the real time media source. This provides the advantage that data may be correctly ordered or re-ordered during processing.

The method may comprise processing the data extracted from the real time media source to remove text duplications.

The method may comprise collecting extracted data in an array or stack. Preferably, the method comprises collecting extracted text data in a mutable array. Advantageously, text elements may be output from the array or stack to the trigger term trie. The array or stack may be of fixed-size. The rolling array or stack may be configured to collect a predetermined number of elements.

The rolling array or stack may be configured to collect between 5 and 30 elements; Preferably the rolling array or stack is configured to collect between 10 and 20 elements; More preferably, the rolling array or stack is configured to collect around 15 elements. The elements may comprise words of text.

This provides the advantage that at the point at which a match is detected, the entire contents of the array may be processed and optionally stored for context analysis.

The array may be a rolling array. This provides the advantage that as one text element is output from the array or stack to the trigger term trie, another text element of extracted data from the data stream is input to the array or stack in a substantially continuous process. Each text element output from the array or stack may comprise a single word.

In certain embodiments, the predetermined information comprises a link to a web page. The predetermined information may comprise images and text. In certain embodiments, the method may comprise modifying the web page in response to the detection.

Optionally, the method comprises creating a trigger term trie from a database of stored trigger terms. Optionally, the method comprises maintaining a database of stored trigger terms associated with the trigger term trie, matching text extracted from the real time media source against the trigger term trie, and influencing the Internet search function or social media site to display a link to a web page of the associated website.

In certain embodiments, the Internet search function or social media site is influenced by the occurrence of the trigger term for only a predetermined period of time. The period of time is preferably around 1 hour or less. The period of time is more preferably around 30 minutes or less. The period of time may be around 15 minutes or less. The period of time may be around 5 minutes or less. In certain embodiments, the period of time is around 30 seconds or less.

The Internet search function or social media site may be influenced substantially immediately following the detection of the trigger term.

The Internet search function or social media site may be influenced within around 2 to 10 seconds or less following the detection of the trigger term.

It will be appreciated that the period of time may depend on the length of a frame of television subtitles.

The Internet search function or social media site may be influenced within around 5 seconds or less following the detection of the trigger term.

Influencing a search function or social media site may comprise providing one or more instructions to a search function, site or application.

In certain embodiments, the Internet search function or social media site may be influenced following a predetermined delay from the detection of the trigger term.

Optionally, one or more keywords are associated with the trigger term, and the Internet search function is influenced by triggering an Adwords campaign based on the keywords associated with the trigger term.

The media source may be one of a television channel, an on-demand video stream, a radio channel, a social media site, a news feed or a weather feed.

The media source may be a television channel or video on demand stream and the trigger term is detected in caption information and/or subtitle information associated with the television programme being broadcast or the video on demand stream.

In certain embodiments, subtitle information is embedded within image data associated with the television broadcast signal or the video on demand stream, and wherein the subtitle information is converted into text using optical character recognition.

Another aspect of the invention provides apparatus for influencing access to website, social media or application content, comprising: a trigger term detector for detecting an occurrence of a trigger term in a real time media source; and a controller for influencing an Internet search function, social media site or application to display predetermined information in response to the detection.

Another aspect of the invention provides a method of influencing website content, comprising: detecting an occurrence of a trigger term in a real time media source by matching data extracted from the real time media source against a trigger term trie; and modifying a web site or page in response to the detection.

Advantageously, the web site or page may be modified to display content related to the trigger term.

The method may comprise maintaining a database of stored trigger terms associated with the trigger term trie, matching text extracted from the real time media source against the trigger term trie, and modifying the associated web site or page.

Optionally, the web site or page is modified by the occurrence of the trigger term for only a predetermined period of time.

The period of time may be around 1 hour or less. The period of time may be around 30 minutes or less. The period of time may be around 15 minutes or less. The period of time may be around 5 minutes or less. In certain embodiments, the period of time is around 30 seconds or less.

In certain embodiments, the web site or page is influenced substantially immediately following the detection of the trigger term.

Advantageously, the Internet search function or social media site may be influenced within around 10 seconds or less following the detection of the trigger term.

In certain embodiments, the web site or page is modified following a predetermined delay from the detection of the trigger term.

The media source may be a television channel or a video on demand stream and the trigger term is detected in caption information and/or subtitle information associated with the television programme being broadcast or the video on demand stream. A further aspect of the invention comprises a method of determining keywords and/or trigger terms that result in a level of consumer Internet activity that is greater than a threshold value comprising: identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the trigger term; based on the correlation data executing a strategy to select and prioritise a subset of the predetermined keywords during a subsequent media broadcast.

Optionally, the method comprises influencing access to website, social media or application content, by detecting an occurrence of a trigger term in a real time media source by matching data extracted from the real time media source against the trigger term trie; and influencing an Internet search function or social media site or application to display predetermined information in response to the detection.

Optionally the method comprises influencing website content, comprising: detecting an occurrence of a trigger term in a real time media source by matching data extracted from the real time media source against the trigger term trie; and modifying a web site or page in response to the detection.

One aspect of the invention provides a method of influencing website, social media or application content, comprising: identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the trigger term; identifying, from said detected variation in consumer internet activity and the correlation data, a strategy to modify the content of a web site, web page, social media site or application; triggering the identified modification to the content of the web site, web page, social media site or application at the point the trigger term occurs during a subsequent media broadcast.

Another aspect of the invention provides an apparatus for influencing access to website, social media or application content, comprising: an Internet activity identifier, for identifying a variation in level of consumer Internet activity in relation to a particular keyword; a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in level of consumer Internet activity and identifying a trigger term associated with the media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to influence an Internet search function, social media site or application to display selected information; a subsequent broadcast identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified strategy at the point the trigger term occurs during the time of the identified subsequent broadcast.

Another aspect of the invention provides an apparatus for influencing website, social media or application content, comprising: an Internet activity identifier, for identifying a variation in level of consumer Internet activity in relation to a particular keyword; a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in level of consumer Internet activity and identifying a trigger term associated with the media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or application; a subsequent broadcaster identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified modification to the content of the web site, web page, social media site or application at the point the trigger term occurs during the time of the identified subsequent broadcast.

According to one aspect of the invention there is provided a method of influencing access to website, social media or mobile application content, comprising: identifying the occurrence of an item of live or previously broadcast media content; identifying consumer Internet activity at the time of the occurrence of the item of media content; identifying, from said detected consumer internet activity, a strategy to influence an Internet search function, social media site or mobile application to display selected information; identifying a subsequent broadcast of the item of media content; and triggering the identified strategy at the point the item occurs during the time of the identified subsequent broadcast.

According to another aspect of the invention, there is provided a method of influencing website, social media or mobile application content, comprising: identifying the occurrence of an item of live or previously broadcast media content; identifying consumer Internet activity at the time of the occurrence of the item of media content; identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or mobile application; identifying a subsequent broadcast of the item of media content; and triggering the identified modification to the content of the web site, web page, social media site or mobile application at the point the item occurs during the time of the identified subsequent broadcast.

Identifying the occurrence of an item of media content may comprise interrogating a scheduling database to determine a broadcast time for the media item. Alternatively, identifying the occurrence of an item of media content may comprise detecting the item of media content in a live media stream.

