Personalized user feed based on monitored activities

1. A system, comprising:

a processor configured to:

determine a feature vector associated with a candidate web document;

filter a plurality of other feature vectors of a feature space associated with the feature vector, the plurality of other feature vectors corresponding to a plurality of other web documents;

determine a density value associated with the candidate web document using a subset of the plurality of other feature vectors remaining in the filtered feature space, wherein the density value is based on one or more user interactions with respect to a subset of the plurality of other web documents corresponding to the subset of the plurality of other feature vectors;

rank the candidate web document with respect to the plurality of other web documents based on the determined density value; and

provide in a content feed the candidate web document based on the ranking; and

a memory coupled with the processor, wherein the memory is configured to provide the processor with instructions.

2. The system of claim 1 , wherein the processor is further configured to receive the candidate web document.

3. The system of claim 1 , wherein the feature space is filtered based on a cosine similarity between the feature vector associated with the candidate web document and the plurality of other feature vectors associated with the plurality of other web documents.

4. The system of claim 3 , wherein at least one of the plurality of other web documents is filtered from the feature space in an event that the cosine similarity between the feature vector associated with the candidate web document and a corresponding one of the plurality of feature vectors associated with the at least one of the plurality of other web documents is greater than a cosine similarity threshold.

5. The system of claim 1 , wherein the processor is further configured to determine a set of nearest neighbors comprising at least one of the subset of the plurality of other web documents based on the filtered feature space.

6. The system of claim 5 , wherein the density value associated with the candidate web document is based on one or more interactions with respect to the set of nearest neighbors.

7. The system of claim 6 , wherein the one or more interactions include one or more positive interactions.

8. The system of claim 7 , wherein the one or more positive interactions include at least one of reading the at least one of the subset of the plurality of other web documents comprising the set of nearest neighbors, commenting on the at least one of the subset of the plurality of other web documents comprising the set of nearest neighbors, sharing the at least one of the subset of the plurality of other web documents comprising the set of nearest neighbors, or providing reaction feedback with at least one of the subset of the plurality of other web documents comprising the set of nearest neighbors.

9. The system of claim 6 , wherein the one or more interactions include one or more negative interactions.

10. The system of claim 9 , wherein the one or more negative interactions include at least one of providing an indication of not being interested in the at least one of the plurality of other web documents comprising the set of nearest neighbors or providing an indication that the at least one of the plurality of other web documents comprising the set of nearest neighbors is off topic.

11. The system of claim 1 , wherein a rank of the candidate web document is boosted based on a probability associated with the candidate web document, wherein the probability indicates a likelihood that the candidate web document will be of interest to a particular user.

12. The system of claim 1 , wherein the density value is one of one or more signals used to determine whether to provide the candidate web document in the content feed.

13. The system of claim 12 , wherein the one or more signals include at least one of a trending signal, a freshness signal, and a relevance signal.

14. The system of claim 12 , wherein the processor is further configured to:

receive feedback associated with content provided in the content feed; and

update the determined density value associated with the candidate web document based on the received feedback.

15. The system of claim 14 , wherein the determined density value associated with the candidate web document is updated in the event a threshold time has passed.

16. A method, comprising:

determining a feature vector associated with a candidate web document;

filtering a plurality of other feature vectors of a feature space associated with the feature vector, the plurality of other feature vectors corresponding to a plurality of other web documents;

determining a density value associated with the candidate web document using a subset of the plurality of other feature vectors remaining in the filtered feature space, wherein the density value is based on one or more user interactions with respect to a subset of the plurality of other web documents corresponding to the subset of the plurality of other feature vectors;

ranking the candidate web document with respect to the plurality of other web documents based on the determined density value; and

providing in a content feed the candidate web document based on the ranking.

17. The method of claim 16 , further comprising determining a set of nearest neighbors comprising a plurality of web documents based on the filtered feature space, wherein the density value associated with the candidate web document is based on one or more interactions with respect to the set of nearest neighbors.

18. The method of claim 17 , wherein the one or more interactions include one or more positive interactions.

19. The method of claim 17 , wherein the one or more interactions include one or more negative interactions.

20. A non-transitory computer readable storage medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

determining a feature vector associated with a candidate web document;

filtering a plurality of other feature vectors of a feature space associated with the feature vector, the plurality of other feature vectors corresponding to a plurality of other web documents;

determining a density value associated with the candidate web document using a subset of the plurality of other feature vectors remaining in the filtered feature space, wherein the density value is based on one or more user interactions with respect to a subset of the plurality of other web documents corresponding to the subset of the plurality of other feature vectors;

ranking the candidate web document with respect to the plurality of other web documents based on the determined density value; and

providing in a content feed the candidate web document based on the ranking.