CN105069047B - A kind of search method and device of geography information - Google Patents
A kind of search method and device of geography information Download PDFInfo
- Publication number
- CN105069047B CN105069047B CN201510437871.4A CN201510437871A CN105069047B CN 105069047 B CN105069047 B CN 105069047B CN 201510437871 A CN201510437871 A CN 201510437871A CN 105069047 B CN105069047 B CN 105069047B
- Authority
- CN
- China
- Prior art keywords
- search
- neural network
- participle
- search modes
- modes
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Navigation (AREA)
Abstract
The invention discloses a kind of search method of geography information and devices, and wherein method includes: the search key for receiving user's input；The search key is segmented, and carries out neural network analysis using the participle of search key, determines the corresponding search modes of the search key；It is retrieved according to the search modes determined and returns to search result.After the present invention receives the search key of user's input, search key is segmented, then which kind of mode carries out neural network analysis to participle is suitble to retrieved to identify that user needs that is retrieved, it realizes and intelligence, accurate identification is carried out to user search keyword, it ensure that the accuracy of search result, meanwhile, user, which only needs to input search key, can be obtained accurate search result, it is easy to operate, improve the operated user experience of geography information inspection.
Description
Technical field
The present invention relates to technical field of information retrieval, in particular to the search method and device of a kind of geography information.
Background technique
With the rapid development of domestic automobile and the raising of people's economic level, the chance of self-driving trip is more and more,
But it because being unfamiliar with to road, takes an unnecessary way, take a wrong way occurs often.Destination is easily positioned for help people, at present market
On navigation software be directed to user's use demands and provided a variety of search functions, such as but relative complex operation and numerous
More search functions allows the first people using navigation to have no way of doing it.For users, search operaqtion is still not easy enough.
Summary of the invention
In view of the above problems, it proposes on the present invention overcomes the above problem or at least be partially solved in order to provide one kind
State the search method and device of a kind of geography information of problem.
It is based on the embodiment of the present invention in a first aspect, a kind of geography information provided in an embodiment of the present invention search method,
Include:
Receive the search key of user's input；
The search key is segmented, and carries out neural network analysis using the participle of search key, is determined
The corresponding search modes of the search key；
It is retrieved according to the search modes determined and returns to search result.
After the present embodiment is segmented using the search key that user inputs, neural network analysis is carried out to participle, from
And identify what user needs to retrieve, it is suitble to which kind of mode is retrieved, realizes and intelligence, essence are carried out to user search keyword
True identification ensure that the accuracy of search result, meanwhile, user, which only needs to input search key, can be obtained accurate inspection
Rope is as a result, easy to operate.
In one embodiment, the search key of user's input is received, comprising:
Receive the search key that user inputs from preset frame retrieval.
The present embodiment provides the input mode of search box for user, realizes frame retrieval, with further simplifying user
The operation of information retrieval is managed, the user experience is improved.
In one embodiment, the search key is segmented, comprising:
Word segmentation processing is carried out to the search key, obtains at least one participle；
The attribute of each participle is analyzed, the corresponding distinguished symbol of each participle, the distinguished symbol packet are marked
It includes:
Administrative region, peripheral extent, classification information, telephone number, intersection, number, point of interest and road.
In one embodiment, each participle is analyzed, according to the attribute of each participle, marks each participle pair
The distinguished symbol answered, specifically includes:
It by the attribute of each participle, is matched, is marked with the decision rule of preset each distinguished symbol and/or database
The corresponding distinguished symbol of each participle out.
It in the present embodiment, is matched, can quickly and accurately be marked using preset decision rule and/or database
The corresponding distinguished symbol of each participle out.
In one embodiment, neural network analysis is carried out using the participle of search key, determines that the retrieval is crucial
The corresponding search modes of word, comprising:
Using the weight of each participle and corresponding distinguished symbol as the backpropagation BP neural network pre-established
Input, obtains the output of the BP neural network；The output of the BP neural network includes: the corresponding score value of each search modes；
The corresponding score value of each search modes is ranked up, it is crucial that the highest search modes of score value are determined as the retrieval
The corresponding search modes of word；Or
The corresponding score value of each search modes is ranked up, and ranking results are pushed to user；User is therefrom selected
Search modes be determined as the corresponding search modes of the search key.
The specific embodiment for how obtaining search modes by neural network analysis is present embodiments provided, nerve is passed through
The analysis of network can accurately reflect that active user wants retrieval what and applicable search modes, by each search modes
Score value feed back to user after, the search modes that user is therefrom selected are as final search modes, with reference to the meaning of user
It is willing to, further ensures the accuracy of search result, while improving the usage experience of user.
In one embodiment, the search modes include following one or more:
Title search modes, systematic searching mode, telephone number search modes, intersection search modes, number inspection
Rope mode, periphery systematic searching mode and periphery title search modes.
In one embodiment, the BP neural network is pre-established by following manner:
Determining the topological structure of BP neural network, the topological structure includes: input layer, hidden layer and output layer, and
The neurode of each layer；
Determine the parametric variable of each layer in BP neural network；
It by being trained respectively to preset each training sample, and inversely feeds back, adjusts the parameter of each layer of neural network
Variable until the neural network desired output and real output value error be less than preset threshold value when, deconditioning,
Obtain stable BP neural network；
The neurode of the input layer includes: each participle and corresponding distinguished symbol；
The neurode of the hidden layer is following one or more: administrative region, peripheral extent, classification information, phone
Number, intersection, number, point of interest and road；
The neurode of the output layer is following one or more: title search modes, systematic searching mode, phone number
Code search modes, intersection search modes, number search modes, periphery systematic searching mode and periphery title retrieve mould
Formula.
