阅读说明：本技术 多需求搜索请求的划分方法、装置、电子设备及存储介质 (Division methods, device, electronic equipment and the storage medium of more demand searching requests ) 是由 汪洪臣 徐昇 李裕东 于 2019-08-28 设计创作，主要内容包括：本申请公开了多需求搜索请求的划分方法、装置、电子设备及存储设备,涉及搜索领域。具体实现方案为：接收多需求搜索请求,所述多需求搜索请求包括至少两个搜索需求；将所述多需求搜索请求输入预先设置的解析模型,利用所述解析模型的输出结果确定至少两个单需求搜索请求,每个所述单需求搜索请求对应所述多需求搜索请求的一个搜索需求。本申请能够划分出多需求搜索请求中包含的多个搜索需求,便于得到较佳的搜索结果。(This application discloses the division methods of more demand searching requests, device, electronic equipment and storage equipment, are related to search field.Specific implementation are as follows: receive more demand searching requests, more demand searching requests include at least two search needs；More demand searching requests are inputted into pre-set analytic modell analytical model, determine that at least two single demand searching requests, each single demand searching request correspond to a search need of more demand searching requests using the output result of the analytic modell analytical model.The application can mark off the multiple search needs for including in more demand searching requests, convenient for obtaining preferable search result.)

1. a kind of division methods of more demand searching requests characterized by comprising

More demand searching requests are received, more demand searching requests include at least two search needs；

More demand searching requests are inputted into pre-set analytic modell analytical model；

At least two single demand searching requests are determined using the output result of the analytic modell analytical model, and each single demand search is asked Seek a search need of corresponding more demand searching requests.

2. the method according to claim 1, wherein further include:

Show at least two single demands searching request；

Receive the search instruction to the single demand searching request；

According to described search instruction execution to the search operation of single demand searching request, and show search result.

3. method according to claim 1 or 2, which is characterized in that the analytic modell analytical model is the first analytic modell analytical model；

The output result using the analytic modell analytical model determines that at least two single demand searching requests include:

The generic items and at least two particular terms of the first analytic modell analytical model output are obtained, the generic items are that more demands are searched The text items that all search needs of rope request all include, the particular term are that the single search of more demand searching requests needs Seek the text items for including；

The generic items are spliced with each particular term respectively, obtain at least two single demand searching requests.

4. according to the method described in claim 3, it is characterized in that, described preset more demand searching request inputs Analytic modell analytical model include:

Identify classification belonging to more demand searching requests；

In the case where more demand searching request generics can not be identified, by more demand searching requests input described the One analytic modell analytical model.

5. according to the method described in claim 3, it is characterized in that, first analytic modell analytical model is sequence labelling model.

6. method according to claim 1 or 2, which is characterized in that the analytic modell analytical model is the second analytic modell analytical model；Described Two analytic modell analytical models include multiple slot position matching templates, and each slot position matching template is made of at least two slot positions, Mei Gesuo State the corresponding output rule of slot position matching template；

The output result using the analytic modell analytical model determines at least two single demand searching requests, comprising:

More demand searching requests are directed to each slot position matching template in second analytic modell analytical model respectively to be filled；

For successful slot position matching template is filled, according to the filling result of each slot position and the slot position matching template pair The output rule answered, determines at least two single demand searching requests.

7. according to the method described in claim 6, it is characterized in that, second analytic modell analytical model is to ask with more demand search Seek the affiliated corresponding analytic modell analytical model of classification；

It is described to include: by the pre-set analytic modell analytical model of more demand searching requests inputs

Identify classification belonging to more demand searching requests；

In the case where more demand searching request generics can be identified, corresponding second parsing of the classification is determined Model；

More demand searching requests are inputted into the second analytic modell analytical model determined.

8. a kind of analytic modell analytical model training method characterized by comprising

More demand searching requests are inputted into the first analytic modell analytical model；

Obtain the generic items of the first analytic modell analytical model output and the predicted value of particular term；The generic items are that more demands are searched The text items that all search needs of rope request all include, the particular term are that the single search of more demand searching requests needs Seek the text items for including；

The generic items and the predicted value of particular term and the true value of generic items and particular term are compared, according to comparing result Adjust the parameter of first analytic modell analytical model.

9. according to the method described in claim 8, it is characterized by further comprising:

Meet in the comparing result of the generic items and the predicted value of particular term and the generic items and the true value of particular term pre- If it is required that when, terminate to the training process of first analytic modell analytical model.

10. method according to claim 8 or claim 9, which is characterized in that first analytic modell analytical model is sequence labelling model.

11. according to the method described in claim 10, it is characterized in that, the sequence labelling model is statistical model or nerve net Network model；Wherein,

The statistical model is hidden Markov model or condition random field algorithm model；

The neural network model is shot and long term memory network-condition random field algorithm model.

12. a kind of dividing device of more demand searching requests characterized by comprising

Request receiving module, for receiving more demand searching requests, more demand searching requests include that at least two search need It asks；

First input module, for more demand searching requests to be inputted pre-set analytic modell analytical model；

Determining module determines at least two single demand searching requests, Mei Gesuo for the output result using the analytic modell analytical model State the search need that single demand searching request corresponds to more demand searching requests.

13. device according to claim 12, which is characterized in that further include:

Display module, for showing at least two single demands searching request；

Command reception module, for receiving the search instruction to the single demand searching request；

Search module for the search operation according to described search instruction execution to single demand searching request, and shows search knot Fruit.

