CN108595620A - Escape recognition methods, device, computer equipment and storage medium - Google Patents
Escape recognition methods, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN108595620A CN108595620A CN201810367116.7A CN201810367116A CN108595620A CN 108595620 A CN108595620 A CN 108595620A CN 201810367116 A CN201810367116 A CN 201810367116A CN 108595620 A CN108595620 A CN 108595620A
- Authority
- CN
- China
- Prior art keywords
- word
- feature vector
- feature
- escape
- object word
- 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.)
- Granted
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application proposes a kind of escape recognition methods, device, computer equipment and storage medium, wherein method includes：Obtain first object word to be identified and the second target word；Determine the corresponding first eigenvector of the first object word and second feature vector and the corresponding third feature vector of the second target word and fourth feature vector；Wherein, the first eigenvector is related to the second target word, and second feature vector is unrelated with the second target word, and third feature vector is related to the first object word, and fourth feature vector is unrelated with the first object word；According to the first eigenvector between the second feature vector at a distance from and the third feature between the fourth feature vector at a distance from, determine the escape probability when first object word and the second target word combination.By this method, the accuracy and reliability of escape identification can be improved, and then improve the accuracy of search result.
Description
Technical field
This application involves search engine technique field more particularly to a kind of escape recognition methods, device computer equipment and
Storage medium.
Background technology
Retrieval is search statement of the search engine according to expression query intention input by user, returns to a certain number of search
As a result process.The search result that search engine returns may be matched only with search statement, but not meet the true inquiry of user
It is intended to, for example, search statement input by user is " diamond ", the search result that search engine returns is the letter of " diamond pad pasting "
Breath, such case are referred to as escape.Escape can seriously affect the search experience of user.
In order to return to the search result for meeting user's query intention, need to carry out escape knowledge to candidate search result
Not.In the related technology, escape identification is realized using the escape identification model that study obtains.In general, the search result showed
Click volume it is higher, the probability that escape does not occur between search statement and search result is higher, and for repeatedly showing and without point
The amount of hitting or the seldom search result of click volume, the probability that escape occurs are higher.Based on this, in the related technology, using the point of user
It hits data to learn to obtain escape identification model as training sample, be identified for escape.
However, the mode for obtaining escape identification model dependent on user's click Behavioral training is more unilateral, for the point of user
Hit the keyword not occurred in data, it is difficult to which escape information is arrived in study, and user is unintentionally overdue to be hit or intentional click cheating is equal
The accuracy of identification that can influence escape identification model causes escape recognition accuracy low.
Invention content
The application is intended to solve at least some of the technical problems in related technologies.
For this purpose, first purpose of the application is to propose a kind of escape recognition methods, with by obtaining first object word
Language and the relevant first eigenvector of the second target word and the second feature unrelated with the second target word vector, and obtain
The fourth feature for taking the second target word unrelated with first object word with the relevant third feature vector sum of first object word
Vector, so according to the distance between first eigenvector and second feature vector and third feature vector and fourth feature to
The distance between amount determines escape probability when first object word and the second target word combination, improves the standard of escape identification
True property and reliability, and then improve the accuracy of search result.
Second purpose of the application is to propose a kind of escape identification device.
The third purpose of the application is to propose a kind of computer equipment.
The 4th purpose of the application is to propose a kind of non-transitorycomputer readable storage medium.
The 5th purpose of the application is to propose a kind of computer program product.
In order to achieve the above object, the application first aspect embodiment proposes a kind of escape recognition methods, including：
Obtain first object word to be identified and the second target word；
Determine the corresponding first eigenvector of the first object word and second feature vector and second target word
The corresponding third feature vector of language and fourth feature vector；Wherein, the first eigenvector and the second target word phase
It closing, second feature vector is unrelated with the second target word, and third feature vector is related to the first object word, and the 4th
Feature vector is unrelated with the first object word；
According to the first eigenvector between the second feature vector at a distance from and the third feature and described the
Distance between four feature vectors determines the first object word and escape probability when the second target word combination.
The escape recognition methods of the embodiment of the present application, by obtaining first object word and the second target word to be identified
Language, and determine first object word and the relevant first eigenvector of the second target word and unrelated with the second target word
Second feature vector, and determine the second target word and the relevant third feature vector sum of first object word and first object
The unrelated fourth feature vector of word, so it is special according to the distance between first eigenvector and second feature vector and third
The distance between sign vector and fourth feature vector, determine that first object word and escape when the second target word combination are general
Rate.As a result, by the influence according to two words to feature vector each other, whether to occur to turn when determining two word combinations
Justice to improve the accuracy and reliability of escape identification, and then improves the accuracy of search result.
In order to achieve the above object, the application second aspect embodiment proposes a kind of escape identification device, including：
Acquisition module, for obtaining first object word to be identified and the second target word；
Determining module, for determine the corresponding first eigenvector of the first object word and second feature vector, and
The corresponding third feature vector of the second target word and fourth feature vector；Wherein, the first eigenvector with it is described
Second target word is related, and second feature vector is unrelated with the second target word, third feature vector and first mesh
It is related to mark word, fourth feature vector is unrelated with the first object word；
Escape probability determination module, at a distance from according to the first eigenvector between the second feature vector,
And the third feature between the fourth feature vector at a distance from, determine the first object word and second target word
Escape probability when language combines.