Second aspect based on the embodiment of the present invention, the embodiment of the invention also provides a kind of retrieval of geography information dresses
It sets, comprising:
Receiving module, for receiving the search key of user's input；
Word segmentation module, for being segmented to the search key；
Analysis module determines the search key for carrying out neural network analysis using the participle of search key
Corresponding search modes；
Retrieval module, for being retrieved according to the search modes determined and returning to search result.
In one embodiment, the receiving module, the inspection inputted from preset frame retrieval specifically for receiving user
Rope keyword.
In one embodiment, the word segmentation module, comprising:
Submodule is segmented, for carrying out word segmentation processing to the search key, obtains at least one participle；
Submodule is marked, is analyzed for the attribute to each participle, the corresponding distinguished symbol of each participle is marked,
The distinguished symbol includes:
Administrative region, peripheral extent, classification information, telephone number, intersection, number, point of interest and road.
In one embodiment, the label submodule, specifically for by the attribute of each participle, with preset each identification
The decision rule and/or database of symbol are matched, and the corresponding distinguished symbol of each participle is marked.
In one embodiment, the analysis module, comprising:
Acquisition submodule, for using the weight of each participle and corresponding distinguished symbol as the reversed biography pre-established
The input for broadcasting BP neural network obtains the output of the BP neural network；The output of the BP neural network includes: each retrieval mould
The corresponding score value of formula；
Sorting sub-module, for the corresponding score value of each search modes to be ranked up；
Submodule is determined, for the highest search modes of score value to be determined as the corresponding retrieval mould of the search key
Formula；Or ranking results are pushed to user, it is corresponding that the search modes that user is therefrom selected are determined as the search key
Search modes.
In one embodiment, the search modes include following one or more:
Title search modes, systematic searching mode, telephone number search modes, intersection search modes, number inspection
Rope mode, periphery systematic searching mode and periphery title search modes.
In one embodiment, above-mentioned apparatus can also include:
BP neural network establishes module, for determining that the topological structure of BP neural network, the topological structure include: input
The neurode of layer, hidden layer and output layer and each layer；Determine the parametric variable of each layer in BP neural network；By to pre-
If each training sample be trained respectively, and inversely feed back, adjust the parametric variable of each layer of neural network until the nerve
When the desired output of network and the error of real output value are less than preset threshold value, deconditioning obtains stable BP nerve
Network；
The neurode of the input layer includes: each participle and corresponding distinguished symbol；
The neurode of the hidden layer is following one or more: administrative region, peripheral extent, classification information, phone
Number, intersection, number, point of interest and road；
The neurode of the output layer is following one or more: title search modes, systematic searching mode, phone number
Code search modes, intersection search modes, number search modes, periphery systematic searching mode and periphery title retrieve mould
Formula.
Technical solution provided in an embodiment of the present invention can include the following benefits:
The search method and device of geography information provided in an embodiment of the present invention receive the search key of user's input
Afterwards, search key is segmented, neural network analysis then is carried out to participle, so that it is assorted to identify that user needs to retrieve
, it is suitble to which kind of mode to be retrieved, realizes and intelligence, accurate identification are carried out to user search keyword, ensure that retrieval knot
The accuracy of fruit, meanwhile, user, which only needs to input search key, can be obtained accurate search result, and it is easy to operate, it is promoted
The operated user experience of geography information inspection.
Other features and advantages of the present invention will be illustrated in the following description, also, partly becomes from specification
It obtains it is clear that understand through the implementation of the invention.The objectives and other advantages of the invention can be by written explanation
Specifically noted structure is achieved and obtained in book, claims and attached drawing.
Below by drawings and examples, technical scheme of the present invention will be described in further detail.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with reality of the invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the flow chart of the search method of geography information provided in an embodiment of the present invention；
Fig. 2 is the flow chart provided in an embodiment of the present invention segmented to search key；
Fig. 3 is the flow chart of the step S13 in the search method of geography information provided in an embodiment of the present invention；
Fig. 4 A is the Establishing process figure of BP neural network provided in an embodiment of the present invention；
Fig. 4 B is an example of the network topological diagram of BP neural network provided in an embodiment of the present invention；
Fig. 5 is the structural schematic diagram of the retrieval device of geography information provided in an embodiment of the present invention；
Fig. 6 is the structural schematic diagram of word segmentation module provided in an embodiment of the present invention；
Fig. 7 is the structural schematic diagram of analysis module provided in an embodiment of the present invention.
Specific embodiment
Exemplary embodiments of the present disclosure are described in more detail below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
It is limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
It is fully disclosed to those skilled in the art.
The search method of geography information provided in an embodiment of the present invention, as shown in Figure 1, comprising:
S11, the search key for receiving user's input；
S12, the search key is segmented；
S13, neural network analysis is carried out using the participle of search key, determines the corresponding retrieval of the search key
Mode；
S14, it is retrieved according to the search modes determined and returns to search result.
The search method of geography information provided in an embodiment of the present invention, after receiving the search key that user inputs, to inspection
Rope keyword is segmented, and then carries out neural network analysis to participle, to identify that user needs that is retrieved, which is suitble to
Kind mode is retrieved, and is realized and is carried out intelligence to user search keyword, accurately identifies, ensure that the accurate of search result
Property, meanwhile, user, which only needs to input search key, can be obtained accurate search result, and it is easy to operate, improve geographical letter
The operated user experience of breath inspection.