14. method according to claim 12 or 13, which is characterized in that the analytic modell analytical model is the first analytic modell analytical model；

The determining module, it is described logical for obtaining the generic items and at least two particular terms of the first analytic modell analytical model output The text items for all including with all search needs that item is more demand searching requests, the particular term are that more demands are searched The text items that the single search need of rope request includes；The generic items are spliced with each particular term respectively, obtain to Few two single demand searching requests.

15. method according to claim 12 or 13, which is characterized in that the analytic modell analytical model is the second analytic modell analytical model；Institute Stating the second analytic modell analytical model includes multiple slot position matching templates, and each slot position matching template is made of at least two slot positions, often A corresponding output rule of the slot position matching template；

The determining module, each slot position for being directed to more demand searching requests respectively in second analytic modell analytical model Matching template is filled；For filling successful slot position matching template, according to the filling result of each slot position and described The corresponding output rule of slot position matching template, determines at least two single demand searching requests.

16. a kind of analytic modell analytical model training device characterized by comprising

Second input module, for more demand searching requests to be inputted the first analytic modell analytical model；

Module is obtained, for obtaining the generic items of the first analytic modell analytical model output and the predicted value of particular term；The generic items The text items that all search needs for more demand searching requests all include, the particular term are that more demand search are asked The text items that the single search need asked includes；

Module is adjusted, for carrying out pair the true value of the predicted value of the generic items and particular term and generic items and particular term Than adjusting the parameter of first analytic modell analytical model according to comparing result.

17. device according to claim 16, which is characterized in that further include:

Control module, pair of the true value for predicted value and the generic items and particular term in the generic items and particular term When meeting preset requirement than result, terminate the training process to first analytic modell analytical model.

18. device according to claim 16 or 17, which is characterized in that first analytic modell analytical model is sequence labelling mould Type.

19. device according to claim 18, which is characterized in that the sequence labelling model is statistical model or nerve net Network model；Wherein,

The statistical model is hidden Markov model or condition random field algorithm model；

The neural network model is shot and long term memory network-condition random field algorithm model.

20. a kind of electronic equipment characterized by comprising

At least one processor；And

The memory being connect at least one described processor communication；Wherein,

The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one It manages device to execute, so that at least one described processor is able to carry out method of any of claims 1-11.

21. a kind of non-transitory computer-readable storage medium for being stored with computer instruction, which is characterized in that the computer refers to It enables for making the computer perform claim require method described in any one of 1-11.

Technical field

This application involves a kind of computer field more particularly to a kind of search fields.

Background technique

More demand searching requests (query) are the searching requests for containing at least two search need or search intention.For example, Following sentence is demand searching request more than one:

" searching the near synonym and antonym put things right once and for all ".

In above-mentioned more demand searching requests, 2 search needs are contained, it may be assumed that 1) near synonym put things right once and for all；2) labor The antonym to escape forever.

Existing search technique can not accurately mark off the multiple search needs for including in more demand searching requests, cause Search result is poor.

Summary of the invention

The embodiment of the present application proposes the partitioning method and device and a kind of analytic modell analytical model instruction of a kind of more demand searching requests Practice method and device, at least to solve the above technical problem in the prior art.

In a first aspect, the embodiment of the present application provides a kind of division methods of more demand searching requests, comprising:

More demand searching requests are received, more demand searching requests include at least two search needs；

More demand searching requests are inputted into pre-set analytic modell analytical model；

At least two single demand searching requests are determined using the output result of the analytic modell analytical model, and each single demand is searched One search need of the corresponding more demand searching requests of rope request.

In one embodiment, further includes:

Show at least two single demands searching request；

Receive the search instruction to the single demand searching request；

According to described search instruction execution to the search operation of single demand searching request, and show search result.

In one embodiment, the analytic modell analytical model is the first analytic modell analytical model；

The output result using the analytic modell analytical model determines that at least two single demand searching requests include:

The generic items and at least two particular terms of the first analytic modell analytical model output are obtained, the generic items need more to be described The text items for asking all search needs of searching request all to include, the particular term are individually searching for more demand searching requests The text items that rope demand includes；

The generic items are spliced with each particular term respectively, obtain at least two single demand searching requests.

In one embodiment, described to include: by the pre-set analytic modell analytical model of more demand searching requests inputs

Identify classification belonging to more demand searching requests；

In the case where more demand searching request generics can not be identified, more demand searching requests are inputted into institute State the first analytic modell analytical model.

In one embodiment, first analytic modell analytical model is sequence labelling model.

In one embodiment, the analytic modell analytical model is the second analytic modell analytical model；Second analytic modell analytical model includes multiple Slot position matching template, each slot position matching template are made of at least two slot positions, and each slot position matching template is corresponding One output rule；

The output result using the analytic modell analytical model determines at least two single demand searching requests, comprising:

More demand searching requests are directed to each slot position matching template in second analytic modell analytical model respectively to carry out Filling；

For successful slot position matching template is filled, mould is matched according to the filling result of each slot position and the slot position The corresponding output rule of plate, determines at least two single demand searching requests.

