Guided search based on query model

1. A method of guided search, comprising:

receiving a new query;

processing the new query to obtain a plurality of models,

wherein the obtaining of the plurality of models comprises:

identifying a central phrase comprising a plurality of words or a central word of the new query; and

wherein:

a model indicates a mapping relationship of a previously stored query and corresponding guidance information;

the model includes information extracted from the new query, information transformed based on the new query, or both; and

the model characterizes the new query;

determining a corresponding plurality of similarities of the plurality of models relative to the new query, wherein the determining of the corresponding plurality of similarities comprises:

computing a similarity of one of the models with the new query based on a property of a model word in the one of the models, a property of a skipped word in the one of the models, or a combination thereof; and

in the event that the skipped word exists, calculating a penalty score based on the skipped word in the one of the models, comprising:

determining a first penalty score of the skipped word based on a part of speech of the skipped word;

determining a second penalty score based on a distance of the skipped word relative to the central phrase in the new query;

determining a third penalty score based on a distance of the skipped word relative to the central word in the new query; and

adjusting the similarity of the one of the models with the new query based on the first, second, and third penalty scores, comprising:

weighing one of the first penalty score, second penalty score or third penalty score by a first weight to obtain a first weighted penalty score;

weighing another one of the first penalty score, second penalty score or third penalty score by a second weight to obtain a second weighted penalty score, the first weight being different from the second weight; and

adjusting the similarity of the one of the models with the new query based on the first and second weighted penalty scores;

selecting at least one of the plurality of models based at least in part on the similarities;

obtaining guidance information by using the selected model as an index to search a database comprising a plurality of mapping relationships of previously stored queries and corresponding guidance information; and

sending the obtained guidance information to be displayed to a user.

2. The method of claim 1 , wherein processing the new query to obtain the plurality of models further includes extracting phrases of specified lengths, the phrases comprising the central phrase or the central word.

3. The method of claim 1 , wherein processing the new query to obtain the plurality of models includes skipping a plurality of words in the new query to generate the plurality of models.

4. The method of claim 1 , further comprising ranking the plurality of models according to their similarities.

5. The method of claim 1 , further comprising determining a corresponding plurality of confidences of the obtained guidance information relative to the selected model, and wherein obtaining the guidance information further includes selecting the at least some of the guidance information based at least in part on the confidences.

6. The method of claim 5 , wherein the corresponding plurality of confidences are specified based on historical data of previously stored queries and user selected guidance information and user selected intermediate information.

7. The method of claim 5 , wherein the corresponding plurality of confidences is specified based on historical data of previously stored queries and user selected guidance information and is determined based at least in part on probability values based on an occurrence rate of a previous stored query and the user selected guidance information occurring concurrently.

8. The method of claim 7 , wherein the corresponding plurality of confidences is determined by:

determining a first probability based on an occurrence rate of the previously stored query and the user selected guidance information occurring concurrently given a total number of occurrences of the previously stored query;

determining a second probability based on an occurrence rate of the previously stored query and user selected intermediate information occurring concurrently given a total number of occurrences of the previously stored query;

calculating a weighted sum of the first probability and the second probability to determine the confidence.

9. The method of claim 1 , further comprising predicting guidance information in a machine learning mode.

10. A system for providing guided search, comprising:

one or more processors configured to:

receive a new query;

process the new query to obtain a plurality of models,

wherein the obtaining of the plurality of models comprises:

identify a central phrase comprising a plurality of words or a central word of the new query; and

wherein:

a model indicates a mapping relationship of a previously stored query and corresponding guidance information;

the model includes information is extracted from the new query, information transformed based on the new query, or both; and

the model characterizes the new query;

determine a corresponding plurality of similarities of the plurality of models relative to the new query, wherein the determining of the corresponding plurality of similarities comprises:

compute a similarity of one of the models with the new query based on a property of a model word in the one of the models, a property of a skipped word in the one of the models, or a combination thereof; and