The specific embodiment of above steps is described in detail separately below.
In embodiments of the present invention, in above-mentioned steps S11, the application program of geographic information retrieval function, Ke Yiti are provided
Options (such as pull-down menu select province, city, area) for search domain and can Freely input frame retrieval, user can be with
Search domain is first selected in options, and search key can be inputted in frame retrieval.
In the present invention, search key can be the search key that user inputs from frame retrieval.That is, connecing
The step of receiving the search key of user's input, the search key that can be inputted from preset frame retrieval by receiving user
To realize.
There is a kind of situation, includes search domain in the search key of user's input, simultaneously also by search domain
Options has selected search domain, under such situation, in the embodiment of the present invention, and content priority in the frame retrieval of user's input,
Ignore the content of the options of search domain, i.e., the search domain in search key inputted in frame retrieval with user is
It is quasi-.
Further, in the embodiment of the present invention, the composition of search key may include: search domain, retrieval object, divide
It is one or more in category information, for the explanation that simply gives some instances:
1. search domain+retrieval object,
A) for example: " the Shenyang City Olympic Sports Center ", " Shenyang City " are using administrative area as search domain, and " Olympic Sports Center " is inspection
Rope object.
B) for example: " the neighbouring Olympic Sports Center ", " neighbouring " it is using peripheral extent as search domain, " Olympic Sports Center " is
Retrieve object.
C) for example: " Shenyang City, literature and art road, youth street " either " Shenyang City, & literature and art road, youth street " or " Shenyang City is green
The crossing in year street and literature and art road ".Wherein " Shenyang City " is search domain, and " young street, literature and art road " is to make intersection
To retrieve object.
D) for example: " Shenyang City youth street 11 ", " Shenyang City " are search domains, and " young street 11 " is doorplate
Number.
E) for example: " Shenyang City 02412345678 ", " Shenyang City " are search domains, and " 02412345678 " is telephone number
2. search domain+classification, such as: " western-style restaurant, Shenyang City ", " Shenyang City " be using administrative area as search domain,
" western-style restaurant " is classification.
3. classification, such as: " western-style restaurant ".
4. object is retrieved, such as: " Olympic Sports Center ".
5. classification+retrieval object, such as: " the gymnasium Olympic Sports Center ", " gymnasium " they are classification, and " Olympic Sports Center " is retrieval
Object.
Further, search key is segmented, as shown in Fig. 2, may include following step:
S21, word segmentation processing is carried out to search key, obtains at least one participle.
S22, the attribute of each participle is analyzed, marks the corresponding distinguished symbol of each participle.
Above-mentioned distinguished symbol may include: administrative region, peripheral extent, classification information, telephone number, intersection, door
The trade mark, point of interest and road.
Further, in the specific implementation, above-mentioned administrative region, peripheral extent, classification information, telephone number, crossroad
Corresponding shorthand notation can be respectively adopted i.e. in mouth, number, point of interest and road: administrative region (AD), peripheral extent (NB),
Classification information (CLS), telephone number (PH), intersection (CRS), number (HSN), point of interest (POI) and road (RD).
In above-mentioned steps S21, word segmentation processing is carried out to search key, is referred to string matching in the related technology
Participle technique, the participle technique based on understanding and the participle technique based on statistics are segmented, and the process specifically segmented is herein not
It repeats again.
Further, in above-mentioned steps S22, the attribute of each participle is analyzed, each participle is marked and corresponds to
Distinguished symbol can be accomplished in the following manner in the specific implementation:
By the attribute of each participle, matched with the decision rule of preset each distinguished symbol and/or database, thus
Mark the corresponding distinguished symbol of each participle.
Specifically, administrative region (AD, Administrator), peripheral extent (NB, Nearby), classification information (CLS,
Class), telephone number (PH, phone number), intersection (CRS, Cross road), number (HSN, House
Number), the labeling method of the participle of point of interest (POI, Point Of Interest) and road (RD, Road) can refer to as
Lower explanation:
1, administrative region is marked: the administrative area data that pre-save is traversed according to each participle of user's input, by save,
City, area three-level fuzzy matching administrative area title.Administrative area title may include the full name, abbreviation, abbreviation in administrative area etc., require
Matching one by one.
If successful match, can mark this participle is administrative area.It is marked using label symbol AD.
2, to the label of peripheral extent: when user's input for example: the word of " nearby " and the expression such as " nearest " peripheral extent, can
To be matched with the dictionary of the word for indicating peripheral extent pre-saved, which is then periphery model by successful match
It encloses.It is marked using label symbol NB.
3, to the label of classification information: the search key that user is inputted can be matched in tag along sort data,
If successful match, marking a certain section of word segmentation result is classification information.It is marked using label symbol CLS.
Labeling process needs priority flag to go out point of search domain (including: administrative region and peripheral extent) and classification information
Word, remaining unlabelled participle will carry out succeeding marker as retrieval object.It can not for search domain and the label of classification
Successively.
Include to retrieval object tag: telephone number label, POI label, pavement marker, intersection label and number
Label.
The sequence of label can be with are as follows: telephone number label → intersection label → doorplate labelled notation → POI (point of interest)
Label and pavement marker.
4, telephone number is marked: when all numbers of the search key for identifying user's input, for current state
Whether family's phone number rule is analyzed, by taking China as an example, expect someone's call number feature letter comprising area code or "-" in analysis input
Breath, when meeting the rule of telephone number really, is marked using label symbol PH.
