HK1183540A1 - Method for searching and system thereof - Google Patents

Abstract

The invention provides a searching method and system, which relate to the technical field of network. The searching method comprises the following steps of utilizing a first separator to splice the header message field area and the attribute information field area of a target object to form a new field, and constructing an index based on the target object; after constructing the index, computing search terms of a user based on the index and the first separator, and computing a total relevancy of the search terms and the new field according to a field area, on which each query term of the new field is located; and returning at least one target object corresponding to the new field based on the total relevancy of each new field and the search terms. According to the searching method and the system, headers and brand messages of products are spliced to form the new field by utilizing the separator, and search engine index construction is carried out on the new field, so that a product result meeting the expectation of a user can be quickly returned; and in addition, text relevancy can be computed at one time only through incrementally updating the index once, so that the computing cost and the hardware resources are greatly reduced.

Description

Searching method and system

Technical Field

The present application relates to the field of network technologies, and in particular, to a search method and system.

Background

A product or article will typically have a brand. Such as sports shoes, with the adidas brand, with the nikker brand, and with the leining brand. Branding is undoubtedly the most powerful evidence for the quality of the goods, being the guarantee of service. With the development of network technology, more and more users purchase commodities on the internet, and because of lack of a link of field experience, brand information of the commodities is more important to the quality of the commodities, so that in electronic commerce, it is more important that a system returns accurate brand information according to search terms of the users.

In the prior art, one method is to perform keyword matching only on the title of the product, which may result in the production of brand products not desired by the user, such as search for adedas, and the search result only gives the product with the title containing different types of keywords such as adedas, addas, etc., but it is possible that the brand attribute of the product is not the adedas brand. In addition, for a product that is an adidas brand, it is easy to miss in search ranking because no adidas-related brand keyword appears in the title.

The other method is to establish two separate engines, namely a title engine of a commodity and a brand information engine, after segmenting the search terms of the user, matching operation is carried out on the segmented query terms in the title engine and the brand information engine respectively, and then the calculation results of the two engines are combined to calculate the overall relevancy. However, if the title and the brand of the product are edited again, the index of 2 engines needs to be updated incrementally at the same time, even if only one of the title or the brand information is modified, the index of 2 engines needs to be updated at the same time, which not only needs to add more extra computing cost, but also has more huge computing cost for storing the product in a non-location mode in each engine when updating the index of the engine, so that the method has the disadvantages of slow processing, higher maintenance cost, higher hardware cost and unfavorable fast updating.

Disclosure of Invention

The technical problem to be solved by the application is to provide a searching method and a searching system, which can quickly process returned results, are convenient to maintain and have low maintenance cost.

In order to solve the above problem, the present application discloses a search method, including:

for a search word input by a user, acquiring each query word corresponding to the search word;

searching each index word corresponding to each query word in an index aiming at each obtained query word, wherein the index is constructed according to a field of a target object, and the field of the target object comprises a new field formed by splicing a title information field area and an attribute information field area of the target object through a first separator;

according to the position of each index word in the new field and the position of the first separator in the new field, determining whether the query word corresponding to the index word belongs to the title information field area or the attribute information field area in the new field;

calculating the total correlation degree of the search word and the new field according to the field area of the new field to which each query word belongs; the total correlation comprises a first correlation calculated according to the weight of the field area of the new field where each query word is located;

and returning the target object corresponding to at least one new field based on the total relevance of each new field and the search word.

Preferably, the new field formed by splicing the header information field area and the attribute information field area of the target object by the first separator includes the following steps:

reading a title information field area and an attribute information field area of the target object;

replacing the same character in the new field as the first delimiter with a blank character;

and splicing the replaced header information field area and the attribute information field area into a new field through a first separator.

Preferably, the index is constructed according to each participle in the field by the following steps:

corresponding the identification of each target object with the corresponding new field through a second separator;

performing word segmentation operation on each new field;

and taking the participles obtained by the participle operation as index words, and corresponding the index words with the identifications of the related target objects and the positions of the index words in the new fields.

