Facilitating content entity annotation while satisfying joint performance conditions
View Patent ↗Facilitation of content entity annotation while maintaining joint quality, coverage and/or completeness performance conditions is provided. In one example, a system includes an aggregation component that aggregates signals indicative of initial entities for content and initial scores associated with the initial entities generated by one or more content annotation sources; and a mapping component that maps the initial scores to calibrated scores within a defined range. The system also includes a linear aggregation component that: applies selected weights to the calibrated scores, wherein the selected weights are based on joint performance conditions; and combines the weighted, calibrated scores based on a selected linear aggregation model of a plurality of linear aggregation models to generate a final score. The system also includes an annotation component that determines whether to annotate the content with one of the initial entities based on a comparison of the final score and a defined threshold value.
1. A non-transitory computer-readable storage medium comprising computer-readable instructions that, in response to execution, cause a computing system to perform operations, comprising:
receiving information indicative of initial entities for content and initial scores associated with the initial entities from one or more content annotation sources;
aggregating the information indicative of initial entities for content and initial scores associated with the initial entities received from the one or more content annotation sources;
defining a range of values;
mapping the initial scores to respective values within the defined range, the mapping producing calibrated scores within the defined range;
selecting weights based on joint performance conditions;
applying the selected weights to the calibrated scores within the defined range to produce weighted, calibrated scores;
selecting a linear aggregation model from among a plurality of linear aggregation models;
combining the weighted, calibrated scores based on the selected linear aggregation model to generate a final score; and
determining whether to annotate the content with at least one of the initial entities based on a comparison of the final score and a defined threshold value.
2. The non-transitory computer-readable storage medium of claim 1 , wherein the operations further comprise comparing the final score to the defined threshold value.
3. The non-transitory computer-readable storage medium of claim 1 , wherein the mapping further comprises mapping the initial scores to respective values within the defined range by employing isotonic regression.
4. The non-transitory computer-readable storage medium of claim 1 , wherein an initial entity of the initial entities is associated with a property based on a calibrated score of the calibrated scores, and wherein the property is indicative of a level of relevance of the initial entity to the content.
5. The non-transitory computer-readable storage medium of claim 4 , wherein the property is indicative of at least one of the initial entity being central to the content, the initial entity being relevant to the content or the initial entity being off-topic relative to the content.
6. The non-transitory computer-readable storage medium of claim 1 , wherein the selected linear aggregation model is based on probabilities associated with the initial entities.
7. The non-transitory computer-readable storage medium of claim 1 , wherein the selected linear aggregation model is based on a logarithm of one or more probabilities associated with the initial entities.
8. The non-transitory computer-readable storage medium of claim 7 , wherein at least one of the selected weights is a non-unit weight.
9. The non-transitory computer-readable storage medium of claim 1 , wherein the operations further comprise:
receiving information indicative of the joint performance conditions.
10. The non-transitory computer-readable storage medium of claim 1 , wherein the joint performance conditions comprise at least two of coverage associated with one or more of the initial entities, completeness associated with one or more of the initial entities or quality associated with one or more of the initial entities.
11. The non-transitory computer-readable storage medium of claim 1 , wherein the operations further comprise:
generating the selected weights.
12. The non-transitory computer-readable storage medium of claim 11 , wherein the operations further comprise performing random initialization and adjustment of one or more step sizes to generate the selected weights.
13. The non-transitory computer-readable storage medium of claim 11 , wherein the operations further comprise performing random initialization and one or more gradient updates to generate the selected weights.
14. A computer server comprising:
a computer processor; and
a non-transitory computer-readable medium storing computer program instructions executable by the processor to perform steps comprising:
receiving information indicative of initial entities for content and initial scores associated with the initial entities from one or more content annotation sources;
aggregating the information indicative of initial entities for content and initial scores associated with the initial entities received from the one or more content annotation sources;
defining a range of values;
mapping the initial scores to respective values within the defined range, the mapping producing calibrated scores within the defined range;
selecting weights based on joint performance conditions;
applying the selected weights to the calibrated scores within the defined range to produce weighted, calibrated scores;
selecting a linear aggregation model from among a plurality of linear aggregation models;
combining the weighted, calibrated scores based on the selected linear aggregation model to generate a final score; and
determining whether to annotate the content with at least one of the initial entities based on a comparison of the final score and a defined threshold value.
15. The server of claim 14 , wherein the steps further comprise comparing the final score to the defined threshold value.
16. The server of claim 14 , wherein the mapping further comprises mapping the initial scores to respective values within the defined range by employing isotonic regression.
17. The server of claim 14 , wherein an initial entity of the initial entities is associated with a property based on a calibrated score of the calibrated scores, and wherein the property is indicative of a level of relevance of the initial entity to the content.
18. The server of claim 14 , wherein the selected linear aggregation model is based on probabilities associated with the initial entities.
19. The server of claim 14 , wherein the selected linear aggregation model is based on a logarithm of one or more probabilities associated with the initial entities.