5, intersection information is marked: because of the particularity of intersection retrieval mode, retrieves intersection and need to input
Two road, if that is just marked as intersecting if the search key of user's input can be identified as the title in two road
Crossing.Such as: " young street literature and art road " is CRS (Cross road) using label symbol.
6, number information flag: the format of number information can generally follow link name+number (number+English/Chinese
Word) this form, such as: " young street 101 ".If can identify that the content before number is link name, that is just
By the later numeral mark of link name at number.System marks symbol is HSN (House Number).
7, POI (point of interest) information: the data volume of POI is relatively more, passes through statistical information to the title of POI, can extract
Some Feature Words are tentatively confirmed whether it is POI by way of Feature Words.It is marked using label symbol POI.
8, road information marks: can extract some Feature Words by the statistical information to title to the label of road, lead to
The mode of Feature Words is crossed, to identify whether being road, is marked using label symbol is RD.
Further, in above-mentioned steps S13, using search key participle carry out neural network analysis, determine described in
The corresponding search modes of search key can be realized in the specific implementation using process as shown in Figure 3:
S31, using the weight of each participle and corresponding distinguished symbol as pre-establish backpropagation (BP,
BackPropagation) the input of neural network.
S32, the output for obtaining BP neural network；The output of the BP neural network includes: the corresponding score value of each search modes.
S33, the corresponding score value of each search modes is ranked up；Then following step S34 or S35 are executed.
S34, the highest search modes of score value are determined as the corresponding search modes of the search key.
S35, ranking results are pushed to user；Then turn to step S36.
S36, the search modes for therefrom selecting user are determined as the corresponding search modes of search key.
In above-mentioned process, using this mathematical tool of BP neural network, identify search key it is corresponding be any
Search modes.
Further, search modes may include: title search modes, systematic searching mode, telephone number search modes,
Intersection search modes, number search modes, periphery systematic searching mode and periphery title search modes.
In order to facilitate understanding, above-mentioned various search modes are briefly described separately below:
1. title retrieval is the relatively high retrieval mode of most users frequency of use.By user select administrative area and
The keyword of input, for keyword, such as: the title of POI, the abbreviation of POI, the initial of POI title, the spelling name of POI
Claim, other voices of the title of other language of POI, the title of road, the initial of road, the spelling title of road and road
Title etc..Retrieving is that POI and road are retrieved within the scope of this administrative area.Administrative area range include, for example: it is provincial,
Prefecture-level, district grade, municipality directly under the Central Government, municipality directly under the Central Government have county, linchpin county, province, special administrative region etc. under its command.
2. systematic searching (or label search) allows the classification of user's selection POI, such as: dining room, gas station, parking
Field, market, convenience store and cinema etc., while also needing to select administrative area, point that user specifies is retrieved within the scope of administrative area
The POI of class.The POI original classification that these classification can be provided by map companies is produced in conjunction with the keyword of the input of user
It is raw.Such as: most users are liked retrieving original classification using keyword " dining room " being the POI of " Chinese-style restaurant ", then can
To increase such classification " dining room ".The classification that the relevant technologies of big data can help us to be arranged is more reasonable.
3. telephone number retrieval is that user is allowed to input phone number and selection administrative area to retrieve POI.Telephone number retrieval
Allow user to input since area code, can also be inputted since number segment.Compare other retrieval content, this retrieval
Retrieval content is number entirely.
4. intersection or branch road that the object of intersection retrieval (or crossing retrieval) retrieval is road and road
Mouthful, retrieving needs to select administrative area, the title of road is inputted, behind the two road for selecting you to be searched, two road
Intersection will provide.
5. number retrieves the doorplate on (or residence retrieval) retrieval road, if user is familiar to address, this
A function may be optimal selection, and user can be helped quickly to position that position.Retrieving needs user to select row
Administrative division and input road name, after selecting a road, can provide all numbers on this road.User only needs to select
Selecting the number to be looked for can located.
6. periphery systematic searching is using central point as the center of circle, radius retrieves POI within the scope of n Km.Central point can be certainly
The position of vehicle, the central point etc. being also possible on map." n " can be arranged according to the retrieval distance of selection.
7. title retrieval in periphery is using central point as the center of circle, radius is within the scope of n Km, according to the key search of input
POI.Central point can be car's location, the central point etc. being also possible on map." n " can be according to the retrieval distance of selection
To be arranged.
In embodiments of the present invention, it in S31 and S32, is realized using this mathematical tool of BP neural network to retrieval mould
The identification of formula.BP neural network can learn and store a large amount of input-output mode map relationship, for the embodiment of the present invention
Speech, is exactly that BP neural network first passes through training sample in advance, learns and store the mapping between the keyword of input and search modes
Relationship reaches a stable BP neural network (the i.e. aforementioned BP neural network pre-established) in this way, subsequent can will use
The participle and corresponding distinguished symbol of the search key of family input obtain corresponding with the search key various as input
The scoring (scoring is higher, then it is assumed that is more suitable for being retrieved using the search modes) of search modes, to judge user
Which kind of it is intended to that is retrieved, it is desirable to be retrieved using search modes.
The following table 1 is an example of the scoring of each search modes of BP neural network output:
Table 1
| Title retrieval | 90% |
| Systematic searching | 60% |
| Periphery retrieval | 55% |
| …… | …… |
When ranking results are pushed to user, upper table 1, which can all enumerate all search modes, to be come, and also be can choose and is pushed away
Several search modes that score is forward are sent, for example, if score less than 10%, there will not be in this table 1.