In one embodiment, second analytic modell analytical model is corresponding with classification belonging to more demand searching requests Analytic modell analytical model；

It is described to include: by the pre-set analytic modell analytical model of more demand searching requests inputs

Identify classification belonging to more demand searching requests；

In the case where more demand searching request generics can be identified, the classification corresponding second is determined Analytic modell analytical model；

More demand searching requests are inputted into the second analytic modell analytical model determined.

Second aspect, the embodiment of the present application provide a kind of analytic modell analytical model training method, comprising:

More demand searching requests are inputted into the first analytic modell analytical model；

Obtain the generic items of the first analytic modell analytical model output and the predicted value of particular term；The generic items need more to be described The text items for asking all search needs of searching request all to include, the particular term are individually searching for more demand searching requests The text items that rope demand includes；

The generic items and the predicted value of particular term and the true value of generic items and particular term are compared, according to comparison As a result the parameter of first analytic modell analytical model is adjusted.

In one embodiment, further includes:

It is full in the comparing result of the generic items and the predicted value of particular term and the generic items and the true value of particular term When sufficient preset requirement, terminate the training process to first analytic modell analytical model.

In one embodiment, first analytic modell analytical model is sequence labelling model.

In one embodiment, the sequence labelling model is statistical model or neural network model；Wherein,

The statistical model is hidden Markov model or condition random field algorithm model；

The neural network model is shot and long term memory network-condition random field algorithm model.

The third aspect, the embodiment of the present application provide a kind of dividing device of more demand searching requests, comprising:

Request receiving module, for receiving more demand searching requests, more demand searching requests are searched including at least two Rope demand；

First input module, for more demand searching requests to be inputted pre-set analytic modell analytical model；

Determining module determines at least two single demand searching requests for the output result using the analytic modell analytical model, often A single demand searching request corresponds to a search need of more demand searching requests.

In one embodiment, further includes:

Display module, for showing at least two single demands searching request；

Command reception module, for receiving the search instruction to the single demand searching request；

Search module for the search operation according to described search instruction execution to single demand searching request, and shows and searches Hitch fruit.

In one embodiment, the analytic modell analytical model is the first analytic modell analytical model；

The determining module, for obtaining the generic items and at least two particular terms of the first analytic modell analytical model output, institute The text items that all search needs that generic items are more demand searching requests all include are stated, the particular term needs more to be described The text items for asking the single search need of searching request to include；The generic items are spliced with each particular term respectively, are obtained To at least two single demand searching requests.

In one embodiment, the analytic modell analytical model is the second analytic modell analytical model；Second analytic modell analytical model includes multiple Slot position matching template, each slot position matching template are made of at least two slot positions, and each slot position matching template is corresponding One output rule；

The determining module, it is each in second analytic modell analytical model for being directed to more demand searching requests respectively Slot position matching template is filled；For filling successful slot position matching template, according to the filling result of each slot position and The corresponding output rule of the slot position matching template, determines at least two single demand searching requests.

Fourth aspect, the embodiment of the present application provide a kind of analytic modell analytical model training device, comprising:

Second input module, for more demand searching requests to be inputted the first analytic modell analytical model；

Module is obtained, for obtaining the generic items of the first analytic modell analytical model output and the predicted value of particular term；It is described logical The text items for all including with all search needs that item is more demand searching requests, the particular term are that more demands are searched The text items that the single search need of rope request includes；

Module is adjusted, for carrying out the true value of the predicted value of the generic items and particular term and generic items and particular term Comparison, the parameter of first analytic modell analytical model is adjusted according to comparing result.

In one embodiment, further includes:

Control module, the true value for predicted value and the generic items and particular term in the generic items and particular term Comparing result when meeting preset requirement, terminate the training process to first analytic modell analytical model.

In one embodiment, first analytic modell analytical model is sequence labelling model.

In one embodiment, the sequence labelling model is statistical model or neural network model；Wherein,

The statistical model is hidden Markov model or condition random field algorithm model；

The neural network model is shot and long term memory network-condition random field algorithm model.

5th aspect, the embodiment of the present application provide a kind of electronic equipment, comprising:

At least one processor；And

The memory being connect at least one processor communication；Wherein,

Memory is stored with the instruction that can be executed by least one processor, and instruction is executed by least one processor, with It is able to carry out at least one processor any in the division methods or analytic modell analytical model training method of above-mentioned more demand searching requests The method of item.

6th aspect, the embodiment of the present application provide a kind of non-instantaneous computer-readable storage for being stored with computer instruction Medium, the division methods or analytic modell analytical model training which is used to that the computer to be made to execute above-mentioned more demand searching requests The method of any one of method.

One embodiment in above-mentioned application has the following advantages that or the utility model has the advantages that the application is defeated by more demand searching requests Enter pre-set parsing module, determine corresponding at least two single demands searching request using the output result of parsing module, To realize the division to more demand searching requests.Further, the embodiment of the present application can also show single demand search phrase Sentence receives the search instruction for being directed to single demand searching request, and executes the instruction, to realize to each single demand searching request Search.The parsing module of the application can have two kinds of forms of the first parsing module and the second parsing module, wherein the first parsing Module can be adapted for the more demand searching requests that can not determine generic, and the second parsing module can be adapted for can not be true More demand searching requests of generic are made, the second different parsing modules corresponds to different classifications.For the first parsing mould Block, the embodiment of the present application can use the generic items of the first parsing module output and particular term generates single demand searching request；It is right In the second parsing module, the embodiment of the present application can use the filling result and matching of the slot position of the second parsing module output Successfully the corresponding output rule of slot position matching template determines single demand searching request, is searched for realizing to different classes of more demands The accurate division of request.