The escape identification device of the embodiment of the present application, by obtaining first object word and the second target word to be identified
Language, and determine first object word and the relevant first eigenvector of the second target word and unrelated with the second target word
Second feature vector, and determine the second target word and the relevant third feature vector sum of first object word and first object
The unrelated fourth feature vector of word, so it is special according to the distance between first eigenvector and second feature vector and third
The distance between sign vector and fourth feature vector, determine that first object word and escape when the second target word combination are general
Rate.As a result, by the influence according to two words to feature vector each other, whether to occur to turn when determining two word combinations
Justice to improve the accuracy and reliability of escape identification, and then improves the accuracy of search result.
In order to achieve the above object, the application third aspect embodiment proposes a kind of computer equipment, including：It processor and deposits
Reservoir；Wherein, the processor is held to run with described by reading the executable program code stored in the memory
The corresponding program of line program code, for realizing the escape recognition methods as described in first aspect embodiment.
In order to achieve the above object, the application fourth aspect embodiment proposes a kind of non-transitory computer-readable storage medium
Matter is stored thereon with computer program, realizes that the escape as described in first aspect embodiment is known when which is executed by processor
Other method.
In order to achieve the above object, the 5th aspect embodiment of the application proposes a kind of computer program product, when the calculating
The escape recognition methods as described in first aspect embodiment is realized when instruction in machine program product is executed by processor.
The additional aspect of the application and advantage will be set forth in part in the description, and will partly become from the following description
It obtains obviously, or recognized by the practice of the application.
Description of the drawings
The application is above-mentioned and/or additional aspect and advantage will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, wherein：
A kind of flow diagram for escape recognition methods that Fig. 1 is provided by the embodiment of the present application；
Fig. 2 is the method flow schematic diagram that first eigenvector and second feature vector are determined according to co-occurrence word；
Fig. 3 is the method flow schematic diagram that first eigenvector and second feature vector are determined according to webpage information；
Fig. 4 is the method flow schematic diagram that first eigenvector and second feature vector are determined according to image content；
The flow diagram for another escape recognition methods that Fig. 5 is provided by the embodiment of the present application；
A kind of structural schematic diagram for escape identification device that Fig. 6 is provided by the embodiment of the present application；
The structural schematic diagram for another escape identification device that Fig. 7 is provided by the embodiment of the present application；
The structural schematic diagram for another escape identification device that Fig. 8 is provided by the embodiment of the present application；
The structural schematic diagram for another escape identification device that Fig. 9 is provided by the embodiment of the present application；
The structural schematic diagram for also a kind of escape identification device that Figure 10 is provided by the embodiment of the present application；And
Figure 11 is the structural schematic diagram for the computer equipment that one embodiment of the application proposes.
Specific implementation mode
Embodiments herein is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the application, and should not be understood as the limitation to the application.
Below with reference to the accompanying drawings the escape recognition methods, device, computer equipment and storage medium of the embodiment of the present application are described.
According to the viewpoint of linguist, the semantic of word is determined by the context distribution of the word.Applicant passes through to word
The example rear discovery for statistical analysis of escape occurs for language, and most escapes are happened at adjacent context, above and below farther out
Text will not cause word escape substantially.However, the semantic of word is determined by the distribution of its context, it is not meant to the semanteme of word
By isolated context environmental offhand decision, therefore to obtain escape may occur for the semanteme that can learn word by big data
Word.In addition, whether there is escape between search statement and title text, can also be sentenced by the information except text
It is disconnected, for example, distinguishing whether search statement occurs escape according to the image results of search statement retrieval.
Based on this, the embodiment of the present application proposes a kind of escape recognition methods, to improve the accuracy rate of escape identification, Jin Erti
The accuracy of high search result.
A kind of flow diagram for escape recognition methods that Fig. 1 is provided by the embodiment of the present application.
As shown in Figure 1, the escape recognition methods may comprise steps of：
Step 101, first object word to be identified and the second target word are obtained.
Wherein, first object word and the second target word can be arbitrary relevant two words, such as to go out simultaneously
Two existing words, alternatively, for the keyword etc. in search term and corresponding search result, the present embodiment does not limit this
It is fixed.
For example, if escape recognition methods provided by the present application, is realized by search engine, then what search engine obtained
First object word, can be search statement in keyword, the second target word, can according to search statement obtain search
Keyword etc. in hitch fruit.
Wherein, the keyword in search result can be the word occurred simultaneously with the keyword in search statement.For example,
It, can be using search statement input by user as first object word when search statement input by user is an individual word
Language, and the internet information collection for including the first object word in network is obtained, it concentrates and determines and the first mesh from internet information
The word that mark word occurs jointly is as the second target word.For example, when user inputs " diamond ", first object word is " to bore
Stone ", the second target word can be " pad pasting ", " grade " " brand " " trump " " how much " etc..
Alternatively, when search statement input by user is a phrase, it is defeated to user that relevant segmenting method may be used
The search statement that enters carries out word segmentation processing, using the word after participle as first object word and the second target word.For example,
When user inputs " apple variety ", word segmentation processing can be carried out to " apple variety " and obtain " apple " and " kind ", by " apple
Fruit " is used as first object word, and " kind " is used as the second target word.
Step 102, the corresponding first eigenvector of first object word and second feature vector and the second target word are determined
The corresponding third feature vector of language and fourth feature vector.
Wherein, first eigenvector is related to the second target word, and second feature vector is unrelated with the second target word, the
Three feature vectors are related to first object word, and fourth feature vector is unrelated with first object word.
