Sponsored search ranking simulation for patterns triggered by natural language queries
The technology disclosed relates to natural language understanding-based search engines, ranking sponsored search results and simulated ranking of sponsored search results. Tools and methods describe how to simulate the ranking of sponsored search results. The tools further identify instances of user queries within the scope of trigger patterns, optionally providing examples both of user queries for which a sponsored search result is likely to be displayed and examples for which the sponsored search result will not rank highly enough to be displayed, at least on the first page of search results.
1. A method of predicting future ad price distributions, the method comprising:
configuring a memory with a database of past natural language queries, the past natural language queries being represented as deep structures;
configuring the memory with a database of active sponsored messages, the active sponsored messages comprising natural language trigger patterns and corresponding prices;
configuring the memory with a database of projected results of the past queries processed against the natural language trigger patterns and the corresponding prices of the active sponsored messages;
indexing data stored in the database of past natural language queries, the indexing being based on the deep structures;
predicting future ad price distributions from the projected results of the past queries processed against the natural language trigger patterns and the corresponding prices of the active sponsored messages; and
reporting the future ad price distributions to an advertiser.
2. A method of indexing data, the method comprising:
configuring a memory with a database of past natural language queries, the past natural language queries being represented as deep structures;
configuring the memory with a database of active sponsored messages, the active sponsored messages comprising natural language trigger patterns and corresponding prices;
configuring the memory with a database of projected ad results of the past natural language queries processed against the natural language trigger patterns and the corresponding prices of the active sponsored messages;
indexing the past natural language queries as a joint distribution of multiple variables; and
reporting statistics about the projected ad results to an advertiser, the statistics including an estimated frequency of past queries that trigger the natural language trigger patterns of the active sponsored messages.
3. The method of claim 1 , wherein at least one active sponsored message of the active sponsored messages is added to the database of active sponsored messages after the past queries have been received and stored in the database of past queries.
4. The method of claim 1 ,
wherein the active sponsored messages include trigger pattern criteria provided by sponsors of results,
wherein the projected results are determined by identifying trigger pattern criteria of an active sponsored message, of the active sponsored messages, that matches a past query of the past queries, and
wherein the method further comprises providing a sponsored result associated with the identified trigger pattern criteria that matches the past query.
5. The method of claim 1 , wherein the past queries are natural language queries obtained from spoken words.
6. The method of claim 5 , wherein the past queries are represented as deep structures, the deep structures using semantic representation that populates predefined categories using words included in the natural language queries.
7. The method of claim 1 , wherein the active sponsored messages are paid-for active sponsored messages.
8. The method of claim 7 , wherein the active sponsored message comprise sponsored ads.
9. The method of claim 1 , wherein the corresponding prices are prices for advertising the active sponsored messages.
10. The method of claim 1 , wherein the corresponding prices are prices related to an advertised products associated with the active sponsored messages.
11. The method of claim 1 , wherein the corresponding prices include (i) prices for advertising the active sponsored messages and (ii) prices related to advertised products associated with the active sponsored messages.
12. The method of claim 1 , further comprising displaying an active sponsored message, of the active sponsored messages, on a search results page or a landing page reached from the search results page.
13. The method of claim 1 , further comprising reporting samples of past queries processed against the natural language trigger patterns.
14. The method of claim 1 , wherein the natural language trigger patterns include geographic criteria and the database of past queries includes categorical data regarding respective geographic focuses of the past queries.
15. The method of claim 1 , wherein the natural language trigger patterns include geographic criteria and the database of past queries includes data identifying respective geographic focuses of the past queries.
16. The method of claim 1 , wherein the natural language trigger patterns include temporal criteria and the database of past queries includes categorical data regarding respective temporal focuses of the past queries.
17. The method of claim 1 , wherein the natural language trigger patterns include temporal criteria and the database of past queries includes data identifying respective temporal focuses of the past queries.
18. The method of claim 1 , further comprising receiving a plurality of geographic criteria and reporting a number of previous active messages that would have been delivered in response to the past queries matched with the plurality of geographic criteria.
19. The method of claim 1 , wherein the natural language trigger patterns include quality criteria and the database of past queries includes data identifying respective quality focuses of the past queries.
20. A system including one or more processors coupled to memory, the memory being loaded with computer instructions to predict future ad price distributions, the computer instructions, when executed on the one or more processors, causing the one or more processors to implement actions comprising:
configuring a memory with a database of past natural language queries, the past natural language queries being represented as deep structures;
configuring the memory with a database of active sponsored messages, the active sponsored messages comprising natural language trigger patterns and corresponding prices;
configuring the memory with a database of projected results of the past queries processed against the natural language trigger patterns and the corresponding prices of the active sponsored messages;
indexing data stored in the database of past natural language queries, the indexing being based on the deep structures;
predicting future ad price distributions from the projected results of the past natural language queries processed against the natural language trigger patterns and the corresponding prices of the active sponsored messages; and
reporting the future ad price distributions to an advertiser.