Other effects possessed by above-mentioned optional way are illustrated hereinafter in conjunction with specific embodiment.

Detailed description of the invention

Attached drawing does not constitute the restriction to the application for more fully understanding this programme.Wherein:

Fig. 1 is a kind of division methods implementation process schematic diagram one of more demand searching requests of the application；

Fig. 2 is a kind of division methods implementation process schematic diagram two of more demand searching requests of the application；

It is true using the output result of analytic modell analytical model in division methods of the Fig. 3 for a kind of more demand searching requests of the application The implementation process schematic diagram one of order demand searching request；

Fig. 4 is to input more demand searching requests preparatory in a kind of division methods of more demand searching requests of the application The implementation process schematic diagram one of the analytic modell analytical model of setting；

It is true using the output result of analytic modell analytical model in division methods of the Fig. 5 for a kind of more demand searching requests of the application The implementation process schematic diagram two of order demand searching request；

Fig. 6 is to input more demand searching requests preparatory in a kind of division methods of more demand searching requests of the application The implementation process schematic diagram two of the analytic modell analytical model of setting；

Fig. 7 is a kind of analytic modell analytical model training method implementation process schematic diagram of the application；

Fig. 8 is a kind of dividing device structural schematic diagram one of more demand searching requests of the application；

Fig. 9 is a kind of dividing device structural schematic diagram two of more demand searching requests of the application；

Figure 10 is a kind of analytic modell analytical model training device structural schematic diagram one of the application；

Figure 11 is a kind of analytic modell analytical model training device structural schematic diagram two of the application；

Figure 12 is for the division methods for realizing more demand searching requests of the embodiment of the present application or analytic modell analytical model training side The block diagram of the electronic equipment of method.

Specific embodiment

It explains below in conjunction with exemplary embodiment of the attached drawing to the application, including the various of the embodiment of the present application Details should think them only exemplary to help understanding.Therefore, those of ordinary skill in the art should recognize It arrives, it can be with various changes and modifications are made to the embodiments described herein, without departing from the scope and spirit of the present application.Together Sample, for clarity and conciseness, descriptions of well-known functions and structures are omitted from the following description.

The application proposes a kind of division methods of more demand searching requests, and Fig. 1 is that a kind of more demand search of the application are asked The division methods implementation process schematic diagram one asked, comprising:

Step S101: receiving more demand searching requests, and more demand searching requests include at least two search needs；

Step S102: more demand searching requests are inputted into pre-set analytic modell analytical model；

Step S103: at least two single demand searching requests, each single demand are determined using the output result of analytic modell analytical model Searching request corresponds to a search need of more demand searching requests.

Fig. 2 is a kind of division methods implementation process schematic diagram two of more demand searching requests of the application.As shown in Fig. 2, Can also include: after step s 103

Step S204: at least two single demand searching requests are shown；

Step S205: the search instruction to single demand searching request is received；

Step S206: the search operation to single demand searching request is executed according to search instruction, and shows search result.

The embodiment of the present application can be applied to search engine.In a kind of possible embodiment, search engine, which receives, to be come From more demand searching requests of search client, after determining corresponding at least two single demands searching request, can recommend These single demand searching requests are shown in list.Later, user can click one or two for wishing to search in search client A above single demand searching request, search client generates the search instruction to the single demand searching request, and is sent to search Engine.Search engine receives the search instruction from search client, executes search operation and shows search result.Alternatively, searching Index, which is held up, can directly search and ask according to the search of these single demands after determining corresponding at least two single demands searching request It asks and scans for, show each single demand searching request and its corresponding search result.

In a kind of possible embodiment, the analytic modell analytical model is the first analytic modell analytical model；

It is true using the output result of analytic modell analytical model in division methods of the Fig. 3 for a kind of more demand searching requests of the application The implementation process schematic diagram one of order demand searching request, comprising:

Step S301: obtaining the generic items and at least two particular terms of the output of the first analytic modell analytical model, and generic items are more demands The text items that all search needs of searching request all include, particular term are that the single search need of more demand searching requests includes Text items；

Step S302: generic items are spliced with each particular term respectively, obtain at least two single demand searching requests.

In a kind of possible embodiment, the first analytic modell analytical model, which can be applied to processing, can not identify classification more and need Seek searching request.Alternatively, the first analytic modell analytical model also can be applied to handle arbitrary more demand searching requests；Therefore, the application Embodiment can not identify the generic of more demand searching requests, but directly by the first parsing of more demand searching requests input Model, by the output of the first analytic modell analytical model to the parsing result of more demand searching requests.

Aforementioned categories can be specially hang down class, such as words class, variety class etc., and each vertical class can correspond to different necks Domain.It can realize that classification identifies in more demand searching requests by the way of keyword using identifying.The embodiment of the present application can adopt Identify classification belonging to more demand searching requests with classifier.The classifier can use convolutional neural networks (Convolutional Neural Networks, CNN) Lai Shixian.

Fig. 4 is to input more demand searching requests preparatory in a kind of division methods of more demand searching requests of the application The implementation process schematic diagram one of the analytic modell analytical model of setting, comprising:

Step S401: identify classification belonging to more demand searching requests；

Step S402: in the case where more demand searching request generics can not be identified, by more demand searching requests Input the first analytic modell analytical model.