In the present embodiment, after obtaining first object word to be identified and the second target word, it can utilize default
Language model, determine first object word and the corresponding each feature vector of the second target word respectively；Alternatively, can also utilize
The methods of deep learning, determines first object word and the corresponding each feature vector of the second target word, the present embodiment to this not
It limits.
Step 103, at a distance from according to first eigenvector between second feature vector and third feature and fourth feature to
Distance between amount determines escape probability when first object word and the second target word combination.
In the present embodiment, it is determined that the corresponding first eigenvector of first object word and second feature vector, Yi Ji
After the corresponding third feature vector sum fourth feature vector of two target words, first eigenvector and second feature can be calculated
The distance between vector and the distance between third feature vector and fourth feature vector calculate the distance between vector
Mode can there are many, such as can calculate Euclidean distance between first eigenvector and second feature vector, mahalanobis distance,
Hamming distance, Chebyshev's distance, manhatton distance etc..The application is to calculating between first eigenvector and second feature vector
Distance and third feature vector between fourth feature vector at a distance from mode be not construed as limiting, but herein it should be noted that
Calculate first eigenvector between second feature vector at a distance from, and calculate third feature vector fourth feature vector between
Apart from when, should use identical calculation, to ensure that computational accuracy is identical.
In turn, according to calculate gained first eigenvector and second feature vector between distance and third feature to
It measures at a distance between fourth feature vector, it may be determined that escape probability when first object word and the second target word combination.
Specifically, when being less than first threshold at a distance from first eigenvector is between second feature vector, and third feature with
When distance between fourth feature vector is more than second threshold, it is determined that turn when first object word and the second target word combination
Adopted probability is more than third threshold value；Alternatively, when being more than second threshold, and the at a distance from first eigenvector is between second feature vector
Three features between fourth feature vector at a distance from be less than first threshold when, it is determined that first object word and the second target word group
Escape probability when conjunction is more than third threshold value.
Wherein, first threshold is less than or equal to second threshold, and first threshold, second threshold and third threshold value are to preset
's.
In actual use, if first object word and the second target word are respectively word and search in search statement
As a result the word in, then when the escape probability when first object word and the second target word combination is more than third threshold value,
Escape occurs when can determine first object word and the second target word combination, and then can be to where the second target word
Search result is screened, to return to the matched search result of query intention with user to user.
The escape recognition methods of the present embodiment, by obtaining first object word and the second target word to be identified, and
Determine first object word and the relevant first eigenvector of the second target word and unrelated with the second target word second
Feature vector, and determine the second target word and the relevant third feature vector sum of first object word and first object word
Unrelated fourth feature vector, so according to the distance between first eigenvector and second feature vector and third feature to
The distance between amount and fourth feature vector, determine escape probability when first object word and the second target word combination.By
This, is by the influence according to two words to feature vector each other, whether escape occurs when determining two word combinations, to
The accuracy and reliability of escape identification is improved, and then improves the accuracy of search result.
In order to determine the corresponding first eigenvector of first object word and second feature vector, this application provides three kinds
Possible realization method.
It, can be in conjunction with the co-occurrence word of first object word, according to the first mesh as the possible realization method of one of which
Word and its co-occurrence word are marked, determines the corresponding first eigenvector of first object word and second feature vector.According to Fig. 2
Co-occurrence word determines the method flow schematic diagram of first eigenvector and second feature vector.
As shown in Fig. 2, on the basis of embodiment as shown in Figure 1, step 102 may comprise steps of：
Step 201, data are carried out to network to crawl, obtains the corresponding first co-occurrence word set of first object word and second total
Existing word set, wherein the first co-occurrence word concentration includes the second target word.
In the present embodiment, according to the first object word of acquisition and the second target word, data can be carried out to network and climbed
It takes, to obtain the first co-occurrence word set corresponding with first object word and the second co-occurrence word set.For example, can be according to first object
Word, acquisition includes the data of first object word from network data, for example, web page text, picture header text can be obtained
Originally, the text datas such as inquiry log of user, and word segmentation processing is carried out to text data and goes the pretreatments such as stop words, from acquisition
Text data in extract first object word and its co-occurrence word, filtered out from co-occurrence word including the second target word
Word form the first co-occurrence word set, the first co-occurrence word set is removed into remaining co-occurrence word composition second after the second target word
Co-occurrence word set.
Step 202, each co-occurrence word for including is concentrated according to the first co-occurrence word, determines corresponding first spy of first object word
Sign vector.
The each co-occurrence word concentrated for the first co-occurrence word, it may be determined that the word between first object word and the co-occurrence word
Vector is used as first eigenvector.
As an example, the method that deep learning may be used, in advance training obtain word insertion vector model, by word into
Row vector, and then first object word and co-occurrence word are input in word insertion vector model, obtain first eigenvector.
As an example, vector space model (Vector Space Model, VSM) may be used by first object word
Language and co-occurrence word are converted to term vector, and then indicate first eigenvector using term vector and corresponding weight.Wherein, weight
It can be determined according to frequency that first object word and co-occurrence word occur in text data, specifically, can be directed to and include
The pretreated text data of first object word and co-occurrence word, counts first object word respectively and co-occurrence word exists
The frequency occurred in text data recycles TF-IDF (term frequency-inverse using the frequency as initial weight
Document frequency) Weight algorithm calculates final weight, it is determined using final weight and term vector final
First eigenvector.For example, may be used to the corresponding term vector of first object word and the corresponding term vector of co-occurrence word into
The mode of row weighted sum obtains first eigenvector.