The method may comprise identifying trigger terms in the item of media content. In this case, identifying consumer activity may comprise identifying Internet searches conducted using the trigger terms, or using keywords associated with the trigger terms.

The media content may be, or have been broadcast in a television channel or video on demand stream. In this case, the trigger term may be detected in caption information and/or subtitle information associated with the television programme being broadcast or the video on demand stream. In one implementation, subtitle information is embedded within image data within the television broadcast signal or the video on demand stream, and the subtitle information is converted into text using optical character recognition.

The method may comprise generating and storing correlation data indicative of the impact of the item of media content on Internet searches conducted using the trigger words, or using keywords associated with the trigger terms. Information to be displayed as part of an Internet search function, social media site or mobile application may be selected in dependence on the correlation information.

Identifying the strategy may comprise specifying an Internet search function to be influenced based on trigger terms and/or key words determined to result in a level of consumer search activity which is greater than a threshold value.

The Internet activity may be an Internet search carried out by a consumer, or more particularly, the number of Internet searches carried out by all consumers during a particular time window. Alternatively, the Internet activity may be consumer access to website, social media or mobile application content received over the Internet, again during a particular time window.

The selected information may be predetermined or dynamically generated, and may comprise a link to a web page, and/or images and text.

According to another aspect of the invention, there is provided an apparatus for influencing access to website, social media or application content, comprising: a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content; an Internet activity identifier, for identifying consumer Internet activity at the time of the occurrence of the item of media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to influence an Internet search function, social media site or application to display selected information; a subsequent broadcast identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified strategy at the point the item occurs during the time of the identified subsequent broadcast.

According to another aspect of the invention, there is provided an apparatus for influencing website, social media or application content, comprising: a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content; an Internet activity identifier, for identifying consumer Internet activity at the time of the occurrence of the item of media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or application; a subsequent broadcaster identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified modification to the content of the web site, web page, social media site or application at the point the item occurs during the time of the identified subsequent broadcast.

According to an aspect of the invention there is provided a method of monitoring the influence of media content on website traffic, comprising: receiving a trigger term and associating the trigger term with a website; monitoring traffic to the website; detecting the trigger term in a media source; and identifying and storing correlation data indicative of an impact of content in the media source on the level of traffic to the website, the content being indicated by the trigger term.

The trigger term may be detected in plural media sources.

The received trigger terms may be stored in a database.

The incidence of trigger terms within media sources may be stored in a database.

The correlation data may store a correlation between the occurrence of the trigger term in the media source with a level of traffic to the website at or during a predetermined period of time after the occurrence of the trigger term.

The media source may be a television channel or on demand video stream and the trigger term may be detected in caption information and/or subtitle information associated with the television programme being broadcast. Subtitle information may be embedded within image data within the television broadcast signal or video on demand signal, and in this case the subtitle information may be converted into text using optical character recognition.

Preferably, the correlation data is displayed.

According to another aspect of the invention, there is provided an apparatus for monitoring the influence of media content on website traffic, comprising: a first database for receiving a trigger term and associating the trigger term with a website; a traffic monitor, for monitoring traffic to the website; a trigger term detector, for detecting the trigger term in a media source; and a match detector, for identifying and storing into a second database correlation data indicative of content in the media source on the level of traffic to the website, the content being indicated by the trigger term.

According to an aspect of the invention there is provided a method of influencing access to website, social media or application content, comprising: detecting an occurrence of a trigger term in a real time media source; and influencing an Internet search function, social media site or application to display predetermined information in response to the detection.

The predetermined information may comprise a link to a web page. Further, the predetermined information may in some cases comprises images and/or text.

The web page may be modified in response to the detection.

The method may comprise maintaining a database of stored trigger terms associated with websites, matching the stored trigger terms against text extracted from the real time media source, and influencing the Internet search function or social media site to display a link to a web page of the associated website.

The Internet search function or social media site may be influenced by the occurrence of the trigger term for only a predetermined period of time.

The Internet search function or social media site may be influenced following a predetermined delay from the detection of the trigger term.

One or more keywords may be associated with the trigger term, and the Internet search function may be influenced by triggering an Adwords campaign based on the keywords associated with the trigger term.

Where the media source is a television channel or video on demand stream the trigger term may be detected in caption information and/or subtitle information associated with the television programme being broadcast or the video on demand stream. Where subtitle information is embedded within image data within the television broadcast signal or the video on demand stream, the subtitle information may be converted into text using optical character recognition.

According to another aspect of the invention there is provided an apparatus for influencing access to website, social media or application content, comprising: a trigger term detector for detecting an occurrence of a trigger term in a real time media source; and a controller for influencing an Internet search function, social media site or application to display predetermined information in response to the detection.

According to an aspect of the invention there is provided a method of influencing website content, comprising: detecting an occurrence of a trigger term in a real time media source; and modifying a web site or page in response to the detection.

The web site or page may be modified to display content related to the trigger term.

The method may comprise maintaining a database of stored trigger terms associated with web sites or pages, matching the stored trigger terms against text extracted from the real time media source, and modifying the associated web site or page.

The web site or page may be modified by the occurrence of the trigger term for only a predetermined period of time.

The web site or page may be influenced substantially immediately following the detection of the trigger term.

The web site or page may be modified following a predetermined delay from the detection of the trigger term.

Where the media source is a television channel or a video on demand stream, the trigger term may be detected in caption information and/or subtitle information associated with the television programme being broadcast or the video on demand stream.

Subtitle information may be embedded within image data within the television broadcast signal or the video on demand stream, and the subtitle information may be converted into text using optical character recognition.

According to another aspect of the invention, there is provided an apparatus for influencing website content, comprising: a trigger detector for detecting an occurrence of a trigger term in a real time media source; and a controller for modifying a web site or page in response to the detection.

To help understanding of the invention, a specific embodiment thereof will now be described by way of example and with reference to the accompanying drawings, in which:

Figure 1 schematically illustrates a system for monitoring media streams and influencing access to web pages based on that monitoring;

Figure 2 schematically illustrates a system for detecting trigger terms in television captions;

Figure 2A schematically illustrates a system for detecting trigger terms in television captions.

Figure 3 is a schematic flow diagram which illustrates how trigger terms can be used to correlate events occurring in a media stream with traffic to a website;

Figure 4 is a schematic flow diagram which illustrates how a user is provided with enhanced access to relevant web page content based on trigger words within media streams;

Figure 5 schematically illustrates a system for identifying past consumer Internet search habits at the time of broadcast of media items; and

Figure 6 is a schematic flow diagram of the operation of Figure 5.

Figure 7 Figure 7 is a schematic flow diagram illustrating how data may be correlated to generate a trigger term database for use in the system of Figures 1 to 6.