After getting the output of BP neural network, specific retrieval mode can be determined using two ways, one is
It automatically selects the highest search modes of score value and is determined as the corresponding search modes of the search key.Another is semi-artificial
It participates in, i.e., each retrieval mode is pushed to user according to the result of the sequence of score height, user is according to oneself wish, therefrom
Select a search modes, have search function application receive the feedback of user after, the search modes that are selected according to user
It is retrieved.By the content intelligence to user search, accurate identification, search result can be improved in above two mode
Accuracy.
BP neural network needs pre-establish.The basic principle of BP neural network establishment process can refer to as described below: defeated
Enter signal Xi and output node acted on by intermediate node (hiding node layer), by non-linear transformations, generates output signal Yk,
Each sample of network training includes input vector X and desired throughput t, the mistake between network output valve Y and desired output t
Difference, by adjusting between input node and the linking intensity value Wij and hiding node layer and output node of hiding node layer
Linking intensity Tjk and threshold value decline the error along gradient direction, by repetition learning training, determination and minimal error phase
Corresponding network parameter (weight and threshold value), training stop stopping, and obtain stable BP neural network.Trained BP at this time
Neural network (stable BP neural network) can voluntarily handle the smallest process of output error to the input information of similar sample
The information of non-linear conversion.
The learning process of BP neural network is the connection weight and error modified process repeatedly that network is determined according to sample.
For the embodiment of the present invention, as shown in Figure 4 A, BP neural network can be pre-established by following manner:
S41, the topological structure for determining BP neural network.
In order in resource-constrained navigation equipment, the performance that can have been shown, preferably, 3 layers of BP can be used
Network.
Wherein, which includes: the neurode of input layer, hidden layer and output layer and each layer.
The example of the topological structure of one neural network is as shown in Figure 4 B.
The neurode of input layer may include: each participle and corresponding distinguished symbol；
Preferably, the performance in order to guarantee BP neural network, is not more than 3 in the quantity of input layer, neurode.
The data structure of the input node of input layer can be such that (participle 1, distinguished symbol), (participle 2, distinguished symbol),
(participle 3, distinguished symbol).Participle 1, participle 2 and participle 3 are identical as the sequence of user's input.When the result of participle is greater than 3
When, all participles later are merged into participle 3 by the 3rd participle.
For example:
" the Shenyang Olympic Sports Center " is inputted after word segmentation processing, participle 1 is " Shenyang ", and distinguished symbol is administrative region
(AD), participle 2 is " Olympic Sports Center ", and identification is point of interest (POI), then the node of input layer is (Shenyang, AD), (in body difficult to understand
The heart, POI).
The neurode of hidden layer be it is following one or more: administrative region, peripheral extent, classification information, telephone number,
Intersection, number, point of interest and road.
The neurode of output layer is following one or more: title search modes, systematic searching mode, telephone number inspection
Rope mode, intersection search modes, number search modes, periphery systematic searching mode and periphery title search modes.
S42, the parametric variable for determining each layer in BP neural network；
The parametric variable of each layer includes: the connection weight between input layer and hidden layer in BP neural network, hidden layer with
Connection weight between output layer, the threshold value etc. of each neurode of hidden layer.
Above-mentioned parameter variable can choose preset historical experience value, and in training sample training process, constantly adjust.
S43, by being trained respectively to preset each training sample, calculate the desired output and reality of neural network
The error of output valve, and inversely feed back, the parametric variable of each layer of neural network is adjusted, S44 is then executed；
Whether the error of S44, the desired output for judging neural network and real output value are less than preset threshold value；If
It is to execute following S45；If it is not, repeating S43；
S45, deconditioning obtain stable BP neural network.
In this step S43, the format of training sample is referred to the following table 2:
Table 2
In training process, it is assumed that the participle 1 of input layer, participle 2 and to segment 3 target value be 0.3,0.5 and 0.2 also need
It will input multiplied by the normalized value of corresponding distinguished symbol, as each participle and corresponding distinguished symbol in input layer.
Such as (Shenyang, AD)=0.3*0.2=0.06.
The normalized value of distinguished symbol can first pass through in advance following manner and obtain:
Choose history search key (such as the previous 1000 times search key of user) conduct of user's big data quantity
After participle, accumulative number is carried out according to different distinguished symbols respectively for sample.The normalized value of each distinguished symbol=each
Distinguished symbol adds up number/whole identification object and adds up number.Available form table as shown in table 3 below:
Table 3
| Identify object | Distinguished symbol | Normalized value |
| Administrative region | AD | 0.2 |
| Peripheral extent | NB | 0.1 |
| Classify (label) | CLS | 0.05 |
| Telephone number | PH | 0.05 |
| Intersection (crossing) | CRS | 0.05 |
| Number (residence) | HSN | 0.15 |
| Point of interest | POI | 0.2 |
| Road | RD | 0.2 |
In above-mentioned S43, such as when the desired output of neural network and the error of real output value are less than 0.001, training
It terminates, has obtained stable BP neural network.
It in one embodiment, can be by the relevant search modes of POI for example when designing BP neural network: title retrieval
With the target value of systematic searching etc., be set as forward direction and be intended to 1, search modes relevant to road for example: intersection retrieval,
The target value of number retrieval, is set as reversely being intended to 0, in this way, in the output of final BP neural network, the relevant inspection of POI
The score of rope mode can immediately arrive at, and the score value that the score of the relevant search modes of road needs to export by 1-BP network
Mode obtain.