In a kind of possible embodiment, the first analytic modell analytical model is sequence labelling model.

If using word for granularity of division, more demand searching requests can be divided into continuous multiple words.For example, will " search put things right once and for all near synonym and antonym " be divided into " " search/mono-/under/mono-/labor/forever/ease// close/justice/word/and/ Instead/justice/word/".One or continuous multiple words can form a text items (term).

For present inventor by investigation discovery, most of more demand searching requests meet following two feature:

1) term in more demand searching requests belongs to one of following three kinds:

The first: generic items (common term can be indicated with C), i.e., each search need of more demand searching requests The item for including；

Second: a search of particular term (special term can be indicated with S), i.e., more demand searching requests needs Seek the item for including；

The third: other (other term), to the item of the not no information gain of each search need, more mainly Spoken word, stop words, conjunction etc..

2) word in each term continuously occurs in more demand searching requests.

Based on These characteristics, the application can use sequence labelling model as the first above-mentioned analytic modell analytical model, such as can To use BIO sequence labelling model.First analytic modell analytical model can determine the generic items for including in more demand searching requests and spy The banner word (Begin can be indicated with B) and middle word (Inside can be indicated with I) of different item, and be both not belonging to general Item, other words for being also not belonging to particular term (Outside can be indicated with O).

For example, for more demand searching requests " searching the near synonym and antonym put things right once and for all ", BIO sequence labelling mould Type can determine the label of each word in the sentence, as follows:

Search --- O；

One --- O；

Under --- O；

One --- C_B；

Labor --- C_I；

Forever --- C_I；

Ease --- C_I；

--- O；

Closely --- S_B；

Justice -- S_I；

Word --- S_I；

With --- O；

Instead --- S_B；

Justice --- S_I；

Word --- S_I；

According to above-mentioned label, a generic items for including in above-mentioned more demand searching requests, i.e. " a Lao Yong can be determined Ease "；And two particular terms, i.e. " near synonym " and " antonym ".Generic items are spliced with each particular term respectively, it can To all single demand searching requests, i.e. " near synonym once and for all " and " antonym once and for all ".

The embodiment of the present application can be pre-designed special second analytic modell analytical model for some vertical classes, to guarantee corresponding There are higher accuracy rate and coverage rate in class of hanging down.If more demand searching requests are not belonging to any one class of hanging down, can use Above-mentioned first analytic modell analytical model is handled.

In a kind of possible embodiment, the second analytic modell analytical model may include multiple slot position matching templates and each The corresponding output rule of slot position matching template.Wherein, slot position matching template is formed by connecting by least two slot positions, and a slot position can With a corresponding matching rule or a dictionary.Second analytic modell analytical model is suitable for the more of more fixed mode or rule Demand searching request, such as more demand searching requests comprising words class demand.

It is true using the output result of analytic modell analytical model in division methods of the Fig. 5 for a kind of more demand searching requests of the application The implementation process schematic diagram two of fixed at least two single demand searching requests, comprising:

Step S501: more demand searching requests are directed to each slot position matching template in the second analytic modell analytical model respectively and are carried out Filling；

Step S502: for filling successful slot position matching template, according to the filling result of each slot position and described The corresponding output rule of slot position matching template, determines at least two single demand searching requests.

Multiple second analytic modell analytical models can be used in the embodiment of the present application, and each second analytic modell analytical model corresponds to different more demands Searching request generic.Correspondingly, before using the second analytic modell analytical model, need to determine more affiliated classes of demand searching request first Not.Fig. 6 is to input more demand searching requests pre-set in a kind of division methods of more demand searching requests of the application The implementation process schematic diagram two of analytic modell analytical model, comprising:

Step S601: identify classification belonging to more demand searching requests；

Step S602: in the case where more demand searching request generics can be identified, determine that the category is corresponding Second analytic modell analytical model；

Step S603: more demand searching requests are inputted to the second analytic modell analytical model determined.

For example, the slot position that a slot position matching template includes is as follows:

[spoken word] [core word] [stop words] [demand word 1] [stop words] [demand word 2].Foregoing teachings indicate slot position It include aforesaid plurality of continuous slot position with template.

Wherein, each slot position definition can be such as the following table 1:

Table 1

The corresponding output rule of the slot position matching template is as follows:

[core word] [demand word 1]；[core word] [demand word 2].

Above-mentioned output Rule Expression can generate 2 single demand searching requests, i.e., need comprising core word and the single of demand word 1 Searching request is sought, and the single demand searching request comprising core word and demand word 2.

Still by taking above-mentioned more demand searching requests " searching the near synonym and antonym put things right once and for all " as an example, which is searched Rope request can successfully fill above-mentioned slot position matching template, the filling result of each slot position are as follows:

Spoken word --- " searching "；

Core word --- " once and for all "；

Stop words --- " "；

Demand word 1 --- " near synonym "；

Stop words --- "and"；

Demand word 2 --- " antonym ".

In this way, according to above-mentioned filling result and output rule, by core word (" once and for all ") and (" the nearly justice of demand word 1 Word ") it is spliced into a single demand searching request, i.e. " near synonym once and for all ", and by core word (" once and for all ") and will need Word 2 (" antonym ") is asked to be spliced into another single demand searching request, i.e. " antonym once and for all ".