Referring to Figure 1, a system for monitoring media streams and influencing Internet search functions, web pages, social media sites or applications (which might be mobile applications (or “apps”) running on portable electronic devices such as mobile phones, or software applications running on personal computers for example) based on those media streams is shown. The system comprises a client front end 1 which a user having an association with a particular web site or page 9 is able to use to enter trigger terms (words or phrases) which he considers to be contextually relevant to his web site 9, and in particular which he thinks the occurrence of which in a media stream might cause viewers of the media stream to be driven to his web site 9. For example, if the web site 9 is a merchant site selling clothes then trigger terms relating to types, styles or makes of clothing may be of interest, since viewers seeing the items or events to which these trigger terms relate might be interested in visiting the web site 9. The user may also enter rules which govern what happens when a trigger term is detected in one or more media streams. The rules may relate to how searches, web pages or other media streams should be influenced or modified if that trigger term is detected. The trigger terms are provided to a trigger term database 2, which can be expected to store large numbers of trigger terms associated with various different client users (and/or their associated web sites), and with various different rules. It will therefore be understood that the occurrence of a particular trigger term in a media stream may invoke a number of rules in relation to different client users and their websites, if that trigger term has been registered by multiple client users. A trigger term detector 3 receives media data streams from multiple media sources 4 (four media sources are shown here - two TV channels 4a, 4b, a Facebook feed 4c and a Twitter feed 4d - but it will be appreciated that any number of TV channels, Internet feeds, radio or other real-time media sources such as news or weather feeds could be utilised) and detects when any of the trigger terms in the trigger term database 2 occur in any of media data streams. More particularly, the data from the media sources includes text which is compared against the trigger terms in the database 2. Certain types of media source, such as news or weather feeds, or social media feeds, are rich in text content, making it relatively straightforward to identify when trigger terms arise within those media sources. In other types of media source, such as television or radio, the media is less text based, making it more difficult to obtain trigger terms describing what is going on in the content. The example of television is discussed in some detail below. In the event that the trigger term detector 3 detects a trigger term in one of the media streams then this detection is recorded as a match in a match database 5. More particularly, the match database stores an indication that a detection of a particular trigger word has occurred, along with the media source originating the trigger word, and the time of the occurrence. It will be understood that the trigger term identified within the media stream is indicative of the content of the media stream.

In parallel with this, a web site monitor 8 monitors traffic to the client user’s web site 9, and records this with respect to time. Again, it will be appreciated that the web site monitor 8 may in fact monitor traffic to a large number of client user web sites, and store traffic information in association with an identification of the web site or client user. The level of traffic to the web site 9 (or in the alternative an indication of an amount of increase in traffic to the web site 9) at or shortly after (usually within a few minutes) the detection of the trigger term is stored into the match database 5 in association with the record of occurrence of the trigger term. The user is able to view, via the client front end 1, how the occurrence of particular trigger terms (and thus the occurrence of the underlying content indicated by the trigger terms) influences the traffic to his web site 9. So, for example if the trigger term was “red dress”, detected in caption information from a television programme in which a celebrity is wearing a red dress, then this might result in large number of television viewers looking up red dresses on the Internet, which might in turn result in an uplift in the number of visitors to the users web site 9. The data stored in the match database may comprise a correlation between the occurrence of the trigger term in the media source with a level of traffic to the website at or during a predetermined period of time after the occurrence of the trigger term. The user may have a number of trigger terms registered at the trigger term database 2, and matches for each of these, and associated web traffic information, are stored at the match database 5. This enables the user to determine which trigger words are the most effective. The traffic information may not simply relate to the number of site visits, but might also relate (in the case of an e-commerce site) to the number of purchases made, or the conversion rate from hits to purchases. The traffic information might also track specific segments or demographics of visitors, such as sex, location, likes and dislikes and so on, making it possible to identify from the traffic information the types of viewers which are being driven to the site by particular trigger words.

The detection of a trigger word by the trigger term detector 3 can also initiate the optimisation of an Internet search function 6 that will cause or at least influence the search function to display predetermined information. For example, the predetermined information could be information relevant to the trigger word when appropriate keywords are used in the search (note that the keywords forming the basis of the internet search and the triggers words may be different). As a result, incidence of events within a media stream which are described by trigger words may be used to effectively tailor the information, which may include advertising content, presented on a search results page in real time or near real time. This may give a user more efficient access to relevant information. Generally, the predetermined information will take the form of a link to the web site 9 (potentially in association with other information to assist a user in understanding the relevance of the web site 9 to their search) when invoked by a user performing a relevant search. In this way, media content from a media source, which may be the trigger for a viewer of that media source to perform an Internet search, can be used as a trigger to cause the web site 9 to appear prominently in an Internet search results page when the viewer performs a search (for example on Google). This could happen either immediately upon detection and processing of the trigger term, or in some cases a short time later. For example, it has been found that if a viewer sees a product which is of interest to them they will tend to search for it immediately, while if they are interested in an audio track or a video clip they may wait until the audio track or video clip has finished before performing an Internet search in relation to that item, or undertaking other consumer internet activity in relation to that item.

One possible implementation of this would be for the Internet search function 6 to trigger a Google Adwords campaign using suitable keywords, or increasing the bidding on an existing campaign. In this case, the web site 9 would be displayed on the search results page as an advertisement. In a similar manner, selected (predetermined or dynamically generated) information such as advertising creative could be displayed a social media site such as Facebook or Linkedin in response to trigger terms appearing in a media stream. In either case, the search function or social media platform may only be influenced to display the selected information for a short period after the trigger term has occurred in a media stream. After this, the search function or social media platform may revert to default content or rules.

The same principle may also be extended to content presented via mobile applications (apps), which may obtain, filter and display information obtained from the Internet. In other words, the information displayed via the app on a mobile device may be modified to include the selected information. This can be particularly useful for mobile devices, where the small screen size means that only a limited amount of information can be displayed at any one time. In fact, on mobile devices, the first page of search results may only display the top two advertisement positions, together with a few organic results.

Similarly, the detection of a trigger word by the trigger term detector 3 can also trigger a web site optimiser 7 to initiate the optimisation of the web site 9, to which a user will hopefully be directed via the search results page. In particular, the web site can be modified to relate to the context or subject of the trigger term. As a result, a user will find the content which they were interested in being readily available on the website, rather than needing to navigate deep into the website to find it. Returning to the “red dress” example, the detection of the trigger term “red dress” may cause the main page of the web site 9 to switching from displaying information about a grey jacket to displaying information about a red dress available for purchase through the web site 9. In this way, a person searching for a red dress will not only find it more easily through the web search function, but will also locate it more easily on the web site to which that person clicks through from the search results page. As with the search optimisation, the web site optimisation may occur substantially straightaway in response to the trigger terms, or might be delayed for a predetermined time. Also similarly with search optimisation, the web site may revert back to its original state after a predetermined period of time has passed and the modified content has thus become less relevant.

Advantageously, the optimisation or modification of search function and/or website may be for a short burst time period before reverting back to a baseline or original state. In certain embodiments, the predetermined time period is around 8 minutes or less. In certain embodiments, the predetermined time period is around 15 minutes or less. In other embodiments, the predetermined time period is around 1 hour or less.

Similarly, initiation of the optimisation or modification of search function and/or website may be extremely fast. In some embodiments, this may be triggered within around 2 to 10 seconds or less of the occurrence of trigger term in a media stream. In some embodiments it may be triggered in under 1 second.

Referring back to the match database, it will be appreciated that influencing the prominence of the user’s web site 9 and the immediate availability of content relevant to the trigger term at the web site 9 could be expected to increase the amount of traffic to the web site 9. As a result, the match database 5 is able to provide a good indication of how much of an enhancement to web site traffic is being achieved by way of influencing search results and web page optimisation. Accordingly, client users are able to identify which trigger words are providing maximum impact in driving viewers to their website.

