Information search method, apparatus, and system

1. A method of searching for information, comprising:

receiving, by a server, a search word sent by a user by using a client;

recognizing, by the server, search intention of the user according to feature information of the search word and/or search context feature information;

performing, by the server, search by using a search policy corresponding to the search intention and according to the search word to acquire information search results associated with the search word, wherein the information search results comprise a local search result and/or a non-local search result; and

sending, by the server, the information search results to the client to enable the client to display the information search results,

wherein recognizing the search intention of the user according to the feature information of the search word and/or the search context feature information comprises:

acquiring, by the server, the feature information of the search word and/or the search context feature information;

determining, by the server, non-local search preference of the user according to the feature information of the search word and/or the search context feature information and by using a pre-trained non-local search preference determination model; and

determining, by the server, the search intention of the user according to the non-local search preference and a preset preference threshold condition,

wherein the non-local search preference determination model is trained by:

constructing, by the server, the non-local search preference determination model by using a feature of the search word and/or a search context feature as an independent variable and using a non-local search preference feature as a dependent variable;

acquiring, by the server, training samples generated according to a plurality of search click historical records; and

training, by the server, the non-local search preference determination model by using the training samples,

wherein constructing the non-local search preference determination model comprises:

constructing, by the server, the non-local search preference determination model by using a logic regression model using the following:

remote prob =sigmoid (w 1 x 1 +w 2 x 2 + . . . +w n x n ), wherein

remote prob represents the non-local search preference feature,

x n represents an eigenvalue of an n-th feature,

w n represents a weight corresponding to the n-th feature, a weight corresponding to each of the n features is determined by training according to chosen training samples, and

the function sigmoid is a threshold function of a neural network, and the threshold function is used to map a variable into a value between 0 and 1.

2. The method according to claim 1 , further comprising:

matching, by the server, the search word against a preset local and non-local entity word library, to determine whether the search word has a possibility of non-local search; and

if there is a possibility of non-local search, recognizing, by the server, the search intention of the user according to the feature information of the search word and/or the search context feature information.

3. The method according to claim 1 , wherein performing search by using the search policy corresponding to the search intention and according to the search word to acquire information search results associated with the search word comprises:

if the search intention is that the user has both local search intention and non-local search intention, performing, by the server, local search and non-local search according to the search word, to obtain the information search results comprising a plurality of local search results and a plurality of non-local search results.

4. The method according to claim 3 , further comprising:

incorporating, by the server, the plurality of local search results into a local search result display region;

incorporating, by the server, the plurality of non-local search results into a non-local search result display region;

determining, by the server, a display order of the local search result display region and the non-local search result display region according to non-local search preference of the user and/or quality of the local and non-local search results; and

generating, by the server, the information search results comprising the local search result display region and the non-local search result display region in the display order.

5. The method according to claim 4 , further comprising:

determining, by the server, an order of the plurality of local search results in the local search result display region according to a pre-estimated click through rate of each of the plurality of the local search results, and

determining, by the server, a quantity of local search results displayed when the local search result display region is in a collapsed state.

6. The method according to claim 4 , further comprising:

determining, by the server, an order of the plurality of the non-local search results in the non-local search result display region according to a pre-estimated click through rate of each of the plurality of the non-local search results, and

determining, by the server, a quantity of non-local search results displayed when the non-local search result display region is in a collapsed state.

7. The method according to claim 1 , further comprising:

acquiring, by the server, a click operation log of the user on the information search results; and

performing, by the server, iterative optimization on the non-local search preference determination model according to the click operation log.

8. A method of searching for information, comprising:

monitoring, by a client, a search word entered by a user;

sending, by the client, the search word to a server;

receiving, by the client, information search results fed back by the server, wherein the server uses a search policy corresponding to search intention of the user to perform search to acquire the information search results, the search intention of the user being recognized by the server according to feature information of the search word and/or search context feature information, and the information search results comprise a local search result and/or a non-local search result; and

displaying, by the client, the information search results,

wherein the search intention of the user being recognized by the server according to feature information of the search word and/or search context feature information comprises:

acquiring, by the server, the feature information of the search word and/or the search context feature information;

determining, by the server, non-local search preference of the user according to the feature information of the search word and/or the search context feature information and by using a pre-trained non-local search preference determination model; and

determining, by the server, the search intention of the user according to the non-local search preference and a preset preference threshold condition,

wherein the non-local search preference determination model is trained by:

constructing, by the server, the non-local search preference determination model by using a feature of the search word and/or a search context feature as an independent variable and using a non-local search preference feature as a dependent variable;

acquiring, by the server, training samples generated according to a plurality of search click historical records; and

training, by the server, the non-local search preference determination model by using the training samples,

wherein constructing the non-local search preference determination model comprises:

constructing, by the server, the non-local search preference determination model by using a logic regression model using the following:

remote prob =sigmoid (w 1 x 1 +w 2 x 2 + . . . +w n x n ), wherein

remote prob represents the non-local search preference feature,

x n represents an eigenvalue of an n-th feature,

w n represents a weight corresponding to the n-th feature, a weight corresponding to each of the n features is determined by training according to chosen training samples, and

the function sigmoid is a threshold function of a neural network, and the threshold function is used to map a variable into a value between 0 and 1.

9. The method according to claim 8 , wherein displaying the information search results comprises:

displaying, by the client, a local search result display region and a non-local search result display region in the information search results on the same screen of the client in a display order of the information search results, wherein the local search result display region comprises at least one local search result, and the non-local search result display region comprises at least one non-local search result.

10. A computer device, comprising:

a memory;

a processor; and

computer programs stored in the memory and executable by the processor;

wherein when executing the computer programs, the processor is caused to perform the method according to claim 1 .

11. A computer device, comprising:

a memory;

a processor; and

computer programs stored in the memory and executable by the processor;

wherein when executing the computer programs, the processor is caused to perform the method according to claim 8 .

12. A computer readable storage medium storing computer programs, wherein when executing the computer programs, a processor is caused to perform the method according to claim 1 .

13. A computer readable storage medium storing computer programs, wherein when executing the computer programs, a processor is caused to perform the method according to claim 8 .