Granted Patent B2

US 12,216,720 · App. 18/112,304 · Granted Feb 4, 2025

Inventors: Arnav Mittal (Noida, IN); Franklin Geo Francis (Bangalore, IN); Taruna Manchanda (Bangalore, IN)

Assignee: Microsoft Technology Licensing, LLC

G06F16/9535G06F16/9536

Abstract

Claims (76)

1. A computer-implemented method comprising:

detecting a post of a user being added to an online service;

determining post interest scores for the post, the post interest scores being for a plurality of interests, each interest being associated with a topic;

calculating a match score for a plurality of groups based on the post interest scores and group interest scores;

using an intent machine-learning (ML) model to calculate the group interest scores, wherein the intent ML model is trained with training data based on past content and assigned labels for corresponding intents;

determining whether to recommend a group, from the plurality of groups, for including the post of the user in a feed of the recommended group based on the match score; and

within an order or less of a hundred milliseconds of the post of the user being added, causing, by a real-time nudge system, a presentation of the recommended group based on the determined recommendation.

2. The method as recited in claim 1 , wherein calculating the match score for a group comprises:

determining group interest scores for the group for the plurality of interests; and

calculating the match score for the group as a dot product of the group interests scores and the post interest scores.

3. The method as recited in claim 1 , wherein determining whether to recommend comprises:

calculating, by the intent ML model, an intent of the post from a plurality of intents, the intent ML model taking the post as an input and generating as output one of the intents from the plurality of intents; and

determining not to generate a recommendation when the intent of the post is within a predefined subset of the plurality of intents.

4. The method as recited in claim 1 , wherein determining whether to recommend comprises:

for each group:

calculating a contribution of interests of a first level, in a hierarchy of interests, to the match score for the group; and

eliminating the group from a possible recommendation when the contribution of interests of the first level is above a predetermined threshold.

5. The method as recited in claim 1 , wherein determining whether to recommend comprises:

applying filtering criteria for the post and the plurality of groups;

if the post is a candidate for the recommendation after the filtering, selecting the group, from the groups remaining after the filtering, with a highest match score; and

recommending the group with the highest match score when the highest match score is above a predetermined threshold.

6. The method as recited in claim 1 , further comprising:

calculating group interest scores for the plurality of groups, wherein calculating the group interest scores for each group comprises:

identifying group posts that have interactions from group members;

calculating interests scores for the identified group posts; and

calculating the group interest scores based on the calculated interest scores for the identified group posts.

7. The method as recited in claim 6 , wherein calculating the group interest score further comprises:

selecting a predetermined number of the highest group interest scores and discarding the remaining group interest scores.

8. The method as recited in claim 1 , wherein causing presentation of the recommended group comprises:

presenting on a user interface a message that the post was successfully posted, a first option to add the post to the recommended group, and a second option to not add the post to the recommended group.

9. The method as recited in claim 1 , wherein the plurality of groups includes groups where the user is a member.

10. The method as recited in claim 1 , wherein the plurality of interests is part of an interest ontology with interests arranged in a hierarchical order, the plurality of interest comprising a subset of first-level interests and some interests having other sub-interests in a next level.

11. A system comprising:

a memory comprising instructions; and

one or more computer processors, wherein the instructions, when executed by the one or more computer processors, cause the system to perform operations comprising:

detecting a post of a user being added to an online service;

determining post interest scores for the post, the post interest scores being for a plurality of interests, each interest being associated with a topic;

calculating a match score for a plurality of groups based on the post interest scores and group interest scores;

using an intent machine-learning (ML) model to calculate the group interest scores, wherein the intent ML model is trained with training data based on past content and assigned labels for corresponding intents;

determining whether to recommend a group, from the plurality of groups, for including the post of the user in a feed of the recommended group based on the match score; and

within an order or less of a hundred milliseconds of the post of the user being added, causing, by a real-time nudge system, a presentation of the recommended group based on the determined recommendation.

12. The system as recited in claim 11 , wherein calculating the match score for a group comprises:

determining group interest scores for the group for the plurality of interests; and

calculating the match score for the group as a dot product of the group interests scores and the post interest scores.

13. The system as recited in claim 11 , wherein determining whether to recommend comprises:

calculating, by the intent ML model, an intent of the post from a plurality of intents, the intent ML model taking the post as an input and generating as output one of the intents from the plurality of intents; and

determining not to generate a recommendation when the intent of the post is within a predefined subset of the plurality of intents.

14. The system as recited in claim 11 , wherein determining whether to recommend comprises:

for each group:

calculating a contribution of interests of a first level, in a hierarchy of interests, to the match score for the group; and

eliminating the group from a possible recommendation when the contribution of interests of the first level is above a predetermined threshold.

15. The system as recited in claim 11 , wherein determining whether to recommend comprises:

applying filtering criteria for the post and the plurality of groups;

if the post is a candidate for the recommendation after the filtering, selecting the group, from the groups remaining after the filtering, with a highest match score; and

recommending the group with the highest match score when the highest match score is above a predetermined threshold.

16. A tangible machine-readable storage medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:

detecting a post of a user being added to an online service;

determining post interest scores for the post, the post interest scores being for a plurality of interests, each interest being associated with a topic;

calculating a match score for a plurality of groups based on the post interest scores and group interest scores;

using an intent machine-learning (ML) model to calculate the group interest scores, wherein the intent ML model is trained with training data based on past content and assigned labels for corresponding intents;

determining whether to recommend a group, from the plurality of groups, for including the post of the user in a feed of the recommended group based on the match score; and

within an order or less of a hundred milliseconds of the post of the user being added, causing, by a real-time nudge system, a presentation of the recommended group based on the determined recommendation.

17. The tangible machine-readable storage medium as recited in claim 16 , wherein calculating the match score for the group comprises:

determining group interest scores for the group for the plurality of interests; and

calculating the match score for the group as a dot product of the group interests scores and the post interest scores.

18. The tangible machine-readable storage medium as recited in claim 16 , wherein determining whether to recommend comprises:

calculating, by the intent ML model, an intent of the post from a plurality of intents, the intent ML model taking the post as an input and generating as output one of the intents from the plurality of intents; and

determining not to generate a recommendation when the intent of the post is within a predefined subset of the plurality of intents.

19. The tangible machine-readable storage medium as recited in claim 16 , wherein determining whether to recommend comprises:

for each group:

calculating a contribution of interests of a first level, in a hierarchy of interests, to the match score for the group; and

eliminating the group from a possible recommendation when the contribution of interests of the first level is above a predetermined threshold.

20. The tangible machine-readable storage medium as recited in claim 16 , wherein determining whether to recommend comprises:

applying filtering criteria for the post and the plurality of groups;

if the post is a candidate for the recommendation after the filtering, selecting the group, from the groups remaining after the filtering, with a highest match score; and

recommending the group with the highest match score when the highest match score is above a predetermined threshold.

Assignments (2)

Continuity (1)

References Cited (14)