Step 203, each co-occurrence word for including is concentrated according to the second co-occurrence word, determines corresponding second spy of first object word
Sign vector.
In the present embodiment, for each co-occurrence word that the second co-occurrence word is concentrated, it may be used and calculate first eigenvector
Identical mode determines that the corresponding second feature vector of first object word, concrete mode are retouched referring to the correlation in step 202
It states, is no longer described in detail herein.
In conclusion by crawling network data, it includes the of the second target word that it is corresponding, which to obtain first object word,
One co-occurrence word set and the second co-occurrence word set for not including the second target word, and then according to the first co-occurrence word set and the second co-occurrence word
Each co-occurrence word that concentration includes determines the corresponding first eigenvector of first object word and second feature vector respectively, can
From the corresponding first eigenvector of co-occurrence word angle-determining first object word and second feature vector, to realize turning for multi-angle
Justice identification, the coverage rate for improving escape identification lay the foundation.
As alternatively possible realization method, the net that can also be obtained according to first object word and the second target word
Page information determines the corresponding first eigenvector of first object word and second feature vector.Fig. 3 is to be determined according to webpage information
The method flow schematic diagram of first eigenvector and second feature vector.
As shown in figure 3, on the basis of embodiment as shown in Figure 1, step 102 may comprise steps of：
Step 301, data are carried out to network to crawl, obtains the first page collection and second page for including first object word
Collection, wherein it includes the second target word that first page, which concentrates at least one page,.
It, can be from network number after obtaining first object word to be identified and the second target word in the present embodiment
First page collection and second page collection including first object word are retrieved in.
Further, it in order to which the first page collection and second page that ensure acquisition concentrate the page quality for including, and keeps away
The page set data volume for exempting to obtain causes greatly data processing difficulty big very much, in a kind of possible realization method of the embodiment of the present application
In, the page that include can also be concentrated to screen first page collection and second page, data volume size is suitable, matter to obtain
Measure higher page set.For example, can according to the number that first object word occurs in the page to the page in page set into
Row screening or the number that is occurred in the page according to the second target word screen the page in page set, or according to the
The total degree that one target word and the second target word occur in the page such as screens at the page in page set, will occur
The page that number is less than predetermined threshold value is deleted from page set；And/or it can also will include sensitive vocabulary, advertisement etc. in the page
The low-quality page deleted from page set.In turn, using after deletion first page collection and second page collection determine fisrt feature
Vector sum second feature vector.
Step 302, the attribute information that each page is concentrated according to first page determines corresponding first spy of first object word
Sign vector.
Wherein, the attribute information of each page includes but not limited to the type of each page or the type of the affiliated website of each page.
Step 303, the attribute information that each page is concentrated according to second page determines corresponding second spy of first object word
Sign vector.
Typically for a word, when being interpreted as different meanings, including the type of the page of the word is often not
Together.For example, for " emerald " and " emerald bean curd ", although all including " emerald " in the two words, first word usually goes out
In the present other page of jewelry, and second word typically occurs in the page of cuisines class.It therefore, can be in the present embodiment
The corresponding first eigenvector of first object word and second feature vector are determined according to the type of the page.
As an example, each page concentrated for first page collection and second page, can first obtain the class of the page
The type of type or the affiliated website of the page, and then vectorial expression is carried out to the type of acquisition, obtain fisrt feature in conjunction with weight
Term vector or second feature term vector.Wherein, weight can be true by counting the frequency that first object word occurs in the page
It is fixed.
In conclusion being crawled by carrying out data to network, the first page collection comprising first object word and the are obtained
Two page sets, it includes the second target word that first page, which concentrates at least one page, and then according to first page collection and second page
The attribute information of each page is concentrated in face, determines the corresponding first eigenvector of first object word and second feature vector, can
The first eigenvector and second feature vector that first object word is determined according to the type of the affiliated page of word to be identified, are real
The escape identification of existing multi-angle, the coverage rate for improving escape identification lay the foundation.
As alternatively possible realization method, corresponding picture search result can be obtained according to first object word,
The corresponding first eigenvector of first object word and second feature vector are determined according to image content.Fig. 4 is according in picture
Hold the method flow schematic diagram for determining first eigenvector and second feature vector.
As shown in figure 4, on the basis of embodiment as shown in Figure 1, step 102 may comprise steps of：
Step 401, it is the first pictures and second picture collection to obtain corresponding with first object word, wherein the first pictures
In picture at least one picture pictures corresponding with the second target word it is identical.
It, can be by the first mesh after obtaining first object word to be identified and the second target word in the present embodiment
Word and the second target word are marked as search statement, relevant picture is obtained from photographic search engine, utilizes first object word
The picture obtained as search statement with the second target word in the picture that language is obtained as search statement is identical at least one
Picture generates the first pictures.
In a kind of possible realization method of the embodiment of the present application, respectively according to first object word and the second target word
After obtaining corresponding picture, the picture of acquisition can be ranked up, and (N is just to N before selecting from the picture after sequence
The value of integer, N can be preset, such as N=1000 or N=10000) a picture, is generated respectively using N number of picture
The data volume of first pictures and second picture collection is controlled in suitable size, is avoided by one pictures and second picture collection
Data processing difficulty is larger and increases handling duration, and then causes the feedback efficiency of search engine low.
Step 402, according to the content of each picture in the first pictures, determine the corresponding fisrt feature of first object word to
Amount.