Preferably, it is confirmed whether the query word belongs to the title information field area or the attribute information field area by:

inquiring a new field corresponding to the identification according to the corresponding relation between the index word and the identification of each related target object;

and comparing the position of the index word in the new field with the position of the first separator in the new field, and confirming that the query word corresponding to the index word belongs to the title information field area or the attribute information field area.

Preferably, the first correlation is obtained by:

dividing the character string length of each query word by the character string length of the field area to obtain the interval correlation degree of each query word and the field area;

and multiplying the correlation degrees by the weight of the field area, and adding to obtain the first correlation degree of the search term and the new field.

Preferably, the search term includes:

taking keywords input by a user as search words;

or, one of the suggested words returned according to the input words of the user selected by the user is used as a search word; the suggested words are obtained by extracting the click relation between the input words input by the user and the corresponding results which are counted in advance.

Preferably, for a search term input by a user, obtaining a query term of the search term includes:

and correcting the error of the search word input by the user by an intelligent error correction engine.

Preferably, the target object comprises a commodity; the attribute information includes brand information of the commodity.

Preferably, when at least one new field is output to the user side based on the total correlation degree of each new field and the search term:

outputting at least one top-ranked target object; the target objects are ranked based on the total relevance of the respective new field to the search terms.

Correspondingly, the application discloses a search apparatus, includes:

the query term acquisition module is used for acquiring each query term corresponding to a search term input by a user;

the index word searching module is used for searching each index word corresponding to each query word in the index, the index is constructed according to the field of the target object, and the field of the target object comprises a new field formed by splicing the title information field area and the attribute information field area of the target object through a first separator;

the position confirmation module is used for confirming whether the query word corresponding to the index word belongs to the title information field area or the attribute information field area in the new field according to the position of each index word in the new field and the position of the first separator in the new field;

the relevancy calculation module is used for calculating the total relevancy between the search word and the new field according to the field area to which each query word of the new field belongs; the total correlation comprises a first correlation calculated according to the weight of the field area of the new field where each query word is located;

and the output module is used for returning the target object corresponding to at least one new field based on the total correlation degree of each new field and the search word.

Compared with the prior art, the method has the following advantages:

according to the method and the device, the title of the commodity and the brand information of the commodity are spliced into the new field by using the separators, then the new field is subjected to search engine index construction, the commodity result meeting the expectation of the user can be quickly returned through the method and the device, only the index needs to be updated incrementally, when the text correlation is calculated, the operation can be completed once, and therefore the calculation cost and the hardware resources are greatly reduced.

Drawings

FIG. 1 is a schematic flow chart of a search method of the present application;

fig. 2 is a schematic structural diagram of a search apparatus according to the present application.

Detailed Description

In order to make the aforementioned objects, features and advantages of the present application more comprehensible, the present application is described in further detail with reference to the accompanying drawings and the detailed description.

Referring to fig. 1, a schematic flow chart of a search method according to the present application is shown, which includes:

step 110, for the search terms input by the user, obtaining each query term corresponding to the search terms.

In practice, for a search term input by a user, such as "adidas clover", its query term, such as "adidas" or "clover", is obtained. Generally, a word segmentation operation may be performed on a search term input by a user, for example, if the search term input by the user is "adidas clover", the search term may be segmented into two query terms "adidas" and "clover" according to information of a commodity and a longest matching principle.

In practice, the search terms input by the user in error can be corrected by the intelligent error correction engine. For example, if the user inputs "addida four" to input "addida four" basically, the intelligent error correction engine may correct the addida four "input by the user to" addida "for subsequent processing according to the actual statistical analysis result.

In addition, for the search term input by the user, the keyword input by the user may be used as the search term.

If the user directly uses the keyword input by the user as a search word, for example, the user inputs "Addi", the user directly clicks the confirmation search and uses the keyword as a search word input to the search engine.