Based on the same inventive concept, the embodiment of the invention also provides a kind of devices of the retrieval of geography information, due to this
The principle of the solved problem of device is similar to the search method of aforementioned geography information, thus the implementation of the device may refer to it is aforementioned
The implementation of method, overlaps will not be repeated.
As shown in figure 5, a kind of retrieval device of geography information provided in an embodiment of the present invention, comprising:
Receiving module 51, for receiving the search key of user's input；
Word segmentation module 52, for being segmented to the search key；
Analysis module 53 determines that the retrieval is crucial for carrying out neural network analysis using the participle of search key
The corresponding search modes of word；
Retrieval module 54, for being retrieved according to the search modes determined and returning to search result.
Further, above-mentioned receiving module 51 is closed specifically for receiving the retrieval that user inputs from preset frame retrieval
Keyword.
Further, above-mentioned word segmentation module 52, as shown in Figure 6, comprising:
Submodule 521 is segmented, for carrying out word segmentation processing to the search key, obtains at least one participle；
Submodule 522 is marked, is analyzed for the attribute to each participle, the corresponding identifier of each participle is marked
Number, the distinguished symbol includes:
Administrative region, peripheral extent, classification information, telephone number, intersection, number, point of interest and road.
Further, above-mentioned label submodule 522, specifically for by the attribute of each participle, with preset each identifier
Number decision rule and/or database matched, mark the corresponding distinguished symbol of each participle.
Further, above-mentioned analysis module 53, as shown in fig. 7, comprises:
Acquisition submodule 531, for the weight of each participle and corresponding distinguished symbol is anti-as what is pre-established
To the input for propagating BP neural network, the output of the BP neural network is obtained；The output of the BP neural network includes: each inspection
The corresponding score value of rope mode；
Sorting sub-module 532, for the corresponding score value of each search modes to be ranked up；
Submodule 533 is determined, for the highest search modes of score value to be determined as the corresponding retrieval of the search key
Mode；Or ranking results are pushed to user, the search modes that user is therefrom selected are determined as the search key pair
The search modes answered.
Further, above-mentioned search modes include following one or more:
Title search modes, systematic searching mode, telephone number search modes, intersection search modes, number inspection
Rope mode, periphery systematic searching mode and periphery title search modes.
Further, the retrieval device of above-mentioned geography information, as shown in figure 5, can also include:
BP neural network establishes module 55, for determining that the topological structure of BP neural network, the topological structure include: defeated
Enter the neurode of layer, hidden layer and output layer and each layer；Determine the parametric variable of each layer in BP neural network；By right
Preset each training sample is trained respectively, and is inversely fed back, and adjusts the parametric variable of each layer of neural network until the mind
When the error of desired output and real output value through network is less than preset threshold value, deconditioning obtains stable BP mind
Through network；
The neurode of above-mentioned input layer includes: each participle and corresponding distinguished symbol；
The neurode of above-mentioned hidden layer is following one or more: administrative region, peripheral extent, classification information, phone
Number, intersection, number, point of interest and road；The neurode of above-mentioned output layer is following one or more: title inspection
Rope mode, systematic searching mode, telephone number search modes, intersection search modes, number search modes, periphery classification
Search modes and periphery title search modes.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The shape for the computer program product implemented in usable storage medium (including but not limited to magnetic disk storage and optical memory etc.)
Formula.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (14)
1. a kind of search method of geography information characterized by comprising
Receive the search key of user's input；
The search key is segmented, and using each participle of search key and corresponding distinguished symbol as mind
Input through network analysis obtains the output of neural network, determines that the search key is corresponding according to the output of neural network
Search modes；
It is retrieved according to the search modes determined and returns to search result.
2. the method as described in claim 1, which is characterized in that receive the search key of user's input, comprising:
Receive the search key that user inputs from preset frame retrieval.
3. the method as described in claim 1, which is characterized in that segmented to the search key, comprising:
Word segmentation processing is carried out to the search key, obtains at least one participle；
The attribute of each participle is analyzed, the corresponding distinguished symbol of each participle is marked, the distinguished symbol includes:
Administrative region, peripheral extent, classification information, telephone number, intersection, number, point of interest and road.
4. method as claimed in claim 3, which is characterized in that each participle is analyzed, according to the attribute of each participle,
The corresponding distinguished symbol of each participle is marked, is specifically included:
It by the attribute of each participle, is matched, is marked every with the decision rule of preset each distinguished symbol and/or database
It is a to segment corresponding distinguished symbol.
5. method as claimed in claim 3, which is characterized in that each participle and corresponding identifier using search key
Input number as neural network analysis, obtains the output of neural network, determines that the retrieval is closed according to the output of neural network
The corresponding search modes of keyword, comprising:
Using each participle and corresponding distinguished symbol as the input of the backpropagation BP neural network pre-established, institute is obtained
State the output of BP neural network；The output of the BP neural network includes: each search modes and the corresponding score value of each search modes；
The corresponding score value of each search modes is ranked up, the highest search modes of score value are determined as the search key pair
The search modes answered；Or
The corresponding score value of each search modes is ranked up, and ranking results are pushed to user；The inspection that user is therefrom selected
Rope mode is determined as the corresponding search modes of the search key.
6. method according to any of claims 1-4, which is characterized in that the search modes include following one or more
:
Title search modes, systematic searching mode, telephone number search modes, intersection search modes, number retrieve mould
Formula, periphery systematic searching mode and periphery title search modes.
