Geographical knowledge graph

1. A method, comprising:

receiving plural terms corresponding to a map-based search query on a device;

determining plural interpretations of the map-based search query, each interpretation based on a respective combination of tagging one or more of the plural terms as at least one of a first type, a second type or a third type, wherein a first interpretation of the plural interpretations comprises tagging a first of the plural terms as the first type and a second of the plural terms as the second type which differs from the first type;

for each interpretation of the plural interpretations,

determining a set of weight values for the interpretation, based on at least one of context data of the device or on a feature of the respective combination of tagging one of more of the plural terms, and

assigning a score for the interpretation based on the set of weight values for the interpretation; and

providing at least one completion suggestion or search result based on the plural interpretations of the map-based search query, and on the assigned scores for the plural interpretations.

2. The method of claim 1 , wherein determining the set of weight values is performed by a machine learning model, and wherein the plural interpretations comprise the first interpretation based on a tagging of the first of the plural terms as the first type and the second of the plural terms as the second type and a second interpretation based on a tagging of the first of the plural terms as the second type and the second of the plural terms as the third type.

3. The method of claim 1 , wherein the context data of the device corresponds to at least one of a location of the device, a geographical region of a map being displayed on the device, or a preferred language for the device.

4. The method of claim 1 , wherein the feature corresponds to at least one of a location of a candidate result corresponding to the interpretation, or a popularity of the candidate result.

5. The method of claim 1 , wherein the first type is a concept corresponding to at least one of a brand name, a category, an attribute or a proximity, the second type is a point of interest, and the third type is a geographical entity.

6. The method of claim 5 , wherein the feature corresponds to a tag collocation value, which corresponds to a frequency with which a first tag of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity occurs with a second tag of one of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity.

7. The method of claim 5 , wherein the feature corresponds to an entity collocation value, which corresponds to a frequency with which a first of the plural terms tagged as one of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity occurs in sequence with a second of the plural terms tagged as one of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity.

8. The method of claim 5 , wherein tagging one or more of the plural terms is based on a geographical knowledge graph which is generated from click log data.

9. The method of claim 8 , wherein the geographical knowledge graph comprises at least one of:

a concept graph for identifying at least one of the brand name, the category, the attribute or the proximity,

a point of interest graph for identifying the point of interest, or

a geography graph for identifying the geographical entity.

10. The method of claim 9 , wherein each of the point of interest graph, the concept graph, and the geography graph is implemented as a weighted finite state transducer which analyzes terms on a per-character basis.

11. The method of claim 5 , wherein the point of interest is a predefined global point of interest for providing as the at least one completion suggestion or search result independent of device location.

12. A device, comprising:

at least one processor; and

a memory including instructions that, when executed by the at least one processor, cause the at least one processor to:

receive plural terms corresponding to a map-based search query;

determine plural interpretations of the map-based search query, each interpretation based on a respective combination of tagging one or more of the plural terms as at least one of a first type, a second type or a third type;

for each interpretation of the plural interpretations,

determine, by a machine learning model, a set of weight values for the interpretation, based on at least one of context data of the device or a feature of the respective combination of tagging one of more of the plural terms, and

assign a score for the interpretation based on the set of weight values for the interpretation; and

provide at least one completion suggestion or search result based on the plural interpretations of the map-based search query, and on the assigned scores for the plural interpretations.

13. The device of claim 12 , wherein the context data of the device corresponds to at least one of a location of the device, a geographical region of a map being displayed on the device, or a preferred language for the device.

14. The device of claim 12 , wherein the feature corresponds to at least one of a location of a candidate result corresponding to the interpretation, or a popularity of the candidate result.

15. The device of claim 12 , wherein the first type is a concept corresponding to at least one of a brand name, a category, an attribute or a proximity, the second type is a point of interest, and the third type is a geographical entity.

16. The device of claim 15 , wherein the feature corresponds to a tag collocation value, which corresponds to a frequency with which a first tag of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity occurs with a second tag of one of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity.

17. The device of claim 15 , wherein the feature corresponds to an entity collocation value, which corresponds to a frequency with which a first of the plural terms tagged as one of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity occurs in sequence with a second of the plural terms tagged as one of the brand name, the category, the attribute, the proximity, the point of interest, or the geographical entity.

18. The device of claim 15 , wherein tagging one of more of the plural terms is based on a geographical knowledge graph which is generated from click log data.

19. The device of claim 18 , wherein the geographical knowledge graph comprises at least one of:

a concept graph for identifying at least one of the brand name, the category, the attribute or the proximity,

a point of interest graph for identifying the point of interest, or

a geography graph for identifying the geographical entity.

20. A computer program product comprising code stored in a non-transitory computer-readable storage medium, the code comprising:

code to receive plural terms corresponding to a map-based search query on a device;

code to determine plural interpretations of the map-based search query, each interpretation based on a respective combination of tagging one or more of the plural terms as at least one of a first type, a second type or a third type, based on a geographical knowledge graph generated from click log data,

code to, for each interpretation of the plural interpretations,

determine a set of weight values for the interpretation, based on at least one of context data of the device or a feature of the respective combination of tagging one of more of the plural terms, and

assign a score for the interpretation based on the set of weight values for the interpretation; and

code to provide at least one completion suggestion or search result based on the plural interpretations of the map-based search query, and on the assigned scores for the plural interpretations.