" me is helped to search the meaning to waste natural products and near synonym and hello prolixity in a great rush with another more demand searching request Near synonym " for, which can successfully fill following slot position matching template:

[spoken word] [core word 1] [stop words] [demand word 1] [stop words] [core word 2] [stop words] [demand word 2] [stop words] [core word 3] [stop words] [demand word 3]；

The filling result of each slot position are as follows:

Spoken word --- " me is helped to search "；

Core word 1 --- " wasting natural products "；

Stop words --- " "；

Demand word 1 --- " meaning "；

Stop words --- "and"；

Core word 2 --- " in a great rush "；

Stop words --- " "；

Demand word 2 --- " near synonym "；

Stop words --- " and "；

Core word 3 --- " hello prolixity "；

Stop words --- " "；

Demand word 3 --- " near synonym ".

The corresponding output rule of the slot position matching template is as follows:

[core word 1] [demand word 1]；

[core word 2] [demand word 2]；

[core word 3] [demand word 3].

According to above-mentioned filling result and output rule, core word 1 (" wasting natural products ") and demand word 1 (" meaning ") are spliced At a single demand searching request, i.e., " waste natural products the meaning ".Core word 2 (" in a great rush ") and demand word 2 (" near synonym ") are spelled It is connected into second single demand searching request, i.e. " near synonym in a great rush ".By core word 3 (" hello prolixity ") and demand word 3 (" near synonym ") It is spliced into third single demand searching request, i.e. " hello prolixity near synonym ".

The division methods of more demand searching requests of the embodiment of the present application proposition are described above.

The application also proposes a kind of analytic modell analytical model training method, for training the division methods of above-mentioned more demand searching requests Used in the first analytic modell analytical model.Fig. 7 is a kind of analytic modell analytical model training method implementation process schematic diagram of the application, comprising:

Step S701: more demand searching requests are inputted into the first analytic modell analytical model；

Step S702: the generic items of the first analytic modell analytical model output and the predicted value of particular term are obtained；Generic items are more demands The text items that all search needs of searching request all include, particular term are that the single search need of more demand searching requests includes Text items；

Step S703: generic items and the predicted value of particular term and the true value of generic items and particular term are compared, root The parameter of first analytic modell analytical model is adjusted according to comparing result.

Wherein, the true value of generic items and particular term can be by being manually labeled.

In a kind of possible embodiment, in the true of generic items and the predicted value of particular term and generic items and particular term When the comparing result of value meets preset requirement, terminate the training process to the first analytic modell analytical model, it is believed that the first analytic modell analytical model Training is completed.

The first analytic modell analytical model that training is completed can export the generic items for including in more demand searching requests and particular term Banner word and middle word, and be both not belonging to generic items or be not belonging to other words of particular term, and obtained according to foregoing teachings more The generic items and particular term for including in demand searching request.The related content of first analytic modell analytical model and above-mentioned more demand searching requests Division methods in the content introduced it is consistent, details are not described herein.

In a kind of possible embodiment, the first analytic modell analytical model can be sequence labelling model, and be specifically as follows BIO Sequence labelling model.

Sequence labelling model can be with statistical model or neural network model.

Wherein, statistical model can using hidden Markov model (Hidden Markov Model, HMM) or condition with Airport algorithm (Conditional Random Field algorithm, CRF) model.Statistical model can be adapted for training sample The lesser situation of this quantity.

Neural network model can use shot and long term memory network (Long Short-Term Memory, LSTM)-condition Random field algorithm (CRF) model.Neural network model can be adapted for the larger situation of training samples number.

The embodiment of the present application also proposes a kind of dividing device of more demand searching requests, and Fig. 8, which is that one kind of the application, to be needed The dividing device structural schematic diagram one of searching request is sought, the dividing device 800 of more demand searching requests shown in Fig. 8 includes:

Request receiving module 801, for receiving more demand searching requests, more demand searching requests include at least two Search need；

First input module 802, for more demand searching requests to be inputted pre-set analytic modell analytical model；

Determining module 803 determines at least two single demand searching requests using the output result using the analytic modell analytical model, Each single demand searching request corresponds to a search need of more demand searching requests.

Fig. 9 is a kind of dividing device structural schematic diagram two of more demand searching requests of the application, more demands shown in Fig. 9 The dividing device 900 of searching request includes:

Request receiving module 801, the first input module 802, determining module 803, display module 904, command reception module 905 and search module 906；

Wherein, request receiving module 801, the first input module 802 and determining module 803 are corresponding in above-described embodiment Model function is identical, repeats no more.

Display module 904, for showing at least two single demands searching request.

Command reception module 905, for receiving the search instruction to the single demand searching request；

Search module 906 for the search operation according to described search instruction execution to single demand searching request, and is shown Search result.

In a kind of possible embodiment, the analytic modell analytical model is the first analytic modell analytical model；

The determining module 803, for obtaining the generic items and at least two particular terms of the first analytic modell analytical model output, The generic items are the text items that all search needs of more demand searching requests all include, and the particular term is described more The text items that the single search need of demand searching request includes；The generic items are spliced with each particular term respectively, Obtain at least two single demand searching requests.

In a kind of possible embodiment, the analytic modell analytical model is the second analytic modell analytical model；The second analytic modell analytical model packet Multiple slot position matching templates are included, each slot position matching template is made of at least two slot positions, and each slot position matches mould Plate corresponds to an output rule；

The determining module 803, for more demand searching requests to be directed to respectively in second analytic modell analytical model Each slot position matching template is filled；For successful slot position matching template is filled, according to the filling knot of each slot position Fruit and the corresponding output rule of the slot position matching template, determine at least two single demand searching requests.