7. method as claimed in claim 5, which is characterized in that the BP neural network is pre-established by following manner:
Determine that the topological structure of BP neural network, the topological structure include: input layer, hidden layer and output layer and each layer
Neurode；
Determine the parametric variable of each layer in BP neural network；
It by being trained respectively to preset each training sample, and inversely feeds back, adjusts the parametric variable of each layer of neural network
Until deconditioning obtains when the desired output of the neural network and the error of real output value are less than preset threshold value
Stable BP neural network；
The neurode of the input layer includes: each participle and corresponding distinguished symbol；
The neurode of the hidden layer be it is following one or more: administrative region, peripheral extent, classification information, telephone number,
Intersection, number, point of interest and road；
The neurode of the output layer is following one or more: title search modes, systematic searching mode, telephone number inspection
Rope mode, intersection search modes, number search modes, periphery systematic searching mode, periphery title search modes and each
The corresponding score value of search modes.
8. a kind of retrieval device of geography information characterized by comprising
Receiving module, for receiving the search key of user's input；
Word segmentation module, for being segmented to the search key；
Analysis module, for using each participle of search key and corresponding distinguished symbol as the defeated of neural network analysis
Enter, obtain the output of neural network, the corresponding search modes of the search key are determined according to the output of neural network；
Retrieval module, for being retrieved according to the search modes determined and returning to search result.
9. device as claimed in claim 8, which is characterized in that the receiving module is specifically used for receiving user from preset
The search key inputted in frame retrieval.
10. device as claimed in claim 9, which is characterized in that the word segmentation module, comprising:
Submodule is segmented, for carrying out word segmentation processing to the search key, obtains at least one participle；
Submodule is marked, is analyzed for the attribute to each participle, the corresponding distinguished symbol of each participle is marked, it is described
Distinguished symbol includes:
Administrative region, peripheral extent, classification information, telephone number, intersection, number, point of interest and road.
11. device as claimed in claim 10, which is characterized in that the label submodule, specifically for by each participle
Attribute is matched with the decision rule of preset each distinguished symbol and/or database, marks the corresponding identification of each participle
Symbol.
12. device as claimed in claim 10, which is characterized in that the analysis module, comprising:
Acquisition submodule, for using each participle and corresponding distinguished symbol as the backpropagation BP nerve net pre-established
The input of network obtains the output of the BP neural network；The output of the BP neural network includes: each search modes and each retrieval
The corresponding score value of mode；
Sorting sub-module, for the corresponding score value of each search modes to be ranked up；
Submodule is determined, for the highest search modes of score value to be determined as the corresponding search modes of the search key；Or
Ranking results are pushed to user by person, and the search modes that user is therefrom selected are determined as the corresponding retrieval of the search key
Mode.
13. such as the described in any item devices of claim 8-12, which is characterized in that the search modes include following one or more
:
Title search modes, systematic searching mode, telephone number search modes, intersection search modes, number retrieve mould
Formula, periphery systematic searching mode and periphery title search modes.
14. device as claimed in claim 12, which is characterized in that further include:
BP neural network establishes module, for determining the topological structure of BP neural network, the topological structure include: input layer,
The neurode of hidden layer and output layer and each layer；Determine the parametric variable of each layer in BP neural network；By to preset
Each training sample is trained respectively, and is inversely fed back, and adjusts the parametric variable of each layer of neural network until the neural network
Desired output and real output value error be less than preset threshold value when, deconditioning obtains stable BP neural network；
The neurode of the input layer includes: each participle and corresponding distinguished symbol；
The neurode of the hidden layer be it is following one or more: administrative region, peripheral extent, classification information, telephone number,
Intersection, number, point of interest and road；
The neurode of the output layer is following one or more: title search modes, systematic searching mode, telephone number inspection
Rope mode, intersection search modes, number search modes, periphery systematic searching mode, periphery title search modes and each
The corresponding score value of search modes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510437871.4A CN105069047B (en) | 2014-07-25 | 2015-07-23 | A kind of search method and device of geography information |
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2014103601984 | 2014-07-25 | ||
| CN201410360198.4A CN104112020A (en) | 2014-07-25 | 2014-07-25 | Frame type retrieval method for navigation equipment |
| CN201510437871.4A CN105069047B (en) | 2014-07-25 | 2015-07-23 | A kind of search method and device of geography information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105069047A CN105069047A (en) | 2015-11-18 |
| CN105069047B true CN105069047B (en) | 2019-05-07 |
Family
ID=51708811
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410360198.4A Pending CN104112020A (en) | 2014-07-25 | 2014-07-25 | Frame type retrieval method for navigation equipment |
