Machine learning techniques for detecting surges in content consumption

{
  "publication_number": "US2023232052A1",
  "country": "US",
  "kind": "A1",
  "title": "Machine learning techniques for detecting surges in content consumption",
  "abstract": "The present disclosure describes a content consumption monitor (CCM) that determines surges in content consumption based on changes in content consumptions scores. The CCM determines the content consumptions scores for domains and/or organizations (orgs) based on session events generated by different devices/users from the org and/or domain, a number of events generated by the org/domain, content and/or user interactions with the content indicated by the events, relevancy scores of the content to one or more topics, and/or other criteria. The CCM detects surges in consumption or interest in a topic for the domain/org when the consumption score reaches a threshold and/or within a period of time. The CCM may adjust the consumption score based on the changes in the relevancy, number of events and/or the number of users over different time periods. Other embodiments may be described and/or claimed.",
  "claims": [
    "1. One or more non-transitory computer readable media (NTCRM) comprising instructions for machine learning (ML), wherein execution of the instructions by one or more processors of a computing device is to cause the computing device to: identify one or more features from training data comprising a set of information objects (InObs) with known classifications, each InOb of the set of InObs comprising one or more nodes, the one or more features including structural semantics for respective InObs of the set of InObs, the structural semantics comprising a data structure representative of relationships between the one or more nodes of the respective InObs; train an ML model to identify classifications of InObs not among the set of InObs based on the features identified from the training data and the known classifications of the set of InObs; identify features from an unclassified InOb with an unknown classification, the identified features of the unclassified InOb including a set of nodes of the unclassified InOb; and apply the identified features of the unclassified InOb to the trained ML model to predict a classification for the unclassified InOb based on structural semantics of the unclassified InOb, the structural semantics of the unclassified InOb being based on relationships among nodes of the set of nodes. 2. The one or more NTCRM of claim 1, wherein execution of the instructions is to cause the computing device to: generate a first set of vectors representing the features of the set of InObs; use the first set of vectors and known classifications of the set of InObs to train the ML model; generate a second set of vectors representing the features of the unclassified InOb; and apply the second set of vectors to the trained ML model to classify the unclassified InOb. 3. The one or more NTCRM of claim 1, wherein the structural semantics of the respective InObs includes relationships between nodes making individual InObs and relationships between nodes of different InObs. 4. The one or more NTCRM of claim 3, wherein execution of the instructions is to cause the computing device to: analyze the InObs of the unclassified InOb to identify links between the InObs on the InOb and links with other InObs on the same InOb and links with InObs on other InObs; and determine the structural semantics of the unclassified InOb based on the identified links. 5. The one or more NTCRM of claim 1, wherein the one or more features comprise content semantics of the one or more nodes of the set of InObs. 6. The one or more NTCRM of claim 5, wherein execution of the instructions is to cause the computing device to: analyze the InObs of the unclassified InOb to identify content types and topics in the InObs; and identify the content semantics of the unclassified InOb based on the identified content types and topics in the InObs of the unclassified InOb. 7. The one or more NTCRM of claim 1, wherein the one or more features further comprise content interaction behavior features with InObs in the one or more nodes of the set of InObs. 8. The one or more NTCRM of claim 7, wherein execution of the instructions is to cause the computing device to: identify user interaction events generated by the one or more nodes based on interactions with the one or more nodes of the set of InObs; determine user interaction types based on the user interaction events; and identify the content interaction behavior features based on the user interaction types of the set of InObs. 9. The one or more NTCRM of claim 1, wherein the one or more features further comprise types of users accessing the one or more nodes of the set of InObs, the types of users including device types used for accessing the one or more nodes. 10. The one or more NTCRM of claim 9, wherein execution of the instructions is to cause the computing device to: identify network session events generated by the one or more nodes based on accesses of the one or more nodes the InObs; determine user data from the network session events; and identify the types of users accessing the InObs based on the determined user data. 11. An apparatus, comprising: memory circuitry to store instructions; and processor circuitry communicatively coupled to the memory circuitry, wherein the processor circuitry is to execute the instructions to: identify, using a trained machine learning (ML) model, one or more structural features of an information object (InOb), the trained ML model being trained on a training data set including a set of InObs, each InOb of the set of InObs comprising one or more nodes, and the trained ML model includes a data object indicating structural features of respective InObs of the set of InObs, the structural features are relationships between the one or more nodes of the respective InObs, and the data object is a representation of the