Method and system to utilize session queries in real time to improve geo precision of sponsored listings
Contextual or cumulative interpretation is applied to individual user queries to improve the accuracy of sponsored and/or non-sponsored search results. Queries are interpreted based on present features derived from a user query in the context of past features derived from qualified past queries such as those previously received during a user query session. Based on contextual interpretation, present features may be augmented using past features. A search is performed using augmented features instead of present features. Contextual interpretation of queries improves search system understanding of the context in which queries are submitted, which is particularly helpful when interpolating query intent to develop features not expressed in a query. Contextual interpretation permits queries to be interpreted and treated more uniformly and more intelligently. As a result, users and advertisers benefit from more precise sponsored and/or non-sponsored search results. As one example, contextual interpretation can improve the geo-precision of sponsored search results.
1. A method executable by a computer having a processor and memory, the method comprising:
receiving a user query input to a search engine from a user computing device;
determining, with the processor, that the user query input to the search engine does not specify at least one present geographic feature identifying a geographic location associated with the user query;
retrieving, from a database, past geographic features associated with past user queries from the user computing device, wherein the past geographic features identify a geographic location associated with the past user queries;
determining, with the processor, at least one qualified past geographic feature for at least one past user query, wherein the at least one qualified past geographic feature is determined based on a time interval between the user query and the at least one past user query;
augmenting, with the processor, the at least one present geographic feature with the at least one qualified past geographic feature to determine an augmented geographic feature, wherein the augmented geographic feature identifies a geographic location to apply to the user query that does not specify the at least one present geographic feature,
further wherein the augmented geographic feature is determined by selecting the most relevant qualified past geographic feature from the past geographic features;
determining, with the processor, at least one sponsored search result using the augmented geographic feature by selecting the sponsored search result using the geographic location identified by the augmented geographic feature;
determining, with the processor, at least one non-sponsored search result using the at least one present geographic feature, such that different geographic features are used to determine the sponsored search results as compared with the non-sponsored search results for the user query; and
delivering, with the processor, the at least one sponsored search result and the at least one non-sponsored search result to be displayed on a web page of the user computing device.
2. The method of claim 1 , further comprising:
qualifying the at least one qualified past geographic feature for at least one past query by determining that the at least one past user query and the user query were received during a same session.
3. The method of claim 2 , wherein the same session comprises a same geographic session based on queries having a same geographic context.
4. The method of claim 3 , wherein query geographic context is determined by query pattern recognition.
5. The method of claim 2 , further comprising:
further qualifying the at least one past geographic feature for at least one past query by determining that the at least one past query and the user query were received within a qualifying time interval.
6. The method of claim 1 , further comprising:
determining weights for the at least one present geographic feature and the at least one qualified past geographic feature;
merging the weighted at least one past geographic feature and the weighted at least one present geographic features into the augmented at least one present geographic feature;
wherein the merging comprises eliminating the weighted at least one qualified past feature and the weighted at least one present geographic features whose weight is less than a threshold weight.
7. The method of claim 1 , further comprising:
determining weights for the at least one present geographic feature and the at least one qualified past geographic feature;
merging the weighted at least one past geographic feature and the weighted at least one present geographic features into the augmented at least one present geographic feature,
wherein the merging comprises eliminating duplicate features and adjusting a weight of a feature whose duplicate was eliminated.
8. The method of claim 1 , further comprising:
storing the augmented at least one present geographic feature as an entry in a data structure used to interpret subsequent queries.
9. The method of claim 8 , wherein a stored entry in the data structure comprises a session identifier, at least one geographic feature and a query timestamp.
10. The method of claim 1 , wherein the method is selectively applied when geographic intent is not expressed in the user query and is not applied when geographic intent is expressed in the user query.
11. The method of claim 1 , where augmenting comprises:
weighting the at least one present geographic feature;
weighting the at least one qualified past geographic feature; and
merging the weighted at least one past geographic feature and the weighted at least one present geographic feature into the augmented at least one present geographic feature.
12. The method of claim 1 , wherein selecting the most relevant qualified past geographic feature from the past geographic features comprises:
assigning a weight to each of the past geographic features based on whether each of the past geographic features was explicitly contained in the user query or derived from the user query, and
adjusting the assigned weights based on elapsed time in order to emphasize more recent user queries and past geographic features.
13. The method of claim 12 , wherein weighted the past geographic features based on whether each of the past geographic features was explicitly contained in the user query or implied by the user query further comprises:
identifying each of the past geographic features that was explicitly contained in the user query and increasing the assigned weight of the past geographic feature when it was explicitly contained in the user query,
identifying each of the past geographic features that was derived from the user query and decreasing the assigned weight of the past geographic feature when it was derived from the user query.
14. A search system, comprising:
a processor and a memory;
a session-based query processing module executable by the processor and that interprets a user query based on the user query and previous queries accumulated during a session with a computing device of the user, wherein the session-based query processing module is configured to:
receive the user query from the user computing device;
determine that user query from the user computing device does not specify at least one present geographic feature;
determine at least one present geographic feature for the user query;
retrieve, from a database, past geographic features associated with past user queries from a related user computing device;
determine at least one past geographic feature for at least one past user query received during the same session as the user query;
augment the user query with the at least one past geographic feature;
determine a set of non-sponsored search results using the at present geographic feature; and
determine a set of sponsored search results using the augmented user query and the at least one past geographic feature, such that different geographic features are used to determine the non-sponsored search results and the sponsored search results; and
deliver the set of non-sponsored search results and the set of sponsored search results to be displayed on a web page of the user computing device.
15. The search system of claim 14 , wherein the session comprises a geographic session.
16. The search system of claim 14 , wherein the session-based query module comprises:
a geographic identification module that identifies geographic features in queries; and
a geographic session cache module configured to store past geographic features for past queries.
17. The search system of claim 14 , where the session-based query processing module is further configured to:
weight the at least one present geographic feature;
weight the at least one qualified past geographic feature; and
merge the weighted at least one past geographic feature and the weighted at least one present geographic feature into the augmented at least one present geographic feature.
18. A non-transitory computer readable medium comprising computer-executable instructions that, when executed by a computer having a processor and memory, perform session-based query processing that interprets a user query based on the user query and previous queries accumulated during a session with a computing device of the user by:
determining, with the processor, at least one present geographic feature for the user query;
determining, with the processor, at least one past geographic feature for at least one past user query received during the session;
augmenting, with the processor, the at least one present geographic feature with the at least one past geographic feature to determine an augmented geographic feature, wherein the augmented geographic feature is determined by selecting the most relevant past geographic feature determined during the session;
applying, with the processor, the augmented geographic feature to the user query to generate an interpreted user query that includes the augmented geographic feature;
determining, with the processor, a set of non-sponsored search results based on the at least one present geographic feature;
determining, with the processor, a set of sponsored search results based on the interpreted user query, such that different geographic features are used to determine the non-sponsored search results and the sponsored search results; and
delivering, with the processor, the set of non-sponsored search results and the set of sponsored search results to be displayed on a web page of the user computing device.
19. The non-transitory computer readable medium of claim 18 , wherein the session comprises a geographic session.
20. The non-transitory computer readable medium of claim 18 , further comprising instructions executable by the processor to:
weight the at least one present geographic feature;
weight the at least one qualified past geographic feature; and
merge the weighted at least one past geographic feature and the weighted at least one present geographic feature into the augmented at least one present geographic feature.