In one example, at least some of the media sources are television channels. Referring to Figure 2, functional blocks for detecting trigger terms within these television channels are shown. In Figures 2, two tuners 21, 22 are provided, each for receiving a respective television channel and for decoding and outputting a sequence of image frames forming a video image. To the output of each tuner is a respective subtitle extraction block 23, 24 for extracting a portion of each image frame which corresponds to a subtitle or caption location. The output of each extraction block 23, 24 is a respective optical character recognition block 25, 26 which is able to determine the text included within the subtitle and output it to a matching unit 27, which matches the recognised text against trigger terms stored in the database 2 shown in Figure 1. The elements 21-27 correspond to the trigger term detector 3 of Figure 1. In the case of a match, the match database 5, search function 6 and web site optimiser 7 are all informed of the detection, as described above in relation to Figure 1. The Figure 2 implementation may be important for territories, such as the United Kingdom, where subtitle data is embedded directly into the image. Where the subtitle data is available as text within the data stream (such as for video signals in the US) then a simplified trigger term detector can be used, omitting the requirement to perform optical character recognition.

In certain embodiments, additional processing of text output to the matching unit 27 may be required prior to matching text against trigger terms. Figure 2A illustrates in more detail processing of subtitle data at the trigger term detector 3.

Frame data is extracted from a subtitle feed. Particularly in UK television broadcasts, subtitle data may be received and extracted as unordered frames and may be captured as images. Each frame is assigned a sequence number, which may be used to reorder frames into correct sequence following OCR. Where frames are extracted as images, each image is assigned a sequence number. Frames are also grouped by TV channel.

Image data with associated sequence information is input to the optical character recognition (OCR) block 25, 26 and converted to text, such that each frame of text output from the OCR block 25, 26 is associated with one or more sequence numbers.

In general, a subtitle frame corresponds to text or images displayed in association with a television broadcast at a particular point during the broadcast (and at a particular location). Each frame is replaced with a subsequent frame following a frame refresh. Typically, on a pre-recorded programme, a single frame comprises a line of text, up to around 15 words. As each unique image is refreshed on screen during a broadcast, the entire frame is replaced.

As illustrated in Figure 2A, the trigger term detector 3 comprises a re-ordering block 28. Text with associated sequence number output from the OCR block 25, 26 is input to the reordering block 28, which provides correct word order using sequence numbers, for proper context interpretation and further processing.

In the case of a missing frame or word in the sequence, the re-ordering block 28 waits for the missing frame for a maximum of two seconds, following which delay, if the missing frame is not received, the re-ordered frames are processed and output from the re-ordering block 28 without the missing frame. A missing or delayed frame received by the re-ordering block after the two second delay will subsequently be processed by the system and output from the re-ordering block, such that no potentially important matches of subtitle text with trigger terms are missed. In some broadcasts (particularly live broadcasts) frames are built up as words are spoken, which may lead to duplicate words being received in a feed. For example:

Frame 1: “The”, is refreshed and replaced with Frame 2: “The caf’, which is subsequently replaced with Frame 3: “The cat sat”.

Reordered frames of text are output from the re-ordering block 28 to a debouncing filter 29, which removes duplicate words. The debouncing filter 29 receives two copies of the stream of reordered frames from the same broadcast, which are offset by one frame. Each frame is compared with the previous and next frames such that duplicate words may be removed to provide a stream of single words of text, in correct order, output to a mutable array 30, which is configured to group around 5 to 30 words (and preferably around 15 words) at any one time. Each word from the array 30 is successively matched against a trigger term trie 31.

Trigger terms are arranged in the trie 31 as a string of nodes, each node representing a letter of one or more trigger terms. The structure allows storage and fast matching of a very large number of trigger terms.

It will be appreciated that in certain embodiments, debouncing (removal of duplicates) and collecting a predetermined number of words may be performed using the same array or data structure.

The trigger term trie 31 is constructed using an algorithm applied to trigger terms from the database 2 and is a dynamic structure. This means that the trie structure may be modified - for example, if a new trigger term is entered, new nodes corresponding to letters and failure cases (re-routing failures) are dynamically added to the existing trie structure. A lock function may limit number or timings of modifications to the trie. This allows new trigger terms to be added via the client front end 1, without significant impact or delay on the matching and triggering functionality of the system. In some cases, addition of new nodes and failure cases to the trie 30 may take around 5-10 seconds, which is significantly faster than rebuilding a new trie.

In certain embodiments, the trie 31 is constructed using all trigger terms in the database 2, which may be a very large number corresponding to multiple client trigger words and multiple campaigns. However, it will be understood that alternative groupings or selections of trigger terms may be utilised to provide single or multiple trie structures. If the trigger term detector 3 identifies a match of subtitle text with any trigger term in the trie 31, a match event is broadcast to further components of the system to influence search function and/or website optimisation. The match database 5 stores data relating to the match event, together with context information.

The context information may include the contents of the mutable array 30 at the time of the match. For example, the system may record a predetermined number of words surrounding a mention of a trigger word in a subtitle feed, creating a “snapshof ’ of the context in which the trigger term was mentioned in a media broadcast. A campaigns activator (not shown) listens for particular trigger term mentions to activate specific campaigns in relation to search function and/or website optimisation based on particular rules. It will be appreciated that the same trigger term mention may trigger multiple different campaigns and rules.

In certain embodiments, the trigger term detector 3 comprises a computer processor that receives media data signals comprising input values corresponding to a stream of state change events. The overall state / extended state or “state of the world” comprises a set of named fields. Each of these named fields comprises only the specific state that has changed. The trigger term detector may itself receive programme and channel metadata (such as electronic programme guide data) or it amy be linked to other processors and/or databases that process this data.

The trigger term detector may comprise a number of matching units 27, each configured to run a set of queries each time a state changes eg broadcast frames and or subtitles change. If one or more of the queries matches then a match is generated. Queries can match one any combination of fields. Therefore when a state change event eg a subtitle word arrives it must be combined with a previously received state to create a full set of fields representing the overall state, giving context. This can create a problem in a concurrent environment as the overall state must be “visible” to all of a plurality of matching units 27, such that all events must be transmitted to all signal matchers or the overall “state of the world” must be stored in a distributed cache so that an update by one process is “seen” by all. Transmitting all messages to all matching processes significantly reduces the performance gains of concurrency, and mirroring all updates to all nodes in a distributed cache can also impact hit performance or overwhelm the cache ability to update over the network.

To improve the performance of the system, which may be required to process a great number of concurrent signals and events, some of the fields are designated as persistent state and some are designated transient state. Transient fields represent substantially instantaneous events or states with a very short life. Transient fields are not saved to the cache as they will never be used to enrich later events, since they are out of date as soon as they have been processed. Examples of signals whose values may be designated as transient fields are television metadata such as subtitle text. In contrast, more slowly changing data fields such as TV programme titles or channels are normally persistent. This structure removes most of the load from the distributed cache while ensuring that all events are enriched with enough state for the signal matching units 27 to recognise every required match.

Although the enrichment phase where the state change message is converted to a full state of the world could be performed by the same process that does the matching, in a preferred embodiment it is split into a separate process in this design. This provides added flexibility and options for resilience. When a state change message arrives, all persistent state is added to it, to create a 'state of the world'.