Or, one of the suggested words returned according to the input words of the user selected by the user can be used as the search word; the suggested words are obtained by extracting the click relation between the input words input by the user and the corresponding results which are counted in advance.

For example, if the user inputs "addis", the system may return suggested words "addis", "addowang", "addis clover", etc. according to the actual statistical analysis result, and the user may select one of the suggested words as a search word according to his own needs and finally input the search word to the search engine.

And step 120, searching each index word corresponding to each query word in the index according to each obtained query word, wherein the index is constructed according to the field of the target object, and the field of the target object comprises a new field formed by splicing the title information field area and the attribute information field area of the target object through a first separator.

In the present application, preferably, the target object includes a commodity; the attribute information includes brand information of the commodity.

Before the application processes the search terms input by the user, the method further comprises the steps of constructing an index and establishing a search engine, and specifically comprises the following steps:

step S101, the title information field area and the attribute information field area of the target object are spliced into a new field through a first separator.

Preferably, the step of forming the header information field region and the attribute information field region of the target object into new fields by the first delimiter splicing is performed by:

step a1, the title information field area and the attribute information field area of the target object are read.

In practice, before splicing, the header information and the brand information of each target object in the database need to be read, in this embodiment, the target object includes a commodity, and the attribute information includes brand information of the commodity.

Step a2, replacing the same character in the new field as the first delimiter with a blank character.

The title information field area and the brand information field area of the commodity are read first, and characters identical to the first separator in the title information field area and the brand information field area are replaced by blank character strings.

The first delimiter is: the characters separating the brand information and the title information are arranged in the text, and the value can be actually taken as a symbol which does not frequently appear in the title or the brand information of the commodity. Such as tab \ t, space, slash, comma, etc., are relatively easy to appear in the title or brand information and are therefore not suitable as separators, whereas such as ctrl + a, ascii code value 0x01 or & &, which would not normally appear in the text string, can be used as the first separator.

In practice, after the first separator is selected according to the above principle, the same character as the first separator may exist in the title information field area and the brand information field area of the product, and then the same character as the first separator needs to be replaced in the title information field area and the brand information field area of the product for subsequent processing.

Step a3, the replaced header information field area and attribute information field area are spliced into a new field by a first separator.

Suppose that the brand of a certain commodity is "clover", the title is "1 discount swing of adidas sports shoes", the first separator is ctrl + a, and the finally formed character string is: the clover Adida sports shoe 3 is folded and sold, and the position of the separator is recorded as 6 (the positions 0-5 are clovers, the position 6 is the separator, and the rest is commodity title information).

And step S102, constructing an index aiming at the new field and establishing a search engine.

Preferably, the index is constructed according to each participle in the field by the following steps:

and step B1, corresponding the identification of each target object to the corresponding new field through a second separator.

The item typically corresponds to a new field of the item by its identification (typically a numeric id).

In practice, the file format when the goods are stored is the goods number id and the new character string, namely, two fields of the animation _ id and the brand _ title: digital type and character type. The 2 columns of numeric and character types are separated by another separator (i.e., a second separator) and must not be the same as the first separator in step a 3. And the same character as the second separator is replaced in the title information field area and the brand information field area for the goods.

For example, for the first delimiter ctrl + a and the second delimiter | |, the storage format for the following two products is:

"12345 | | clover adidas sports shoes 1 turn over to get rid of selling"

"83635789 | | clover 2011 autumn new product blackboard shoe 8 is folded and mailed".

And step B2, performing word segmentation operation on each new field.

In step B1, performing word segmentation on the new field following the second separator | |, where the word segmentation results are in turn:

"three-leaf clover Adida sports shoes 1 fold for getting rid of sales"

Three-leaf clover 2011 autumn new product plank shoe 8-fold bag stamp "

In practice, when each new field is processed during the index establishment, words can be segmented according to actual requirements, for example, for the clover Adida sports shoe 1 discount and throw-away, "in addition to the above word segmentation result, words such as" clover "," Addi "and the like can be segmented.

