Real time information feed processing

A real time information feed processing method implemented by a data processing system comprising an information interface, a data mining engine, a recommendation engine, and a subscriber interface, the method comprising the steps of:

• the interface receiving a real time information feed;
• the data mining engine retrieving data concerning a subscriber;
• the recommendation engine automatically modifying the real time information feed according to the mined data;
• and
• the subscriber interface outputting the modified real time information for delivery to the subscriber
• wherein the data mining engine performs the steps of given a user, u, and a set of articles, R, of a first real time information feed and a set of micro-blogging service messages, T, separately indexing the feed and the messages to produce at least first and second indexes, and extracting resulting index terms from said indexes as the basis to produce first and second term vectors, MR and MT, respectively;
• the recommendation engine identifies a set of words or terms, t, that co-occur in MT and MR;
• as words or terms that are present in most recent micro-blogging service messages and the most recent real time information feed items and uses each word or term, ti as a query against the first index to retrieve a set of information feed items A that contain t along with their associated frequency score, wherein each co-occurring term, ti is associated with a set of items A1, . . . An, which contain t, and the frequency score for t in each of A1, . . . An is used to produce a matrix;
• and
• the recommendation engine calculates an overall score for each article by computing the sum of the frequency scores across all of the terms associated with that item wherein items which contain many message terms with higher frequency scores are preferred to items which contain fewer message terms with lower frequency scores;
• and
• wherein the recommendation engine selects the top k articles with the highest scores, and each time the interface gathers an individual feed item from a source, the item is copied into both a user's individual item pool and a community item pool;
• and wherein each information item has a differing relevance score in either pool, and as their frequency score changes based on other content in a local directory, results-lists are generated, and a recency-based list is gathered by collecting most recent information feed items.

A non-transitory computer readable medium comprising computer software embodied therein and being adapted to perform the following method steps when executing on a digital processor:

• receiving a real time information feed;
• retrieving data concerning a subscriber;
• automatically modifying the real time information feed according to the mined data;
• and
• outputting the modified real time information for delivery to the subscriber wherein the data mining engine is adapted to perform the steps of given a user, u, and a set of articles, R, of a first real time information feed and a set of micro-blogging service messages, T, separately index the feed and the messages to produce at least first and second indexes, and to extract resulting index terms from said indexes as the basis to produce first and second term vectors, MR and MT, respectively;
• the recommendation engine is adapted to identify a set of words or terms, t, that co-occur in MT and MR;
• as words or terms that are present in most recent micro-blogging service messages and the most recent real time information feed items and to use each word or term, ti as a query against the first index to retrieve a set of information feed items A that contain t along with their associated frequency score, wherein each co-occurring term, ti is associated with a set of items A1, . . . An, which contain t, and the frequency score for t in each of A1, . . . A1 is used to produce a matrix;
• and
• the recommendation engine is adapted to calculate an overall score for each article by computing the sum of the frequency scores across all of the terms associated with that item wherein items which contain many message terms with higher frequency scores are preferred to items which contain fewer message terms with lower frequency scores;
• and
• wherein the recommendation engine is adapted to select the top k articles with the highest scores, and each time the interface gathers an individual feed item from a source, the item is copied into both a user's individual item pool and a community item pool;
• and wherein each information item has a differing relevance score in either pool, and as their frequency score changes based on other content in a local directory, results-lists are generated, and a recency-based list is gathered by collecting most recent information feed items.

A real time information feed system comprising a processor that comprises:

• an interface adapted to receive a real time information feed;
• a data mining engine adapted to retrieve data concerning a subscriber;
• a recommendation engine adapted to automatically modify the real time information feed according to the mined data;
• and
• a subscriber interface adapted to output the modified real time information for delivery to the subscriber
• wherein the data mining engine is adapted to perform the steps of given a user, u, and a set of articles, R, of a first real time information feed and a set of micro-blogging service messages, T, separately index the feed and the messages to produce at least first and second indexes, and to extract resulting index terms from said indexes as the basis to produce first and second term vectors, MR and MT, respectively;
• the recommendation engine is adapted to identify a set of words or terms, t, that co-occur in MT and MR;
• as words or terms that are present in most recent micro-blogging service messages and the most recent real time information feed items and to use each word or term, ti, as a query against the first index to retrieve a set of information feed items A that contain t along with their associated frequency score, wherein each co-occurring term, ti is associated with a set of items A1, . . . An, which contain t, and the frequency score for t in each of A1, . . . An is used to produce a matrix;
• and
• wherein the recommendation engine is adapted to select the top k articles with the highest scores, and each time the interface gathers an individual feed item from a source, the item is copied into both a user's individual item pool and a community item pool;
• and wherein each information item has a differing relevance score in either pool, and as their frequency score changes based on other content in a local directory, results-lists are generated, and a recency-based list is gathered by collecting most recent information feed items;
• and wherein the recommendation engine is adapted to take a first item from each pool, to collect said items into a list, to randomize them, and to insert them into a master result list.