The embodiment of the present application also proposes that a kind of analytic modell analytical model training device, Figure 10 are a kind of analytic modell analytical model training of the application Apparatus structure schematic diagram one, analytic modell analytical model training device 1000 shown in Fig. 10 include:

Second input module 1001, for more demand searching requests to be inputted the first analytic modell analytical model；

Module 1002 is obtained, for obtaining the generic items of the first analytic modell analytical model output and the predicted value of particular term；Institute The text items that all search needs that generic items are more demand searching requests all include are stated, the particular term needs more to be described The text items for asking the single search need of searching request to include；

Module 1003 is adjusted, for by the true value of the predicted value of the generic items and particular term and generic items and particular term It compares, the parameter of first analytic modell analytical model is adjusted according to comparing result.

Figure 11 is a kind of analytic modell analytical model training device structural schematic diagram two of the application, the training of analytic modell analytical model shown in Figure 11 Device 1100 includes: the second input module 1001, obtains module 1002, adjustment module 1003 and control module 1104；

Wherein, the second input module 1001, acquisition module 1002 and adjustment module 1003 are corresponding in above-described embodiment Model function is identical, repeats no more；

Control module 1104, in the generic items and particular term predicted value and the generic items and particular term it is true When the comparing result of real value meets preset requirement, terminate the training process to first analytic modell analytical model.

In a kind of possible embodiment, the first analytic modell analytical model is sequence labelling model.

In a kind of possible embodiment, sequence labelling model is statistical model or neural network model；Wherein,

Statistical model is hidden Markov model or condition random field algorithm model；

Neural network model is shot and long term memory network-condition random field algorithm model.

According to an embodiment of the present application, present invention also provides a kind of electronic equipment and a kind of readable storage medium storing program for executing.

It as shown in figure 12, is according to the training of the division methods or analytic modell analytical model of more demand searching requests of the embodiment of the present application The block diagram of the electronic equipment of method.Electronic equipment is intended to indicate that various forms of digital computers, such as, laptop computer, Desktop computer, workbench, personal digital assistant, server, blade server, mainframe computer and other suitable meter Calculation machine.Electronic equipment also may indicate that various forms of mobile devices, such as, personal digital assistant, cellular phone, intelligence electricity Words, wearable device and other similar computing devices.Component, their connection and relationship shown in this article and they Function is merely exemplary, and is not intended to limit the realization of the application that is described herein and/or requiring.

As shown in figure 12, which includes: one or more processors 1201, memory 1202, and for connecting Connect the interface of each component, including high-speed interface and low-speed interface.All parts are interconnected using different bus, and can be with It is installed on public mainboard or installs in other ways as needed.Processor can be to the finger executed in electronic equipment Order is handled, including storage in memory or on memory (such as, to be coupled to and connect in external input/output device Mouthful display equipment) on show graphic user interface (Graphical User Interface, GUI) graphical information finger It enables.In other embodiments, if desired, by multiple processors and/or multiple bus and multiple memories and multiple can deposit Reservoir is used together.It is also possible to connect multiple electronic equipments, each equipment provides the necessary operation in part (for example, as clothes Business device array, one group of blade server or multicomputer system).In Figure 12 by taking a processor 1201 as an example.

Memory 1202 is non-transitory computer-readable storage medium provided herein.Wherein, memory stores There is the instruction that can be executed by least one processor, so that at least one processor executes more demands search provided herein The division methods or analytic modell analytical model training method of request.The non-transitory computer-readable storage medium storage computer of the application refers to It enables, the division methods or parsing mould which is used to that computer to be made to execute more demand searching requests provided herein Type training method.

Memory 1202 be used as a kind of non-transitory computer-readable storage medium, can be used for storing non-instantaneous software program, Non-instantaneous computer executable program and module, such as the division methods or solution of more demand searching requests in the embodiment of the present application Corresponding program instruction/the module of model training method is analysed (for example, attached request receiving module shown in Fig. 8 801, first inputs mould Block 802 and determining module 803 or attached second input module 1001 shown in Fig. 10 obtain module 1002 and adjustment module 1003).Non-instantaneous software program, instruction and the module that processor 1201 is stored in memory 1202 by operation, thus The various function application and data processing of execute server realize more demand searching requests in above method embodiment Division methods or analytic modell analytical model training method.

Memory 1202 may include storing program area and storage data area, wherein storing program area can store operation system Application program required for system, at least one function；Storage data area can store the division methods according to more demand searching requests Electronic equipment or the electronic equipment of analytic modell analytical model training method use created data etc..In addition, memory 1202 can It can also include non-transitory memory, for example, at least disk memory, a flash memory to include high-speed random access memory Device or other non-instantaneous solid-state memories.In some embodiments, it includes relative to processor that memory 1202 is optional 1201 remotely located memories, these remote memories can pass through the division methods of network connection at most demand searching request Or the electronic equipment of analytic modell analytical model training method.The example of above-mentioned network includes but is not limited to internet, intranet, local Net, mobile radio communication and combinations thereof.