And step B3, using the participle obtained by the participle operation as the index word, and making the index word correspond to the mark of each related target object and the position of the index word in each new field.

For example, an index is established for the word segmentation result, and a commodity number id (animation _ id) is followed by an index word, so that the result of establishing the index is as follows:

clover-12345 _0, 83635789_0

Adida-12345 _7

Sport footwear-12345 _15

1 is folded to-12345 _21

Swing-12345 _24

2011 autumn- - -83635789_7

New product-83635789 _13

Plank shoe-83635789 _17

8-fold- - -83635789_21

Bag mail- -83535789_24

The index is preceded by an index word and followed by the commodity id that can be referred by the word and the position where the word appears (for the position label, a _ "or the like can be adopted), and the index is pressed into the memory, so that the high-speed query efficiency is ensured. The chinese character can be written as 2 bytes.

Then, for the above index, according to the query words "adidas" and "clover" obtained in step 110, the search engine is searched by using the above two words, and the result is: adida: 12345_ 7; clover: 12345_0 and 83635789_ 0.

Step 130, according to the position of each index word in the new field and the position of the first separator in the new field, determining whether the query word corresponding to the index word belongs to the title information field region or the attribute information field region in the new field.

For the search result, the string index position of each search term in the character string is calculated and compared with the position of the separator, for example, in the foregoing example, if the brand information field area is before and the title information field area is after, then if the position of the index term corresponding to the query term in the new field is smaller than the first separator position, it indicates that the participle exists in the brand information field area, and if the position of the index term corresponding to the query term in the new field is larger than the first separator position, it indicates that the participle exists in the commodity title information field area.

For example, the search results in the index in the previous example are: adida: 12345_ 7; clover: 12345_0 and 83635789_0, the matched commodity id and the position of the first separator in the new field can be determined according to the established structure of the index. Then the location of "adidas" in the new field corresponding to 12345 is 7, which is larger than the location 6 of the first delimiter in the new field, and then the search word "adidas" belongs to the title information field area of the new field corresponding to 12345; the location of "clover" in the new field corresponding to 12345 is 0, which is smaller than the location 6 of the first delimiter in the new field, and the search term "clover" belongs to the brand information field area corresponding to 12345 for the new field.

Step 140, calculating the total correlation degree of the search word and the new field according to the field area to which each query word of the new field belongs; and the total correlation comprises a first correlation obtained by calculation according to the weight of the field area of the new field to which each query word belongs.

And according to the processing result of each participle of the search word in the steps, comprehensively calculating the interval relevance of whether the search word input by the user is related to the brand information field area or the title information field area, and calculating the total relevance of the search word and the new field. In practice, the following 4 categories can be distinguished:

a) the search terms match the brand and title simultaneously;

b) search terms only match brands;

c) the search term only matches the title;

d) the search terms do not match the brand and title.

Preferably, the first correlation is obtained by:

and step C1, dividing the character string length of each query word by the character string length of the field area to obtain the correlation degree of each query word and the field area.

In practice, the correlation degree between the query word and the brand interval of the brand information field area is calculated through length/length (brand information); calculating the title interval correlation degree of the query word and the title information field area through length/length; where length (participle) represents the character string length of the query word, length (brand information) represents the character string length of the brand information field area, and length (title information) represents the character string length of the title information field area.

And step C2, multiplying the relevance of each interval by the weight of the field area and adding to obtain the first relevance of the search term and the new field.

In practice, the total correlation of the product is defined as "weight of brand information field region length (participle)/length (brand information) + weight of title information field region length (participle)/length (title information)". For example, if the weight of the brand information field area and the weight of the title information field area are 0.3 and 0.7, respectively, the formula is: 0.3 length (participle)/length (brand information) +0.7 length (participle)/length (title information).