However, some persistent state is only applicable to a subset of the incoming state change messages. These scoped persistent fields have an associated key field that must be present in the state change event. For instance, a subtitle (transient state change) should be enriched with the relevant programme title (scoped persistent field); but the program title depends on the TV channel (key field). The subtitle update message does contain the channel name, so this is used as part of the lookup for the program title field.

The signal matchers may also apply a post match filter because if a query references only a persistent field then it may be matched many times (ie every time a state change is enriched with this field it will go forward to the matchers and will match) Ideally, a query that references only persistent fields should only match if one or more of those fields changes. The post match filter discards any matches on persistent fields if none of those fields have changed state.

Referring to Figure 3, a method of monitoring the influence of media content on website traffic is shown in the form of a flow diagram. At a step A1 web traffic (for example the number of visits) to a web site is monitored. At a step A2, a client enters trigger terms they deem relevant to their website. At a step A3, media streams are monitored with respect to those trigger terms, and if at a step A4 a trigger term is detected in a media stream then that occurrence is correlated with the monitored website traffic and stored in a database at a step A5. The correlated and stored data, or analytics conducted on that data, is then displayed at a step A6.

Referring to Figure 4, a method of influencing access to website content is shown in the form of a flow diagram. At a step Bl, a media stream, in this case a television channel, is playing. At a step B2, a user is viewing the television channel.

In parallel with the step B2, the system described above, and in particular the trigger term detector 3, is monitoring the television channel for trigger words. At a step B4, a particular event occurs which both gives rise to a trigger word or phrase being detected at a step B5, and is also (in parallel) of interest to a user at a step B8. In response to the detection of the trigger term, a search function relating to the event is modified at a step B6, and a website bearing content relevant to the event is also modified. The steps B5, B6 and B7 occur very quickly, typically in less than a minute.

In a preferred embodiment, the steps B5 and B6 and/or B7 occur within less than about 10 seconds from an event occurring at B4. In some embodiments the steps occur in less than about 5 seconds from an event occurring at B4. A large proportion of this time relates to the processing of image data to convert images to text. In general, only UK TV subtitle feeds require the image processing steps. Text data input to the re-ordering block and/or debouncing filter is typically processed by the system to trigger optimisation/modification of a search function or site in less than 1 second, and in certain embodiments, less than 600ms. This provides significant advantages for reacting quickly to events occurring in a media stream. For example, an advertising campaign may need to be withdrawn or shut down, or content removed from a website in the event of a negative event occurring.

Accordingly, by the time the user initiates a web search at a step B9, the modifications have already been made, resulting in a search function carried out at a step BIO being the one modified at the step B6. Results of the modified search function are presented to the user at a step B11. Based on the results, the user may select the website at a step B12, which by this time will have been modified at the step B7, and thus when it opens the content of the website will have been tailored in view of the event, and in particular in view of the detection of the trigger term. The user is therefore able to access the website content at a step B14, without needing to navigate through the website to find the information which he was looking for based on the event witnessed on the television channel. In parallel with the above steps, the user is also using social media - for example Facebook or Twitter. These represent media streams in themselves, which are monitored by the step B3 and which represent events which might trigger the user (at the step B8) to perform a web search. Also though, the social media site experienced by the user may itself be controlled, at a step B16, in response to the detection of a trigger term, so as to place specific selected context relevant content where it can be readily found by the user.

The invention is not intended to be restricted to the details of the above described embodiment. For instance, while the above explanations are based on a search function, social media platform, web site or mobile application being influenced based on a single trigger term, the triggering could instead occur in response to a certain number of instances of a trigger word occurring within a predetermined time window, or on particular ones of the media streams, or could instead occur in response to a certain combination of different trigger words occurring within a predetermined time window or on particular ones of the media streams.

Similarly, detection of a trigger term in a predetermined time period across a combination of media streams may invoke a trigger. For example, occurrence of a trigger term in a TV feed followed by a Twitter feed may be indicative of TV content sparking interest with viewers, which may increase likelihood of increased Internet search activity.

It will be appreciated that the same item of media content, for example a film or a TV show, may be repeated a number of times on the same or a different channel, or in the same or a different country. A repeat of an item of media content is likely to have much the same influence on consumer behaviour in accessing Internet content as did the original broadcast of that media item. Accordingly, knowledge of consumer Internet activity resulting from the first (or each previous) broadcast may be used to either improve access to information which the viewer of a repeat of that media item is likely to find interesting (for example by influencing an Internet search function, or a social media site, or a mobile application which that consumer is likely to use while the repeat of the media item is ongoing to display selected information), or by influencing the information itself (for example by modifying a particular web page or web site, or social media site or mobile application content), again while the repeat of the media item is ongoing.

Referring to Figure 5, a system for identifying past consumer Internet search habits at the time of broadcast of media items is schematically illustrated. In Figure 5, a correlation processor 122 is able to identify a time of broadcast of a particular media item - which might be a particular movie or TV show - from historical scheduling information 110. An ingestion processor 115 is also able to identify trigger terms associated with media items stored in a media database 120. Techniques for identifying trigger terms associated with media items are discussed in relation to earlier figures, and may be used here. An Internet monitor 125 is able to identify consumer Internet activity at the time of broadcast of the media item. This Internet activity may be the amount of Internet search activity in relation to one of the trigger terms identified by the ingestion processor 115, or to key words associated with one of the trigger terms. Such Internet search activity may be obtained from Google Analytics for example, and may comprise an indication of the number of times a particular search word or phrase has been used within a particular time window. It will be understood that by looking at a time window associated with the identified time of broadcast of the media item, an indication of the impact of the broadcast of that media item on consumer Internet activity can be inferred. Preferably, the “background” level of consumer Internet activity in relation to that search word or phrase can also be identified, and an increase in consumer Internet activity calculated. The background level might be single number with the averaged number of searches per unit time, or might be a function of time of day. For example, if a background level of search for a particular term is 800 searches within a given 10 minute time window during the day (for example between 17:00 and 17:10), based on historical figures, then if 2400 searches are made during that time window when a particular media item was being broadcast at that time, it is reasonable to assume that the broadcast of the media item has influenced consumer interest in making that search.

In this case the impact of the media item on the Internet activity may be expressed as an uplift (200%) on the expected background level of search. An identity of the media item, the trigger words associated with that media item, and the Internet activity associated with the media item (and in particular with the trigger words) are then correlated together and stored as correlation information in a correlation database 130.

In another example, detection of an increase in consumer Internet search activity may be determined by calculating both the mean average of the Internet

Search data set, and the standard deviation, and then identifying points on the data that are 2 standard deviations above the mean. This method may also exclude periods that are 66% zero data, in order to exclude, for instance, periods during the night where the internet search activity may be substantially zero.

It will be appreciated that a similar technique could be used in real time rather than looking at historical scheduling information and Internet search activity. This is shown in Figure 5 by the live media stream 105, which is ingested by the ingestion processor 115 in this case (to extract trigger words) instead of stored media in the database 120. Internet activity in relation to the trigger terms (or related key words) is identified by the Internet monitor 125 in the manner described above. In this case, there is no need to identify the time of broadcast from scheduling information, since the time in question is the current time. Similarly to above, the identity of the media item, the Internet activity associated with that media item (and in particular with the trigger words) are then correlated together and stored as correlation information in the correlation database 130.