The division methods of more demand searching requests or the electronic equipment of analytic modell analytical model training method can also include: input dress Set 1203 and output device 1204.Processor 1201, memory 1202, input unit 1203 and output device 1204 can pass through Bus or other modes connect, in Figure 12 for being connected by bus.

Input unit 1203 can receive the number or character information of input, and generate the division with more demand searching requests The related key signals input of the user setting and function control of the electronic equipment of method or analytic modell analytical model training method, such as touch It is defeated to touch screen, keypad, mouse, track pad, touch tablet, indicating arm, one or more mouse button, trace ball, control stick etc. Enter device.Output device 1204 may include display equipment, auxiliary lighting apparatus (for example, LED) and haptic feedback devices (example Such as, vibrating motor) etc..The display equipment can include but is not limited to, liquid crystal display (Liquid Crystal Display, LCD), light emitting diode (Light Emitting Diode, LED) display and plasma scope.In some embodiment party In formula, display equipment can be touch screen.

The various embodiments of system and technology described herein can be in digital electronic circuitry, integrated circuit system System, is consolidated specific integrated circuit (Application Specific Integrated Circuits, ASIC), computer hardware It is realized in part, software, and/or their combination.These various embodiments may include: to implement in one or more calculating In machine program, which can hold in programmable system containing at least one programmable processor Row and/or explain, which can be dedicated or general purpose programmable processors, can from storage system, at least One input unit and at least one output device receive data and instruction, and data and instruction is transmitted to the storage system System, at least one input unit and at least one output device.

These calculation procedures (also referred to as program, software, software application or code) include the machine of programmable processor Instruction, and can use programming language, and/or the compilation/machine language of level process and/or object-oriented to implement these Calculation procedure.As used herein, term " machine readable media " and " computer-readable medium " are referred to for referring to machine It enables and/or data is supplied to any computer program product, equipment, and/or the device of programmable processor (for example, disk, light Disk, memory, programmable logic device (programmable logic device, PLD)), including, receiving can as machine The machine readable media of the machine instruction of read signal.Term " machine-readable signal " is referred to for by machine instruction and/or number According to any signal for being supplied to programmable processor.

In order to provide the interaction with user, system and technology described herein, the computer can be implemented on computers Include for user show information display device (for example, CRT (Cathode Ray Tube, cathode-ray tube) or LCD (liquid crystal display) monitor)；And keyboard and indicator device (for example, mouse or trace ball), user can be by this Keyboard and the indicator device provide input to computer.The device of other types can be also used for providing the friendship with user Mutually；For example, the feedback for being supplied to user may be any type of sensory feedback (for example, visual feedback, audio feedback or Touch feedback)；And it can be received with any form (including vocal input, voice input or tactile input) from user Input.

System described herein and technology can be implemented including the computing system of background component (for example, as data Server) or the computing system (for example, application server) including middleware component or the calculating including front end component System is (for example, the subscriber computer with graphic user interface or web browser, user can pass through graphical user circle Face or the web browser to interact with the embodiment of system described herein and technology) or including this backstage portion In any combination of computing system of part, middleware component or front end component.Any form or the number of medium can be passed through Digital data communicates (for example, communication network) and is connected with each other the component of system.The example of communication network includes: local area network (LoCal Area Network, LAN), wide area network (Wide Area Network, WAN) and internet.

Computer system may include client and server.Client and server is generally off-site from each other and usually logical Communication network is crossed to interact.By being run on corresponding computer and each other with the meter of client-server relation Calculation machine program generates the relationship of client and server.

According to the technical solution of the embodiment of the present application, more demand searching requests are inputted into pre-set analytic modell analytical model；And At least two single demand searching requests are determined according to the output result of analytic modell analytical model, and more demand searching requests are drawn to realize Point.Later, single demand searching request can be shown, after receiving search instruction of the user for single demand searching request, root The search operation to single demand searching request is executed according to search instruction, and shows search result.Single demand is searched in this way, realizing The search of rope request.Analytic modell analytical model can be the first analytic modell analytical model；By the first analytic modell analytical model output generic items and particular term into Row splicing, available at least two single demands searching request.Alternatively, analytic modell analytical model can be the second analytic modell analytical model.Second solution Analysing model includes multiple slot position matching templates, the corresponding output rule of each slot position matching template；Utilize the filling knot of slot position Fruit and the corresponding output rule of slot position matching template, can determine single demand searching request.The application can be directed to different The second different analytic modell analytical models is arranged in classification, after the classification for determining more demand searching requests, more demand searching requests are defeated Enter corresponding second analytic modell analytical model of the category, the parsing to more demand searching requests is realized by second analytic modell analytical model.Due to Two analytic modell analytical models can be improved the accuracy of division for different classes of design.Identifying more demand searching requests Generic when, the application is identified using pre-set classifier, and the technical effect accurately quickly identified is reached.

It should be understood that various forms of processes illustrated above can be used, rearrangement increases or deletes step.Example Such as, each step recorded in the application of this hair can be performed in parallel or be sequentially performed the order that can also be different and execute, As long as it is desired as a result, being not limited herein to can be realized technical solution disclosed in the present application.

Above-mentioned specific embodiment does not constitute the limitation to the application protection scope.Those skilled in the art should be bright White, according to design requirement and other factors, various modifications can be carried out, combination, sub-portfolio and substitution.It is any in the application Spirit and principle within made modifications, equivalent substitutions and improvements etc., should be included within the application protection scope.