Step 403, the content that each picture is concentrated according to second picture, determine the corresponding second feature of first object word to
Amount.
In the present embodiment, for each picture that the first pictures and second picture are concentrated, it can be determined according to image content
The corresponding first eigenvector of first object word or second feature vector.
As an example, the color characteristic, textural characteristics and shape feature of picture can be extracted, and samples relevant retouch
Method is stated, vectorization expression is carried out respectively to the color characteristic, textural characteristics and shape feature of picture, for example, for color spy
Sign, may be used histogram method and quantifies to the color characteristic of image；For textural characteristics, gray level co-occurrence matrixes may be used
Textural characteristics are quantified；For shape feature, area invariant moment method may be used, shape feature is quantified.In turn,
It is indicated using the color characteristic expression of picture, textural characteristics and shape feature indicates, determine the fisrt feature of first object word
Vector or second feature vector.
As an example, a large amount of picture sample (including picture and its class label) can be acquired, by picture sample
As input, initial depth neural network model is trained, obtains trained picture classification deep neural network model.
Wherein, the obtained picture classification deep neural network model of training, can first by image content be expressed as corresponding feature to
Amount exports the class label of picture later further according to feature vector in output layer.In turn, it obtains and first object word pair
After the first pictures and second picture collection answered, you can the picture in the first pictures is input to picture classification depth nerve
In network model, the feature vector for indicating image content can be extracted from picture classification deep neural network model later,
The feature vector of extraction is determined as first eigenvector.Similarly, the picture that second picture is concentrated is input to picture classification
In deep neural network model, you can extracted from picture classification deep neural network model indicate image content feature to
The feature vector of extraction is determined as second feature vector by amount.
In conclusion by obtaining corresponding first pictures of first object word and second picture collection, according to the first figure
Piece collection and second picture concentrate the content of each picture determine the corresponding first eigenvector of first object word and second feature to
Amount can determine first eigenvector and second feature vector according to image content, to realize the escape identification of multi-angle, improving
The coverage rate of escape identification lays the foundation.
It should be noted that the side of aforementioned determining first object word corresponding first eigenvector and second feature vector
Method is also applied for determining the corresponding third feature vector sum fourth feature vector of the second target word, and the application is no longer directed to true
The mode of the corresponding third feature vector sum fourth feature vector of fixed second target word is described in detail.
In addition, the method for the determination feature vector described in above-described embodiment can be used alone, can also be used in combination, this
Application is not construed as limiting the method for determination of feature vector.When determining feature vector using aforementioned at least two methods,
It may be implemented to carry out escape identification from different angles, increase the basis for estimation of escape identification, be conducive to improve escape identification
Coverage rate.
Applicants experimentally found that in the case where ensureing accuracy rate, the escape recognition methods of the embodiment of the present application
Coverage rate is obviously improved.
In search process scene, according to transfer recognition methods provided by the present application, escape identification is carried out to word, with
The search result for meeting user's query intention is provided a user, the search experience of user is promoted.So, the application determine turn
After adopted probability, shown after can also being ranked up to search result according to escape probability.Fig. 5 is provided by the embodiment of the present application
Another flow diagram of escape recognition methods.
As shown in figure 5, the escape recognition methods may comprise steps of：
Step 501, according to query statement and candidate result, first object word to be identified and the second target word are determined
Language.
In the present embodiment, after user input query sentence, search engine can be obtained first according to query statement input by user
It takes candidate result, then is directed to each candidate result text, after being segmented, going the pretreatments such as stop words, from pretreated
The word of appearance adjacent with query statement input by user is obtained in candidate result text, and query statement input by user is determined
For first object word, using the word of the appearance adjacent with search statement obtained from candidate result text as the second target word
Language.
In a kind of possible realization method of the embodiment of the present application, when obtained from candidate result text and query statement
When the word of adjacent appearance is multiple, the number that the word of each adjacent appearance occurs in candidate result text can be counted, will be gone out
The most word of occurrence number is determined as the second target word.
Step 502, the corresponding first eigenvector of first object word and second feature vector and the second target word are determined
The corresponding third feature vector of language and fourth feature vector.
Wherein, first eigenvector is related to the second target word, and second feature vector is unrelated with the second target word, the
Three feature vectors are related to first object word, and fourth feature vector is unrelated with first object word.
Step 503, at a distance from according to first eigenvector between second feature vector and third feature and fourth feature to
Distance between amount determines escape probability when first object word and the second target word combination.
Herein it should be noted that in the present embodiment, the description to step 502- steps 503 may refer to aforementioned implementation
Description in example to step 102- steps 103, details are not described herein again.
Escape probability when step 504, according to first object word and the second target word combination, determines candidate result
Display order.
In the present embodiment, it is determined that, can be with after escape probability when first object word and the second target word combination
Candidate result is ranked up according to escape probability, to determine the display order of candidate result, search engine is according to display order
The corresponding search result of search statement is shown to user.Since escape probability is higher, the first search statement and the second search statement
It is bigger to be combined into the current possibility that escape occurs, it therefore, can be according to escape probability, according to escape probability in the present embodiment
Sequence from low to high is ranked up candidate result, so that search engine preferentially shows the low search result of escape possibility.