It will be appreciated that the resulting content of the correlation database 130 provides a resource of how particular media items influence Internet activity. A client front end 140 may be able to access a view of this information to assist them in identifying a strategy for influencing a future search, or a social media site, or a mobile application, to display selected information. In particular, the user may be able to identify (either automatically based on an algorithm, or using manual intervention), selected information which is likely to be of interest to consumers in view of the searches which they made. A user interface may be provided to enable a user to set up one or more strategies to be triggered the next time that media item is broadcast. The strategy might include upping a bid for a keyword within an advertising campaign such as Google AdWords (which will influence a search to display selected information), it might be placement of product on a website at a particular time (that is, modifying the content of a website to include content which is contextually relevant to a particular media item being broadcast), it might be launching a promotion in relation to that product, or many other actions. The selected information may (as per above) be a link to a particular web page, and may include text and or image data.

The selected information may be provided in response to a request from a computing device, such as a mobile computing device, over a network eg the Internet. For example, a mobile computing device may comprise a web browser, providing an interface through which search terms may be input or links clicked to generate a request, and web pages can be displayed and interaction with web pages is possible. The selected information may comprise files, documents, text, images, audio, video elements, graphics, search results, web pages, webpage listings, discussion threads, hyperlinks, numerical values, embedded content, advertisements, ad extensions, banners etc or any combination thereof

In addition to this specific information regarding the relationship between a particular media item (and potentially a particular time segment within a media item) and Internet activity, the correlation information also associates together particular trigger terms with particular Internet search statistics. This association represents the likely impact of media content containing a particular trigger term giving rise to particular consumer searches (or other Internet activity - as discussed below). This enables trigger terms to be evaluated as to how valuable they are likely to be as triggers for influencing access to website, social media or mobile application content, or for modifying website, social media or mobile application content. A trigger term which is associated with only a modest increase in Internet traffic is likely to be a less useful trigger term than one which is associated with a larger increase in Internet traffic. By utilising historic scheduling information and ingesting stored media it is possible to build up a detailed picture of the value of particular trigger words, which can be used to infer the likely impact of future media content (containing those trigger words) on consumer Internet activity.

Lists of popular Internet search terms are available in databases such as Google Trends. When bidding on keywords for an advertising campaign such as Google AdWords auctions, there is generally most intensive competition and cost associated with “head terms”, which are the popular and short keywords/phrases most clearly linked to high search volumes. There may be less competition for “long-tail” keywords, which usually include longer and more specific phrases. For example, a large number of furniture retailers may compete for the keyword head term “sofa” because there is a higher probability that consumer search activity for living room furniture will include that term, where there may be less competition in an auction for a long tail keyword such as “small green stripy sofa”. Such long tail keywords may be less expensive but less likely to drive high search volume to a retailer’s website. A method of determining keywords and/or trigger terms that result in a level of consumer Internet activity that is greater than a threshold value is illustrated as a flow diagram at Figure 7.

At a step Dl, a database comprising keyword data representing a large number of keywords is created. The keywords may, for example, include brand names, product names and other terms of interest to a particular client, which they may already bid on in Google AdWords campaigns. The keyword data is sorted and ranked according to historical search volume statistics and a subset of keyword of data is selected. At a step D2, the keyword data subset is used to query an Internet Search Activity Database and an algorithm is applied to detect anomalies in consumer Internet search activity in relation to the keyword data subset in the database.

In certain embodiments the system may detect anomalies in other data sources, such as social media data eg “tweets” around particular words or phrases. Anomaly data from multiple sources corresponding to the same or similar words or phrases may be correlated to provide greater accuracy.

Various approaches to detecting anomalies in time series data such as Google Trends may provide data for use in the system of the invention. The anomalies may represent marked increases or “spikes” in digital activity (eg search for a particular keyword).

These increases or “spikes” may be determined by calculating the mean average of the Internet Search data set and the standard deviation, then identifying points on the data that are a number of standard deviations above the mean (eg 2 standard deviations). The average may be calculated as a rolling average and the method may also exclude periods that include a certain percentage zero data (eg 66% zero data, in order to exclude, for instance, periods during the night where the internet search activity in relation to a particular keyword may be substantially zero). In this way any “spikes” in internet search activity in relation to the keywords are identified and recorded. The Internet Search data set may be limited to a particular time period, such as the preceding month or year.

In order to determine more accurately anomalies that are representative of true instances of increased digital activity prompted by media broadcasts, time series data points are classified as a “spike” x if x > μ + ησ where μ is the median of the time-series, σ is the interquartile range and n is an arbitrary constant.

In identifying the most relevant “spike” anomalies, data which both reach 0 and have high nonzero values in normalised data provided by eg Google Trends, is filtered out by the system. Data which both reach zero and have enough points whose second differences (discrete second derivatives) are either 0 or 1 is also filtered out.

For some keywords, there is a trend where the amount of search traffic increases as the day goes on. This suggests using a rolling median and rolling interquartile range. The system may classify a point Xj as a spike if Xi > μ^ + ησι where μ^ is the centre-aligned rolling median and is the centre-aligned rolling interquartile range. If the rolling window goes past the end it is trimmed.

In certain embodiments, the system detects anomalies in time series data by utilising left-aligned rolling statistics and right-aligned using statistics. In other words, it forms a window to the left of a point and computes the statistic for that window, and does the same for the right, then checks whether both statistics are sufficiently high.

False positives are filtered out by noting that the difference between the value at a “spike” and the value at the previous point in time is usually high. If not that, then the difference between the value at the spike and at the point two units of time previous to it is high. As such, the candidate spikes x^ are filtered down by checking if either x^ — Xi_i or x^ — Xi_2 is high enough.

The system generates a relevant anomaly event indicative of a variation in the level of consumer Internet activity in relation to a keyword within a defined time window by interrogating and filtering time series data where either: • both the minimum nonzero value is high and there is a 0 value • or in which there is a 0 value and the number of points where the second derivative is 0 or 1 is sufficiently high.

Only points where either x^ — Xi_i or Xj — Xj_2 is high enough are then considered. Finally, all remaining points x^ are detected where either: • Xi> μ+ ησ where μ is median and σ is interquartile range. • Xi> μι + kai where μ^ is rolling median and is rolling interquartile range.

At step D3, any increases or “spikes” in consumer Internet search activity detected in relation to the keyword data subset are correlated with media broadcast metadata, which may include TV subtitle data and/or electronic programme guide (EPG) data.

In particular, the time of the increase in consumer Internet search activity may be mapped to a particular point in time in a media broadcast and/or to words and/or images associated with the broadcast at that time or within a short time window. This provides an indication of likely entities, audio, images and/or text causing the spike in Internet searching around that time window.

Data including text and/or images associated with the broadcast is generated at D4 and is categorised, sorted and analysed for potential triggers that might influence an increase in Internet search activity. At step D5, the potential triggers data is used to query an Internet Search Activity Database (or other digital activity database) and an algorithm is applied to detect any increases in consumer Internet search activity in relation to the potential triggers.

In a further step at D6, increases in consumer Internet search activity in relation to the keyword data subset is correlated with increases in consumer Internet search activity in relation to the potential triggers. The correlation data is utilised to generate trigger term data in a database at step D7. The trigger term data may be further analysed by finding further variations in search activity in relation to those terms.