| CN201510437871.4A Active CN105069047B (en) | 2014-07-25 | 2015-07-23 | A kind of search method and device of geography information |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410360198.4A Pending CN104112020A (en) | 2014-07-25 | 2014-07-25 | Frame type retrieval method for navigation equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN104112020A (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105183908B (en) * | 2015-09-30 | 2019-05-28 | 北京奇虎科技有限公司 | A kind of classification method and device of point of interest POI data |
| CN106919569B (en) * | 2015-12-24 | 2020-09-29 | 北京四维图新科技股份有限公司 | Method and device for acquiring administrative division information of point of interest (POI) |
| CN106446235B (en) * | 2016-10-10 | 2021-04-06 | Tcl科技集团股份有限公司 | Video searching method and device |
| CN106528644B (en) * | 2016-10-14 | 2020-07-31 | 航天恒星科技有限公司 | Remote sensing data retrieval method and device |
| TWI773658B (en) * | 2017-01-17 | 2022-08-11 | 香港商阿里巴巴集團服務有限公司 | POI service providing method, POI data processing method and device |
| CN108280195A (en) * | 2018-01-26 | 2018-07-13 | 五八有限公司 | Querying method, device, equipment based on index slip and storage medium |
| CN112988750A (en) * | 2018-03-14 | 2021-06-18 | 北京忆芯科技有限公司 | Key and value generation method and device for KV storage based on structured information |
| CN109033222B (en) * | 2018-06-29 | 2021-07-13 | 北京奇虎科技有限公司 | Method and device for analyzing correlation between POI (point of interest) and search keyword |
| CN110929176A (en) * | 2018-09-03 | 2020-03-27 | 北京搜狗科技发展有限公司 | Information recommendation method and device and electronic equipment |
| CN109992695B (en) * | 2019-04-10 | 2021-08-27 | 北京奇艺世纪科技有限公司 | Video information query method and device |
| CN110442759B (en) * | 2019-07-25 | 2022-05-13 | 深圳供电局有限公司 | Knowledge retrieval method and system, computer equipment and readable storage medium |
| CN111831929B (en) * | 2019-09-24 | 2024-01-02 | 北京嘀嘀无限科技发展有限公司 | Method and device for acquiring POI information |
| CN112487313A (en) * | 2020-11-30 | 2021-03-12 | 北京百度网讯科技有限公司 | Interest point identification method and device, electronic equipment and storage medium |
| CN112508532A (en) * | 2020-12-16 | 2021-03-16 | 深圳市中博科创信息技术有限公司 | Traffic statistical method for second batch service management system |
| CN112861022A (en) * | 2021-02-01 | 2021-05-28 | 杭州全拓科技有限公司 | Artificial intelligence-based personnel activity big data record query method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103336848A (en) * | 2013-07-22 | 2013-10-02 | 五八同城信息技术有限公司 | Ordering method for classified information |
| CN103678684A (en) * | 2013-12-25 | 2014-03-26 | 沈阳美行科技有限公司 | Chinese word segmentation method based on navigation information retrieval |
| CN104199822A (en) * | 2014-07-11 | 2014-12-10 | 五八同城信息技术有限公司 | Method and system for identifying demand classification corresponding to searching |
-
2014
- 2014-07-25 CN CN201410360198.4A patent/CN104112020A/en active Pending
-
2015
- 2015-07-23 CN CN201510437871.4A patent/CN105069047B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103336848A (en) * | 2013-07-22 | 2013-10-02 | 五八同城信息技术有限公司 | Ordering method for classified information |
| CN103678684A (en) * | 2013-12-25 | 2014-03-26 | 沈阳美行科技有限公司 | Chinese word segmentation method based on navigation information retrieval |
| CN104199822A (en) * | 2014-07-11 | 2014-12-10 | 五八同城信息技术有限公司 | Method and system for identifying demand classification corresponding to searching |
Non-Patent Citations (1)
| Title |
|---|
| 图吧导航软件使用说明书;sunny521000;《百度文库》;20101230;第6页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105069047A (en) | 2015-11-18 |
| CN104112020A (en) | 2014-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105069047B (en) | A kind of search method and device of geography information | |
| CN105045858B (en) | Taxi pickup point based on ballot recommends method | |
| CN108681586B (en) | Tourist route personalized recommendation method based on crowd sensing | |
| CN109145169A (en) | A kind of address matching method based on statistics participle | |
| Moral-Muñoz et al. | Analyzing highly cited papers in intelligent transportation systems | |
| CN110472066A (en) | A kind of construction method of urban geography semantic knowledge map | |
| CN106991161A (en) | A kind of method for automatically generating open-ended question answer | |
| CN106462624A (en) | Tile-based geocoder | |
| CN110162591A (en) | A kind of entity alignment schemes and system towards digital education resource | |
| CN110413886A (en) | A kind of point of interest methods of exhibiting and device | |
| CN104182517A (en) | Data processing method and data processing device | |
| CN108897887A (en) | A kind of teaching resource recommended method of knowledge based map and user's similarity | |
| CN103593474A (en) | Image retrieval ranking method based on deep learning | |
| CN112163101B (en) | Geographic entity matching and fusing method for spatial knowledge graph | |
| CN107084735A (en) | Guidance path framework suitable for reducing redundancy navigation | |
| CN110019617A (en) | The determination method and apparatus of address mark, storage medium, electronic device | |
| CN104636457B (en) | A kind of method and device of location finding cognition | |
| CN114297532B (en) | High-definition city functional area identification method and system based on crowd movement track | |
| CN108345609A (en) | A kind of method and apparatus of processing POI information | |
| CN104112366B (en) | Method for traffic signal optimization based on latent semantic model | |
| CN111625732A (en) | Address matching method and device | |
| CN107368480A (en) | A kind of interest point data type of error positioning, repeat recognition methods and device | |
| CN108090220A (en) | Point of interest search sort method and system | |
| CN110457706B (en) | Point-of-interest name selection model training method, using method, device and storage medium | |
| Contractor et al. | Joint Spatio-textual reasoning for answering tourism questions |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 110000 No. 861-6, shangshengou village, Hunnan District, Shenyang City, Liaoning Province Patentee after: Shenyang Meihang Technology Co.,Ltd. Address before: 110004 Room 202, 1t international, No. 3, Yuanhang West Road, Hunnan New District, Shenyang, Liaoning Patentee before: SHENYANG MXNAVI Co.,Ltd. |
|
| CP03 | Change of name, title or address |