The escape recognition methods of the present embodiment, by determining first object to be identified according to query statement and candidate result
Escape can be identified and is limited within the scope of candidate result by word and the second target word so that escape identification has specific aim,
Reduce data processing amount when escape identification；By general according to first object word and escape when the second target word combination
Rate determines the display order of candidate result, enables to search engine preferentially to show the low search result of escape probability, ensures to search
The matching degree of the query intention of hitch fruit and user, improves the accuracy of search result.
In order to realize that above-described embodiment, the application also propose a kind of escape identification device.
A kind of structural schematic diagram for escape identification device that Fig. 6 is provided by the embodiment of the present application.
As shown in fig. 6, the escape identification device 50 may include：Acquisition module 510, determining module 520 and escape are general
Rate determining module 530.Wherein,
Acquisition module 510, for obtaining first object word to be identified and the second target word.
Determining module 520, for determining the corresponding first eigenvector of first object word and second feature vector and the
The corresponding third feature vector of two target words and fourth feature vector；Wherein, first eigenvector and the second target word phase
It closes, second feature vector is unrelated with the second target word, and third feature vector is related to first object word, fourth feature vector
It is unrelated with first object word.
Escape probability determination module 530, for according to first eigenvector between second feature vector at a distance from and third
Feature between fourth feature vector at a distance from, determine escape probability when first object word and the second target word combination.
Specifically, escape probability determination module 530 be used for when first eigenvector between second feature vector at a distance from it is small
In first threshold, and third feature between fourth feature vector at a distance from when being more than second threshold, it is determined that first object word
It is more than third threshold value with escape probability when the second target word combination；Alternatively, when first eigenvector and second feature vector
Between distance be more than second threshold, and third feature between fourth feature vector at a distance from when being less than first threshold, it is determined that the
Escape probability when one target word and the second target word combination is more than third threshold value.Wherein, first threshold is less than or equal to
Second threshold.
Further, in a kind of possible realization method of the embodiment of the present application, as shown in fig. 7, implementing as shown in Figure 6
On the basis of example, determining module 520 includes：
Co-occurrence word set acquiring unit 5201 is crawled for carrying out data to network, obtains first object word corresponding the
One co-occurrence word set and the second co-occurrence word set, wherein the first co-occurrence word concentration includes the second target word.
First determination unit 5202 determines first object word for concentrating each co-occurrence word for including according to the first co-occurrence word
The corresponding first eigenvector of language；And each co-occurrence word for including is concentrated according to the second co-occurrence word, determine first object word pair
The second feature vector answered.
By crawling network data, obtain first object word it is corresponding include the second target word the first co-occurrence word set
Do not include the second co-occurrence word set of the second target word, and then includes according to the first co-occurrence word set and the second co-occurrence word concentration
Each co-occurrence word determines the corresponding first eigenvector of first object word and second feature vector respectively, can be from co-occurrence word angle
Degree determines the corresponding first eigenvector of first object word and second feature vector, to realize the escape identification of multi-angle, carrying
The coverage rate of high escape identification lays the foundation.
In a kind of possible realization method of the embodiment of the present application, as shown in figure 8, on the basis of embodiment as shown in Figure 6
On, determining module 520 includes：
Page set acquiring unit 5211 is crawled for carrying out data to network, and acquisition includes the first of first object word
Page set and second page collection, wherein it includes the second target word that first page, which concentrates at least one page,.
Second determination unit 5212, the attribute information for concentrating each page according to first page determine first object word
The corresponding first eigenvector of language, wherein the attribute information of each page includes but not limited to type or each page institute of each page
Belong to the type of website；And the attribute information of each page is concentrated according to second page, determine first object word corresponding second
Feature vector.
It is crawled by carrying out data to network, obtains the first page collection and second page collection for including first object word,
It includes the second target word that first page, which concentrates at least one page, and then concentrates each page according to first page collection and second page
The attribute information in face determines the corresponding first eigenvector of first object word and second feature vector, can be according to be identified
The type of the affiliated page of word determines the first eigenvector and second feature vector of first object word, to realize multi-angle
Escape identification, the coverage rate for improving escape identification lay the foundation.
In a kind of possible realization method of the embodiment of the present application, as shown in figure 9, on the basis of embodiment as shown in Figure 6
On, determining module 520 includes：
Pictures acquiring unit 5221 is the first pictures and second picture for obtaining corresponding with first object word
Collection, wherein the picture in the first pictures at least one picture pictures corresponding with the second target word is identical.
Third determination unit 5222 determines first object word pair for the content according to each picture in the first pictures
The first eigenvector answered；And the content of each picture is concentrated according to second picture, determine first object word corresponding second
Feature vector.
By obtaining corresponding first pictures of first object word and second picture collection, according to the first pictures and second
The content of each picture determines the corresponding first eigenvector of first object word and second feature vector in pictures, being capable of basis
Image content determines first eigenvector and second feature vector, to realize the escape identification of multi-angle, improving escape identification
Coverage rate lays the foundation.
It should be noted that determining module 520 determines the corresponding first eigenvector of first object word and second feature
The mode of vector is also applied for determining the corresponding third feature vector sum fourth feature vector of the second target word, and the application is to true
Cover half block 520 determines that the process of the corresponding third feature vector sum fourth feature vector of the second target word repeats no more.
In addition, determining module 520 can determine feature vector only with a kind of mode, various ways can also be used to determine
Feature vector, the application determine that the mode of feature vector is not construed as limiting to determining module 520.When determining module 520 is using at least
It when two ways determines feature vector, may be implemented to carry out escape identification from different angles, increase the judgement of escape identification
Foundation is conducive to the coverage rate for improving escape identification.