Such as for the previous example:

for 12345, i.e. 0.3 × 6/6+0.7 × 8/21 ═ 0.56

For 83635789, i.e. 0.3 × 6/6+0.7 × 0/21 ═ 0.3

Then, for the search word of "adidas clover", the total relevance of "1 folding and getting rid of selling of clover adidas sports shoes" is higher than that of "2011 folding and mailing of 8 folding and packing of autumn new-style shoes", that is, the total relevance of the commodity 12345 is higher than that of the commodity 83635789.

The relevance of the search term to the new field may also be calculated in other ways, which are not limited in this application.

The total correlation includes the first correlation, and obviously, other contents, such as sales volume, reputation, and the like, may also be referred to, and the total correlation is finally obtained.

And 150, outputting at least one target object corresponding to the new field to the user side based on the total correlation degree of each new field and the search word.

In practice, when at least one new field is output to the user side based on the total correlation degree between each new field and the search term, the following steps may be performed:

outputting at least one top-ranked target object; the target objects are ranked based on the total relevance of the respective new field to the search terms.

In the foregoing example, the correlation between each new field and the search term, that is, the correlation between the target object corresponding to each new field, that is, the product and the search term, is obtained, and then the product and the product information returned to the user end can be sorted and returned according to the total correlation. In practice, for a plurality of commodities in the same category, the final sorting can be performed according to the comprehensive conditions of commodity sales, browsing, attention and corresponding sellers, and then the final sorting is returned to the user side.

Referring to fig. 2, a schematic structural diagram of a search apparatus of the present application is shown, including:

a query term obtaining module 210, configured to obtain, for a search term input by a user, each query term corresponding to the search term;

the index word searching module 220 is configured to search, for each obtained query word, each index word corresponding to each query word in an index, where the index is constructed according to a field of a target object, and the field of the target object includes a new field formed by splicing a title information field region and an attribute information field region of the target object by a first separator;

a position confirmation module 230, configured to confirm whether a query word corresponding to each index word belongs to the title information field region or the attribute information field region in the new field according to the position of each index word in the new field and the position of the first separator in the new field;

a correlation calculation module 240, configured to calculate a total correlation between the search term and the new field according to the field area to which each query term of the new field belongs; the total correlation comprises a first correlation obtained by calculation according to the weight of the field area of the new field to which each query word belongs;

and the output module 250 is used for returning and outputting at least one target object corresponding to the new field based on the total relevance of each new field and the search word.

When at least one new field is output to the user side based on the total correlation degree of each new field and the search terms: outputting at least one top-ranked target object; the target objects are ranked based on the total relevance of the respective new field to the search terms.

Preferably, the splicing of the header information field area and the attribute information field area of the target object into a new field by the first delimiter is performed by:

the information acquisition module is used for reading a title information field area and an attribute information field area of the target object;

a character replacement module for replacing a character in the new field that is the same as the first delimiter with a blank character;

and the splicing module is used for splicing the replaced title information field area and the replaced attribute information field area into a new field through a first separator.

Preferably, the index is constructed according to each participle in the field by the following modules:

the target object corresponding module is used for corresponding the identification of each target object with the corresponding new field through a second separator;

the new field query word acquisition module is used for performing word segmentation operation on each new field;

and the index construction module is used for taking the participle obtained by the participle operation as an index word and corresponding the index word with the mark of each related target object and the position of the index word in each new field.

Preferably, the query term is confirmed to belong to the title information field area or the attribute information field area by the following modules:

the new field query module is used for querying a new field corresponding to the identification according to the corresponding relation between the index word and the identification of each relevant target object;

and the query word position confirmation module is used for comparing the position of the index word in the new field with the position of the first separator in the new field and confirming that the query word corresponding to the index word belongs to the title information field area or the attribute information field area.

Preferably, the total correlation is obtained by:

the field area relevancy calculation module is used for dividing the character string length of each query word by the character string length of the field area to obtain the relevancy of each query word and the field area;

and the total correlation degree calculation module is used for multiplying each correlation degree by the weight of the field area and adding the correlation degrees to obtain the total correlation degree of the search word and the new field.

