Labeling context slices to produce a storyline from mobile device data
Embodiments create and label context slices from observation data that together define a storyline of a user's movements. A context is a (possibly partial) specification of what a user was doing in the dimensions of time, place, and activity. Contexts can vary in their specificity, their semantic content, and their likelihood. A storyline is composed of a time-ordered sequence of contexts that partition a given span of time. A storyline is created through a process of data collection, slicing and labeling. Raw context data can be collected from a variety of observation sources with various error characteristics. Slicing refines the chaotic collection of contexts produced by data collection into a single consistent storyline composed of a sequence of contexts representing homogeneous time intervals. Labeling adds more specific and semantically meaningful data (e.g., geography, venue, activity) to the storyline produced by slicing.
1. A method of labeling context slices of a storyline of a user's movements, the method comprising:
receiving a plurality of context slices derived from associated context data collected from a plurality of observation sources; and
for each of the plurality of context slices:
determining an uncertainty in a location of the slice based on a type of observation source from which the associated context data originated;
identifying a set of candidate labels based on context data associated with the context slice, each candidate label matching the location of the slice within the predetermined uncertainty, the candidate labels each comprising semantic data describing the context data associated with the context slice, the semantic data selected from a group consisting of geography, venue, and activity;
ranking the set of candidate labels by likelihood;
applying one or more of the candidate labels to the context slice; and
storing a correspondence between the applied one or more labels and the context slice.
2. The method of claim 1 , wherein the set of candidate labels are from public venue records, a user-specified database of places, a database of descriptions of activities commonly applicable to venue categories, calendar data, and user corrections to automatically applied labels.
3. The method of claim 1 , wherein identifying a set of candidate labels comprises searching for candidate labels for geography and venue based on the context slice's location.
4. The method of claim 1 , wherein identifying a set of candidate labels comprises searching for candidate labels for events based on the context slice's location and time boundaries.
5. The method of claim 1 , wherein identifying a set of candidate labels comprises searching for candidate labels for activities based on the context slice's labeled venue.
6. The method of claim 1 , wherein identifying a set of candidate labels comprises searching for candidate labels for activities based on a label applied to a similar slice in the past.
7. The method of claim 1 , wherein ranking based on likelihood comprises determining a likelihood based on a score of a model of features of the context slice including slice location, user, and beginning and end times of the context slice.
8. The method of claim 7 , wherein the model is further informed by venue hours of operation, venue popularity, category popularity, user routine, or social interest.
9. The method of claim 1 , further comprising providing a plurality of candidate labels to a user for possible selection.
10. The method of claim 1 , further comprising providing the storyline for display to the user, the storyline including the plurality of labeled context slices.
11. A non-transitory computer-readable storage medium having computer program instructions embodied therein for labeling context slices of a storyline of a user's movements, the computer program instructions comprising instructions for:
receiving a plurality of context slices derived from associated context data collected from a plurality of observation sources; and
for each of the plurality of context slices:
determining an uncertainty in a location of the slice based on a type of observation source from which the associated context data originated;
identifying a set of candidate labels based on context data associated with the context slice, each candidate label matching the location of the slice within the determined uncertainty, the candidate labels each comprising semantic data describing the context data associated with the context slice, the semantic data selected from a group consisting of geography, venue, and activity;
ranking the set of candidate labels by likelihood;
applying one or more of the candidate labels to the context slice; and
storing a correspondence between the applied one or more labels and the context slice.
12. The computer-readable storage medium of claim 11 , wherein the set of candidate labels are from public venue records, a user-specified database of places, a database of descriptions of activities commonly applicable to venue categories, calendar data, and user corrections to automatically applied labels.
13. The computer-readable storage medium of claim 11 , wherein the instructions for identifying a set of candidate labels comprise instructions for searching for candidate labels for geography and venue based on the context slice's location.
14. The computer-readable storage medium of claim 11 , wherein the instructions for identifying a set of candidate labels comprise instructions for searching for candidate labels for events based on the context slice's location and time boundaries.
15. The computer-readable storage medium of claim 11 , wherein the instructions for identifying a set of candidate labels comprise instructions for searching for candidate labels for activities based on the context slice's labeled venue.
16. The computer-readable storage medium of claim 11 , wherein the instructions for identifying a set of candidate labels comprise instructions for searching for candidate labels for activities based on a label applied to a similar slice in the past.
17. The computer-readable storage medium of claim 11 , wherein the instructions for ranking based on likelihood comprise instructions for determining a likelihood based on a score of a model of features of the context slice including slice location, user, and beginning and end times of the context slice.
18. The computer-readable storage medium of claim 17 , wherein the model is further informed by venue hours of operation, venue popularity, category popularity, user routine, or social interest.
19. The computer-readable storage medium of claim 11 , the instructions further comprising instructions for providing a plurality of candidate labels to a user for possible selection.
20. The computer-readable storage medium of claim 11 , the instructions further comprising instructions for providing the storyline for display to the user, the storyline including the plurality of labeled context slices.