In a kind of possible realization method of the embodiment of the present application, as shown in Figure 10, on the basis of embodiment as shown in Figure 6
On, which can also include：
Display order determining module 540, for general according to first object word and escape when the second target word combination
Rate determines the display order of candidate result.
In the present embodiment, acquisition module 510 is specifically used for, according to query statement and candidate result, determining to be identified first
Target word and the second target word.
It, can be with by determining first object word to be identified and the second target word according to query statement and candidate result
Escape identification is limited within the scope of candidate result so that escape identification has specific aim, at data when reducing escape identification
Reason amount；Escape probability when by according to first object word and the second target word combination, determines that the display of candidate result is suitable
Sequence enables to search engine preferentially to show the low search result of escape probability, ensures the query intention of search result and user
Matching degree, improve the accuracy of search result.
It should be noted that the aforementioned escape for being also applied for the embodiment to the explanation of escape recognition methods embodiment
Identification device, realization principle is similar, and details are not described herein again.
The escape identification device of the present embodiment, by obtaining first object word and the second target word to be identified, and
Determine first object word and the relevant first eigenvector of the second target word and unrelated with the second target word second
Feature vector, and determine the second target word and the relevant third feature vector sum of first object word and first object word
Unrelated fourth feature vector, so according to the distance between first eigenvector and second feature vector and third feature to
The distance between amount and fourth feature vector, determine escape probability when first object word and the second target word combination.By
This, is by the influence according to two words to feature vector each other, whether escape occurs when determining two word combinations, to
The accuracy and reliability of escape identification is improved, and then improves the accuracy of search result.
In order to realize that above-described embodiment, the application also propose a kind of computer equipment, including：Processor and memory.Its
In, processor runs journey corresponding with executable program code by reading the executable program code stored in memory
Sequence, for realizing escape recognition methods as in the foregoing embodiment.
Figure 11 is the structural schematic diagram for the computer equipment that one embodiment of the application proposes, is shown suitable for being used for realizing this
Apply for the block diagram of the exemplary computer device 90 of embodiment.The computer equipment 90 that Figure 11 is shown is only an example,
Any restrictions should not be brought to the function and use scope of the embodiment of the present application.
As shown in figure 11, computer equipment 90 is showed in the form of general purpose computing device.The component of computer equipment 90
It can include but is not limited to：One or more processor or processing unit 906, system storage 910 connect different system
The bus 908 of component (including system storage 910 and processing unit 906).
Bus 908 indicates one or more in a few class bus structures, including memory bus or Memory Controller,
Peripheral bus, graphics acceleration port, processor or the local bus using the arbitrary bus structures in a variety of bus structures.It lifts
For example, these architectures include but not limited to industry standard architecture (Industry Standard
Architecture；Hereinafter referred to as：ISA) bus, microchannel architecture (Micro Channel Architecture；Below
Referred to as：MAC) bus, enhanced isa bus, Video Electronics Standards Association (Video Electronics Standards
Association；Hereinafter referred to as：VESA) local bus and peripheral component interconnection (Peripheral Component
Interconnection；Hereinafter referred to as：PCI) bus.
Computer equipment 90 typically comprises a variety of computer system readable media.These media can be it is any can be by
The usable medium that computer equipment 90 accesses, including volatile and non-volatile media, moveable and immovable medium.
System storage 610 may include the computer system readable media of form of volatile memory, such as deposit at random
Access to memory (Random Access Memory；Hereinafter referred to as：RAM) 911 and/or cache memory 912.Computer is set
Standby 90 may further include other removable/nonremovable, volatile/non-volatile computer system storage mediums.Only
As an example, storage system 913 can be used for reading and writing immovable, non-volatile magnetic media (Figure 11 do not show, commonly referred to as
" hard disk drive ").Although being not shown in Figure 11, can provide for reading removable non-volatile magnetic disk (such as " floppy disk ")
The disc driver write, and to removable anonvolatile optical disk (such as：Compact disc read-only memory (Compact Disc Read
Only Memory；Hereinafter referred to as：CD-ROM), digital multi CD-ROM (Digital Video Disc Read Only
Memory；Hereinafter referred to as：DVD-ROM) or other optical mediums) read-write CD drive.In these cases, each driving
Device can be connected by one or more data media interfaces with bus 908.System storage 910 may include at least one
There is one group of (for example, at least one) program module, these program modules to be configured to perform this for program product, the program product
Apply for the function of each embodiment.
Computer-readable signal media may include in a base band or as the data-signal that a carrier wave part is propagated,
Wherein carry computer-readable program code.Diversified forms may be used in the data-signal of this propagation, including --- but
It is not limited to --- electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be
Any computer-readable medium other than computer readable storage medium, which can send, propagate or
Transmission for by instruction execution system, device either device use or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In --- wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
Can with one or more programming languages or combinations thereof come write for execute the application operation computer
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
Further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partly executes or executed on a remote computer or server completely on the remote computer on the user computer.
Program/utility 914 with one group of (at least one) program module 9140 can be stored in such as system and deposit
In reservoir 610, such program module 9140 includes but not limited to operating system, one or more application program, Qi Tacheng
Sequence module and program data may include the realization of network environment in each or certain combination in these examples.Program
Module 9140 usually executes function and/or method in embodiments described herein.