Preferably, the present application further comprises:

and the intelligent engine is used for correcting the error of the search words input by the user by the intelligent error correction engine.

Preferably, the present application further comprises:

and the suggested word engine is used for returning suggested words according to the input words of the user.

For the system embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.

The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The above detailed description is given to a search method and a search system provided by the present application, and a specific example is applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiment is only used to help understand the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims (10)

1. A method of searching, comprising:

for a search word input by a user, acquiring each query word corresponding to the search word;

searching each index word corresponding to each query word in an index aiming at each obtained query word, wherein the index is constructed according to a field of a target object, and the field of the target object comprises a new field formed by splicing a title information field area and an attribute information field area of the target object through a first separator;

according to the position of each index word in the new field and the position of the first separator in the new field, determining whether the query word corresponding to the index word belongs to the title information field area or the attribute information field area in the new field;

calculating the total correlation degree of the search word and the new field according to the field area of the new field to which each query word belongs; the total correlation comprises a first correlation calculated according to the weight of the field area of the new field where each query word is located;

and returning the target object corresponding to at least one new field based on the total relevance of each new field and the search word.

2. The method according to claim 1, wherein the new field formed by splicing the header information field area and the attribute information field area of the target object by the first delimiter comprises the following steps:

reading a title information field area and an attribute information field area of the target object;

replacing the same character in the new field as the first delimiter with a blank character;

and splicing the replaced header information field area and the attribute information field area into a new field through a first separator.

3. The method of claim 1, wherein constructing an index from the tokens in the field is performed by:

corresponding the identification of each target object with the corresponding new field through a second separator;

performing word segmentation operation on each new field;

and taking the participles obtained by the participle operation as index words, and corresponding the index words with the identifications of the related target objects and the positions of the index words in the new fields.

4. The method of claim 3, wherein the query word is confirmed to belong to a title information field area or an attribute information field area by:

inquiring a new field corresponding to the identification according to the corresponding relation between the index word and the identification of each related target object;

and comparing the position of the index word in the new field with the position of the first separator in the new field, and confirming that the query word corresponding to the index word belongs to the title information field area or the attribute information field area.

5. Method according to one of the claims 1, characterized in that the first degree of correlation is obtained by:

dividing the character string length of each query word by the character string length of the field area to obtain the interval correlation degree of each query word and the field area;

and multiplying the correlation degrees by the weight of the field area, and adding to obtain the first correlation degree of the search term and the new field.

6. The method of claim 1, wherein the search term comprises:

taking keywords input by a user as search words;

or, one of the suggested words returned according to the input words of the user selected by the user is used as a search word; the suggested words are obtained by extracting the click relation between the input words input by the user and the corresponding results which are counted in advance.

7. The method of claim 1, wherein for a search term input by a user, obtaining a query term of the search term comprises:

and correcting the error of the search word input by the user by an intelligent error correction engine.

8. The method of claim 1, wherein:

the target object comprises a commodity; the attribute information includes brand information of the commodity.

9. The method of claim 1, wherein when outputting at least one new field to the user side based on the total correlation between each new field and the search term:

outputting at least one top-ranked target object; the target objects are ranked based on the total relevance of the respective new field to the search terms.

10. A search apparatus, comprising:

the query term acquisition module is used for acquiring each query term corresponding to a search term input by a user;

the index word searching module is used for searching each index word corresponding to each query word in the index, the index is constructed according to the field of the target object, and the field of the target object comprises a new field formed by splicing the title information field area and the attribute information field area of the target object through a first separator;

the position confirmation module is used for confirming whether the query word corresponding to the index word belongs to the title information field area or the attribute information field area in the new field according to the position of each index word in the new field and the position of the first separator in the new field;

the relevancy calculation module is used for calculating the total relevancy between the search word and the new field according to the field area to which each query word of the new field belongs; the total correlation comprises a first correlation calculated according to the weight of the field area of the new field where each query word is located;

and the output module is used for returning the target object corresponding to at least one new field based on the total correlation degree of each new field and the search word.