DYNAMIC COMPUTATION ENGINE IN SEARCH STACK
A data search and retrieval system that, in response to a search query, dynamically selects and applies a model of information to be returned to a user. The model may be selected based on the search query directly, or indirectly based on data returned by a search engine applying the query. For this purpose, the system may include an index of models, similar to a search index. Models may include expressions that may be descriptions of computations, equations, constraints or rules to define information to be returned. These elements of a selected model may be executed within to the search stack based on data identified in a search of web pages by executing the query or other dynamically obtained data. Execution of a model may lead to collecting new information, such as through further searching or performing computations, or may result in ordering or other formatting of data.
1 . A search and retrieval system comprising:
at least one processor for implementing a plurality of computer-executable components, the components comprising:
an input component adapted to receive from a user input defining a search query;
a data search engine adapted to generate a data portion of search results based on the search query;
a model selection component adapted to select a model defining information to be generated from among a plurality of models, the model selection component selecting the model based at least in part on the search query and/or the data portion of the search results;
a model execution engine adapted to execute the selected model to generate information; and
an output component adapted to return to the user in response to the received search query at least the information generated by the model execution engine.
2 . The search and retrieval system of claim 1 , further comprising a computer storage medium encoded storing the plurality of models, each of the plurality of models comprising declarative statements.
3 . The search and retrieval system of claim 1 , further comprising a computer storage medium encoded storing the plurality of models, each of the plurality of models comprising a combination of elements, the elements including one or more of calculations, equations, constraints and rules.
4 . The search and retrieval system of claim 2 , wherein:
at least a first of the plurality of models comprises an equation; and
when the first model is executed by the model execution engine, the model execution engine performs a mathematical operation on data in the data portion of the search results to generate information.
5 . The search and retrieval system of claim 2 , wherein:
at least a third of the plurality of models comprises a rule; and
when the third model is executed by the model execution engine, the model execution engine triggers a search by the data search engine in response to an evaluation of the rule.
6 . The search and retrieval system of claim 2 , wherein:
at least a fourth of the plurality of models comprises a calculation; and
when the fourth model is executed by the model execution engine, the model execution engine performs a numerical calculation on the data portion of the search results to generate information.
7 . The search and retrieval system of claim 1 , wherein the data portion of search results based on the search query comprises data related to medical records of a plurality of users.
8 . The search and retrieval system of claim 7 , wherein executing the selected model comprises generating a metric comparing a lab result of the user to lab results in the medical records.
9 . The search and retrieval system of claim 1 , wherein:
the system further comprises:
an index of data pages relating each of a plurality of search terms to one or more data pages; and
an index of models, relating each of a plurality of search terms to one or more models of the plurality of models;
the data search engine generates the data portion of search results based at least in part on the index of data pages; and
the model selection component selects a model based at least in part on the index of models.
10 . A method of operating a search system, the method comprising:
at least one processor, in response to a search query:
identifying a declarative model of information based at least in part on the search query;
dynamically identifying data based at least in part on the search query;
generating information in compliance with the model based at least in part on the dynamically identified data; and
generating a result based on the generated information.
11 . The method of claim 10 , wherein the generating information comprises:
performing a computation using as an input to the computation the dynamically identified data.
12 . The method of claim 10 , wherein the generating information comprises:
performing an Internet search with a search query generated by application of the declarative model to the dynamically identified data.
13 . The method of claim 10 , wherein:
the model identifies a time order of a plurality of events; and
the generating information comprises ordering elements within the dynamically identified data relating to the plurality of events in accordance with the time order of the model.
14 . The method of claim 10 , wherein:
the search query is received from a user over the Internet; and
the method further comprises conveying the result to the user over the Internet.
15 . The method of claim 10 , further comprising:
inferring a search query based on contextual information received from a user over the Internet.
16 . At least one non-transitory computer storage medium comprising computer-executable instructions that, when executed by at least one processor, perform a method within a search stack in a search system, the method comprising:
processing information representing a search query to:
identify a model based on an index of models;
dynamically identify data based on an index of data pages;
generating information by evaluating declarative statements representing equations in the identified model based on the dynamically identified data.
17 . The at least one non-transitory computer storage medium of claim 16 , wherein the method further comprises generating information comprising calculating a numerical value based on an expression in the identified model.
18 . The at least one non-transitory computer storage medium of claim 16 , wherein the method further comprises generating the search query based on context information associated with a user.
19 . The at least one non-transitory computer storage medium of claim 16 , wherein the generating information is further based on received input comprising the search query and a user context.
20 . The at least one non-transitory computer storage medium of claim 16 , wherein the method further comprises modifying the index of models based on user feedback.