in the event that the skipped word exists, calculate a penalty score based on the skipped word in the one of the models, comprising:

determine a first penalty score of the skipped word based on a part of speech of the skipped word;

determine a second penalty score based on a distance of the skipped word relative to the central phrase in the new query;

determine a third penalty score based on a distance of the skipped word relative to the central word in the new query; and

adjust the similarity of the one of the models with the new query based on the first, second, and third penalty scores, comprising to:

 weigh one of the first penalty score, second penalty score or third penalty score by a first weight to obtain a first weighted penalty score;

 weigh another one of the first penalty score, second penalty score or third penalty score by a second weight to obtain a second weighted penalty score, the first weight being different from the second weight; and

adjust the similarity of the one of the models with the new query based on the first and second weighted penalty scores;

select at least one of the plurality of models based at least in part on the similarities;

obtain guidance information by using the selected model as an index to search a database, wherein the database comprises a plurality of mapping relationships of previously stored queries and corresponding guidance information; and

send the obtained guidance information to be displayed to a user; and

one or more memories coupled to the one or more processors, configured to provide the one or more processors with instructions.

11. The system of claim 10 , wherein processing the new query to obtain the plurality of models includes skipping a plurality of words in the new query to generate the plurality of models.

12. The system of claim 10 , wherein obtaining the guidance information includes using a selected one of the at least some of the plurality of models as an index to search the database.

13. The system of claim 10 , wherein the one or more processors are further configured to determine a corresponding plurality of confidences of the obtained guidance information to the plurality of models, and wherein obtaining the guidance information further includes selecting the at least some of the guidance information based at least in part on the confidences.

14. A computer program product for providing guided search, the computer program product being embodied in a non-transitory computer readable storage medium and comprising computer instructions for:

receiving a new query;

processing the new query to obtain a plurality of models,

wherein the obtaining of the plurality of models comprises:

identifying a central phrase comprising a plurality of words or a central word of the new query; and

wherein:

a model indicates a mapping relationship of a previously stored query and corresponding guidance information;

the model includes information is extracted from the new query, information transformed based on the new query, or both; and

the model characterizes the new query;

determining a corresponding plurality of similarities of the plurality of models relative to the new query, wherein the determining of the corresponding plurality of similarities comprises:

computing a similarity of one of the models with the new query based on a property of a model word in the one of the models, a property of a skipped word in the one of the models, or a combination thereof; and

in the event that the skipped word exists, calculating a penalty score based on the skipped word in the one of the models, comprising:

determining a first penalty score of the skipped word based on a part of speech of the skipped word;

determining a second penalty score based on a distance of the skipped word relative to the central phrase in the new query;

determining a third penalty score based on a distance of the skipped word relative to the central word in the new query; and

adjusting the similarity of the one of the models with the new query based on the first, second, and third penalty scores; comprising:

weighing one of the first penalty score, second penalty score or third penalty score by a first weight to obtain a first weighted penalty score;

weighing another one of the first penalty score, second penalty score or third penalty score by a second weight to obtain a second weighted penalty score, the first weight being different from the second weight; and

adjusting the similarity of the one of the models with the new query based on the first and second weighted penalty scores;

selecting at least one of the plurality of models based at least in part on the similarities;

obtaining guidance information by using the selected model as an index to search the database, wherein the database comprises a plurality of mapping relationships of previously stored queries and corresponding guidance information; and

sending the obtained guidance information to be displayed to a user.

15. The method of claim 1 , wherein the obtaining of the model comprises performing an N-gram technique or a Skip-Gram technique on the new query.

16. The method of claim 1 :

wherein the part of speech is adverb, adjective, numeral, verb, or noun.

17. The method of claim 1 , wherein:

the one penalty score corresponds to the third penalty score;

the other penalty score corresponds to the second penalty score; and

the first weight is greater than the second weight.