In another embodiment, the subset of keyword data may be used to query a database of media broadcast data, which may include TV subtitle data. The occurrence of the keywords in media broadcasts may be correlated with increases in consumer Internet search activity at the time of the occurrence of the keywords in a media broadcast. A weighting may be assigned to particular a keyword based on search activity prompted by its occurrence in a media broadcast. Similarly, long tail keywords may also be generated by analysing increases in search activity around keywords and noting words surrounding their occurrence in a media broadcast.

Words or phrases may then be combined to automatically generate strings comprising long tail keywords and/or phrases known to trigger internet search activity.

It will be appreciated that the trigger terms generated may be contextually and conceptually unrelated to the keywords. This means that the trigger term may not be descriptive of any characteristic or conceptual association with the keyword/phrase. For example, many fashion items in media content that drive internet searches are not mentioned in audio and thus subtitle content. Similarly, the trigger terms may be conceptually and contextually unrelated to the internet search activity ie the search terms input by consumers. A trigger term may be a word or phrase that is broadcast at a particular point in a media broadcast when an image displayed on screen is prompting internet search activity. As such, the trigger term may serve only as a time stamp, marking a particular point in the broadcast, even in circumstances in which the word or phrase is not descriptive of or contextually/conceptually associated in any way with the image prompting the search.

Because a trigger term known to occur at a particular point in a particular media broadcast can be correlated to an increase in search activity at that particular point in the broadcast, if the same broadcast is subsequently broadcast, the trigger term can be used to mark a position during the broadcast when internet search activity is predicted to increase. The trigger term will give an accurate indication of position within the broadcast even if the broadcast is divided into different time segments (such as via advertisement breaks of differing lengths or positions) on a subsequent broadcast.

Broadcast of a trigger term as a predictor of increased digital activity is much more accurate when given context, for example by its occurrence during a particular television programme, on a certain channel, at a particular time of day, in association with one or more other terms and/or with the sentiment of the mention. As such, the trigger term may be associated with any one of these elements and upon detection of a trigger term in a media stream, the website optimiser 7, search function 6 and/or campaigns activator may only be triggered to execute a strategy if certain conditions or rules are fulfilled eg the trigger word is broadcast during a particular TV programme.

This filtering step may also be performed with reference to a historical correlations database. For example, when a match or mention is detected, the system may only be triggered to execute a strategy if the context of the match is comparable to a historic correlation between same or similar trigger term in the same or similar context (eg during the same programme or time of day) in which historical instance the correlated level of digital (eg search or social media) activity has been determined to be above a threshold level.

At a step D8, trigger term data from the database is used to construct a trigger term trie. The trigger term trie is constructed using an algorithm applied to trigger terms from the database and is a dynamic structure. The trigger term data and trie may be utilised in a method of influencing access to website, social media or application content. As such, the trigger term database and trie at step D8 may be equivalent to database 2 and trie 31 illustrated in Figure 2A.

In certain embodiments, the correlation data also represents the likely impact of media content containing a particular trigger term giving rise to particular consumer searches or other Internet activity. Such data may be further analysed, segmented and assigned a certain weighting, which may be applied to rules for optimising and/or influencing an internet search function, website or application.

The most promising trigger words can therefore be used to trigger the influencing of access to and/or the modification of Internet-related content in the manner described above with reference to Figures 1 to 4. It will be understood that the correlation between a media item and Internet activity, or between trigger words and Internet activity, can be improved by looking at each repeat of that media item and the impact it has on web searches, rather than just relying on data from the first broadcast of that media item.

Instead of using consumer search statistics, the Internet activity may be based on detected consumer access to website, social media or mobile application content received over the Internet. This may require reporting by software installed on user devices (for example a browser plug-in), or reporting from web sites, social media platforms or mobile applications. A strategy controller 135 is able to formulate a strategy to influence an Internet search function, social media site or mobile application to display selected information, and/or a strategy to modify the content of a web site, web page, social media site or mobile application. The strategy is based on the correlation information, and may include identifying suitable content to be displayed in an Internet search, or on a social media site, or via a mobile application, or modifications to be made to a web page or web site, or to a social media output stream, or to information provided to a user through a mobile application. The selected information and/or modifications may be predetermined information and/or modifications previously associated with trigger terms. In this case, the formulation of a strategy involves (a) determining which trigger words associated with the media item have given rise to a substantial increase in Internet activity, and (b) identifying what predetermined information (for display) or modification is associated with those trigger words. More generally, the selected information may be predetermined or dynamically generated. Future or current scheduling information 145 may be made available to the strategy controller 135 in order to enable the strategy controller 135 to trigger the strategy at an appropriate time, and in particular at a time when the media item is being broadcast again.

Referring to Figure 6, a schematic flow diagram illustrating the operation of the system of Figure 5 is provided. At a step Cl, the occurrence of an item of media content is identified from historic scheduling information. At a step C2, trigger terms within or associated with (but not necessarily conceptually/contextually associated with) the item of media content are identified. This may be achieved, as described above, by obtaining the media item from a content database, and processing it to extract trigger terms as discussed above in relation to Figures 2 and 3. At a step C3, a volume of Internet searches made using the trigger terms, or made using keywords associated with those trigger terms, at the time of the broadcast, are identified. An activity metric based on the identified volume as compared with a normal (background) volume of Internet searches made using the same search terms can be evaluated. At a step C4, correlation information can be generated which associates together the activity metric with the media item and/or with a time point or window within the media item and/or with the trigger words. At a step C5, the correlation information is stored in a database. At a step C6, a strategy to influence an Internet search function, social media site or mobile application to display selected information, and/or a strategy to modify the content of a web site, web page, social media site or mobile application is determined. Identifying the strategy may involve specifying an Internet search function to be influenced based on trigger terms and/or key words determined to result in a level of consumer search activity (for example the activity metric) which is greater than a threshold value. In one example, a Google AdWords campaign (defining selected content to be displayed in response to a particular search) based on those trigger words/key words may form part of the strategy. At a step C7, a future broadcast time for the media item is identified, and then at a step C8 the determined strategy is triggered at the time of broadcast of the media item. It will be understood that the strategy may be triggered at a particular time, and for a particular period, during the broadcast, rather than being triggered at the start and persisting throughout the broadcast. The system of the invention enables triggering at the level of seconds, more particularly within around 2 to 10 seconds of the occurrence of a trigger term in a broadcast. In some embodiments triggering may occur in under 1 second. Further, the system provides short burst activation of strategy, which may default to baseline or original state after a number of minutes such as after around 15 minutes. In certain embodiments, it may be desirable to invoke the strategy or optimisation for the duration of a particular broadcast or for longer time periods. Returning to the step C7, it will be appreciated that in one approach the identification of the future broadcast may simply be achieved by monitoring media sources and detecting the occurrence of the media item (for example based on metadata received with the broadcast of the media item), and in which case the step C8 may be triggered in response to such detection.

The particular strategy triggered at C8 may be subject to evaluation. For example, clicks, impressions, conversion rates or other key performance indicators may be calculated during the time period of the execution (eg modified content or access to content). Based on a conversion calculation, a feedback loop is created, such that the strategy may be automatically and dynamically optimised.

Embodiments of the invention and techniques described in this specification can be implemented in digital electronic circuitry, computer software, firmware or hardware, including the structures equivalent to those disclosed in this specification, or in combination thereof.