relationships; and predict a classification for the InOb based on the identified one or more structural features of the InOb. 12. The apparatus of claim 11, wherein the processor circuitry is to execute the instructions to: identify user interaction events generated by the InOb or users that interact with the InOb; determine user interaction types based on the user interaction events; identify one or more content interaction behavior features for the InOb based on the determined user interaction types, the one or more content interaction behavior features being patterns of user interaction with content of the InOb. 13. The apparatus of claim 12, wherein the processor circuitry is to execute the instructions to: generate a structural feature vector comprising the one or more structural features of the InOb; generate a content interaction behavior feature vector comprising the one or more content interaction behavior features of the InOb; and feed the structural feature vector and the content interaction behavior feature vector into the ML model to predict the classification for the InOb. 14. The apparatus of claim 13, wherein the user interaction events indicate an event type and an engagement metric, and each content interaction behavior feature in the content interaction behavior feature vector represents a percentage or average value of the engagement metric for an associated event type for a time period. 15. The apparatus of claim 13, wherein the one or more content interaction behavior features include one or more of a time of day, day of week, date, total amount of content consumed by respective users, percentages of different device types used for accessing the InOb, duration of time users spend on individual InObs of the InOb, total engagement the respective users have on the individual InObs, a number of distinct user profiles accessing the individual InObs versus a total number of user interaction events for the individual InObs, a dwell time, a scroll depth, a scroll velocity, and variance in content consumption over time. 16. The apparatus of claim 13, wherein, to generate the structural feature vector, the processor circuitry is to execute the instructions to: generate respective structural feature vectors for each individual InOb of the InOb; and average the respective structural feature vectors for each individual InOb to obtain the structural feature vector for the InOb. 17. The apparatus of claim 13, wherein, to generate the content interaction behavior feature vector, the processor circuitry is to execute the instructions to: generate respective content interaction behavior feature vectors for each individual InOb of the InOb; and average the respective content interaction behavior feature vectors for each individual InOb to obtain the content interaction behavior feature vector for the InOb. 18. The apparatus of claim 12, wherein the processor circuitry is to execute the instructions to: generate the one or more content interaction behavior features for the InOb based on types of businesses accessing InObs of the InOb. 19. The apparatus of claim 11, wherein the processor circuitry is to execute the instructions to: determine the one or more structural features of the InOb based on links between InObs of the InOb and links to other InObs of other InObs from the InObs of the InOb. 20. The apparatus of claim 19, wherein the processor circuitry is to execute the instructions to: analyze the InObs of the InOb to identify the links between the InObs of the InOb and the links to the other InObs."
  ],
  "description_excerpt": "Embodiments described herein generally relate to machine learning (ML) and artificial intelligence (AI), and in particular, ML/AI techniques for associating network addresses with locations from which content and/or information objects is/are accessed.\n\nUsers receive a random variety of different information from a random variety of different businesses. For example, users may constantly receive promotional announcements, advertisements, information notices, event notifications, and/or the like. Users request some of this information. For example, a user may register on a company website to receive sales or information announcements. However, much of the information is of little or no interest to the user. For example, the user may receive emails announcing every upcoming seminar, regardless of the subject matter.\n\nThe user may also receive unsolicited information. For example, a user may register on a website to download a white paper on a particular subject. A lead service then may sell the email address to companies that send the user unsolicited advertisements. Users end up ignoring most or all of these emails since most of the information has no relevance or interest. Alternatively, the user directs all of these emails into a junk email folder.\n\nFIG. 1 depicts an example content consumption monitor (CCM). FIG. 2 depicts an example of the CCM in more detail. FIG. 3 depicts an example operation of a CCM tag. FIG. 4 depicts example events processed by the CCM. FIG. 5 depicts an example user intent vector. FIG. 6 depicts an example process for segmenting users. FIG.",
  "cpc": [
    "H04N 21/231",
    "G06Q 30/0201",
    "G06Q 30/0242",
    "H04L 67/1044",
    "H04L 67/14",
    "H04L 67/535",
    "H04N 21/251"
  ],
  "ipc": [
    "H04L 67/104",
    "H04L 67/14",
    "H04N 21/231",
    "H04N 21/25"
  ],
  "assignees": [
    "Bombora Inc"
  ],
  "inventors": [
    "Oleg Valentin KHAVRONIN",
    "Benny Lin",
    "Anthony LIVHITS",
    "Erik Gregory MATLICK",
    "Christian Michael BURTON",
    "Robert James Armstrong",
    "Nicholaus E. HALECKY"
  ],
  "filing_date": "2023-02-13",
  "publication_date": "2023-07-20",
  "priority_date": "2014-09-26",
  "application_number": "US-202318168440-A",
  "family_id": "85222640",
  "cited_by_count": 16,
  "citations": [
    "US20110258049A1",
    "US20090092299A1",
    "US20130132468A1",
    "US20130166485A1",
    "US20140188830A1"
  ]
}