The invention can be implemented as one or more computer program products, ie one or more modules of computer program instructions encoded on a computer-readable medium for execution by, or to control the operation of, data processing apparatus. The computer program product may written in any programming language. The method of the invention may be performed by one or more computer processors executing a program tangibly embodied in a machine readable storage device for execution by a computer processor, such a microprocessor. A computing device suitable for implementation of the invention may comprise any combination of any number of a processor, a memory or other storage medium readable and/or writable by the processor -including, for example, volatile and non-volatile memory and/or storage elements which may have a distributed architecture), an input device, and an output device and, or other peripherals that may be communicatively coupled via a local interface, such as one or more buses, wired or unwired connections. The processor may be any custom or commercially available processor that can process instructions for execution within the computing device.

In certain implementations, multiple processors and/or multiple buses may be used, as appropriate, along with multiple memories and types of memory including but not limited to any one or combination of volatile memory elements (e.g., random access memory (RAM, such as DRAM, SRAM, SDRAM) and nonvolatile memory elements (e.g., ROM, hard drive, tape, CDROM). Multiple computing devices may be connected, with each device providing portions of the necessary operations.

It will be appreciated that simplified exemplary embodiments of the invention are described and illustrated in the specification to aid understanding, where in fact, in application the invention may comprise a highly complex system for managing many hundreds of thousands of content items, websites and applications for numerous different client campaigns over one or more networks and processing of a large amount of disparate signal data.

One of the advantages of the invention is that it provides a system for providing optimised content or access to content to an unknown user, based on aggregate statistical data.

The network 102 may include any type of network, for example, the Internet, wi-fi network, local area network (LAN), wide area network (WAN) or wireless telephone network.

It will be understood that although in the illustrative example shown in Figure 1, a single user device is shown but in fact the system 100 may include many thousands of user devices 101, publishers 103 and ad servers 104.

User devices 101 may include desktop computers, laptop computers, tablets, mobile telephones personal digital assistants or any other devices capable of receiving content via the network 102 and presenting it to a user. Typically, a user device will comprise an application eg a web browser, providing a user interface through which web pages are displayed to the user and through which the user may interact with the web pages. When a user device sends a request to a server of a publisher for a publication (typically a web page) the content generation engine 105 dynamically selects and assembles content items 106 to be presented to the user on the user device 101 in response to the request.

The content items may be retrieved from a server 104 (such as an advertisement server) or directly from a publisher 103. The content items 106 may comprise files, documents, text, images, audio, video elements, graphics, search results, web pages, webpage listings, discussion threads, hyperlinks, numerical values, embedded content, advertisements, ad extensions, banners etc or any combination thereof

Claims

1. A method of influencing website, social media or application content or access to content, comprising: identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content substantially at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the item of live or previously broadcast media content trigger term; identifying, from said detected variation in consumer internet activity and the correlation data, a strategy to modify the content or access to the content of a web site, web page, social media site or application; triggering the identified modification substantially at the point the trigger term occurs during a subsequent media broadcast.

2. A method according to claim 1, wherein identifying the occurrence of an item of media content comprises interrogating a scheduling database to determine a broadcast time for the media item or detecting the item of media content in a live media stream.

3. A method according to claim 1 or 2, wherein the step of identifying the trigger term associated with the media content comprises interrogating a scheduling database to determine programme information.

3. A method according to claim 4, wherein the step of identifying the trigger term associated with the media content comprises extracting a list of entities from the programme information.

6. A method according to any preceding claim, wherein the step of identifying a trigger term associated with the media content comprises extracting caption information and/or subtitle information associated with a television programme broadcast or video on demand stream.

6. A method according to claim 5, wherein the step of identifying a trigger term associated with the media content comprises extracting caption information and/or subtitle information associated a television programme broadcast or video on demand stream substantially at the time of the occurrence of the variation in consumer Internet activity. 7 A method according to any preceding claim, wherein identifying the variation in the level of consumer Internet activity in relation to the predetermined keyword comprises identifying an increase in Internet searches conducted using the predetermined keyword.

8. A method according to claim 1, wherein the step of identifying the strategy comprises specifying an Internet search function to be influenced based on trigger terms and/or keywords determined to result in a level of consumer search activity which is greater than a threshold value.

9. A method according to any preceding claim, comprising identifying a correlation between the variation in consumer Internet activity in relation to the predetermined keyword and a variation in consumer Internet activity in relation to the trigger term.

10. A method according to claim 9, wherein identifying the variation in the level of consumer Internet activity in relation to the trigger term comprises identifying an increase in Internet searches conducted using the trigger term.

11. A method according to any preceding claim, wherein the keyword and/or trigger term comprises a single word or a plurality of words.

12. A method according to any preceding claim, wherein the keyword is conceptually unrelated to the trigger term.

13. A method according to any preceding claim, wherein the Internet activity is an Internet search carried out by a consumer or interaction with a social media site.

14. A method according to Claim 1, wherein the Internet activity is consumer access to website, social media or application content received over the Internet.

15. A method according to any preceding claim, wherein the step of generating and storing correlation data indicative of an association between the keyword and the trigger term comprises identifying a correlation between the occurrence of a trigger term associated with media content and an increase in Internet searches conducted using the predetermined keyword.

16. A method according to claim 15, wherein the step of generating and storing correlation data indicative of an association between the keyword and the trigger term comprises storing trigger terms in a trigger term database.

17. A method according to claim 16, comprising maintaining a database of stored trigger terms associated with websites.

18. An apparatus for influencing website, social media or application content or access to the content, comprising; an Internet activity identifier, for identifying a variation in level of consumer Internet activity in relation to a particular keyword; a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in level of consumer Internet activity and identifying a trigger term associated with the media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or application; a subsequent broadcaster identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified modification to the content of the web site, web page, social media site or application at the point the trigger term occurs during the time of the identified subsequent broadcast.

19. A non-transitory computer readable storage medium providing instructions for influencing website, social media or application content and/or access to the content; the instructions when executed by a computing device, performing a method comprising identifying a variation in a level of consumer Internet activity in relation to a predetermined keyword; identifying the occurrence of an item of live or previously broadcast media content substantially at the time of the occurrence of the variation in consumer Internet activity; identifying a trigger term associated with the media content; generating and storing correlation data indicative of an association between the keyword and the item of live or previously broadcast media content trigger term; identifying, from said detected variation in consumer internet activity and the correlation data, a strategy to modify the content or access to the content of a web site, web page, social media site or application; triggering the identified modification substantially at the point the trigger term occurs during a subsequent media broadcast.

20. A system for influencing website, social media or application content and/or access to the content comprising processing apparatus and a non-transitoiy computer readable storage medium providing instructions for influencing website, social media or application content and/or access to content; wherein the processing apparatus comprises an Internet activity identifier, for identifying a variation in level of consumer Internet activity in relation to a particular keyword; a media content identifier, for identifying the occurrence of an item of live or previously broadcast media content at the time of the occurrence of the variation in level of consumer Internet activity and identifying a trigger term associated with the media content; a strategy identifier, for identifying, from said detected consumer internet activity, a strategy to modify the content of a web site, web page, social media site or application; a subsequent broadcaster identifier, for identifying a subsequent broadcast of the item of media content; and a strategy controller, for triggering the identified modification to the content of the web site, web page, social media site or application at the point the trigger term occurs during the time of the identified subsequent broadcast.