Computer equipment 90 can also be with one or more external equipments 10 (such as keyboard, sensing equipment, display 100
Deng) communication, can also be enabled a user to one or more equipment interact with the terminal device 90 communicate, and/or with make
Any equipment that the computer equipment 90 can be communicated with one or more of the other computing device (such as network interface card, modulation /demodulation
Device etc.) communication.This communication can be carried out by input/output (I/O) interface 902.Also, computer equipment 90 can be with
Pass through network adapter 900 and one or more network (such as LAN (Local Area Network；Hereinafter referred to as：
LAN), wide area network (Wide Area Network；Hereinafter referred to as：WAN) and/or public network, for example, internet) communication.Such as figure
Shown in 11, network adapter 900 is communicated by bus 908 with other modules of computer equipment 90.Although should be understood that Figure 11
In be not shown, can in conjunction with computer equipment 90 use other hardware and/or software module, including but not limited to：Microcode is set
Standby driver, redundant processing unit, external disk drive array, RAID system, tape drive and data backup storage system
System etc..
Processing unit 906 is stored in program in system storage 910 by operation, to perform various functions using with
And data processing, such as realize the escape recognition methods referred in previous embodiment.
In order to realize that above-described embodiment, the application also propose a kind of non-transitorycomputer readable storage medium, deposit thereon
Computer program is contained, when which is executed by processor, realizes escape recognition methods as in the foregoing embodiment.
In order to realize that above-described embodiment, the application also propose a kind of computer program product, when the computer program produces
When instruction in product is executed by processor, escape recognition methods as in the foregoing embodiment is realized.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is contained at least one embodiment or example of the application.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiments or example.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
In addition, term " first ", " second " are used for description purposes only, it is not understood to indicate or imply relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present application, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing custom logic function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the application includes other realization, wherein can not press shown or discuss suitable
Sequence, include according to involved function by it is basic simultaneously in the way of or in the opposite order, to execute function, this should be by the application
Embodiment person of ordinary skill in the field understood.
Expression or logic and/or step described otherwise above herein in flow charts, for example, being considered use
In the order list for the executable instruction for realizing logic function, may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (system of such as computer based system including processor or other can be held from instruction
The instruction fetch of row system, device or equipment and the system executed instruction) it uses, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium " can any can be included, store, communicating, propagating or passing
Defeated program is for instruction execution system, device or equipment or the dress used in conjunction with these instruction execution systems, device or equipment
It sets.The more specific example (non-exhaustive list) of computer-readable medium includes following：Electricity with one or more wiring
Interconnecting piece (electronic device), portable computer diskette box (magnetic device), random access memory (RAM), read-only memory
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, computer-readable medium can even is that the paper that can print described program on it or other are suitable
Medium, because can be for example by carrying out optical scanner to paper or other media, then into edlin, interpretation or when necessary with it
His suitable method is handled electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each section of the application can be realized with hardware, software, firmware or combination thereof.Above-mentioned
In embodiment, software that multiple steps or method can in memory and by suitable instruction execution system be executed with storage
Or firmware is realized.Such as, if realized in another embodiment with hardware, following skill well known in the art can be used
Any one of art or their combination are realized：With for data-signal realize logic function logic gates from
Logic circuit is dissipated, the application-specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile
Journey gate array (FPGA) etc..
Those skilled in the art are appreciated that realize all or part of step that above-described embodiment method carries
Suddenly it is that relevant hardware can be instructed to complete by program, the program can be stored in a kind of computer-readable storage medium
In matter, which includes the steps that one or a combination set of embodiment of the method when being executed.
In addition, each functional unit in each embodiment of the application can be integrated in a processing module, it can also
That each unit physically exists alone, can also two or more units be integrated in a module.Above-mentioned integrated mould
The form that hardware had both may be used in block is realized, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized in the form of software function module and when sold or used as an independent product, can also be stored in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only memory, disk or CD etc..Although having been shown and retouching above
Embodiments herein is stated, it is to be understood that above-described embodiment is exemplary, and should not be understood as the limit to the application
System, those skilled in the art can be changed above-described embodiment, change, replace and become within the scope of application
Type.
Claims (11)
1. a kind of escape recognition methods, which is characterized in that including：
Obtain first object word to be identified and the second target word；
Determine the corresponding first eigenvector of the first object word and second feature vector and the second target word pair
The third feature vector and fourth feature vector answered；Wherein, the first eigenvector is related to the second target word, the
Two feature vectors are unrelated with the second target word, and third feature vector is related to the first object word, fourth feature
It is vectorial unrelated with the first object word；
According to the first eigenvector between the second feature vector at a distance from and the third feature and the described 4th special
Distance between sign vector, determines the first object word and escape probability when the second target word combination.
2. the method as described in claim 1, which is characterized in that the corresponding fisrt feature of the determination first object word to
Amount and second feature vector, including：
Data are carried out to network to crawl, and obtain the corresponding first co-occurrence word set of the first object word and the second co-occurrence word set,
The wherein described first co-occurrence word concentration includes the second target word；
According to first co-occurrence word concentrate include each co-occurrence word, determine the corresponding fisrt feature of the first object word to
Amount；
According to second co-occurrence word concentrate include each co-occurrence word, determine the corresponding second feature of the first object word to
Amount.
3. the method as described in claim 1, which is characterized in that the corresponding fisrt feature of the determination first object word to
Amount and second feature vector, including：
Data are carried out to network to crawl, and obtain the first page collection and second page collection for including the first object word, wherein
It includes the second target word that the first page, which concentrates at least one page,；
The attribute information that each page is concentrated according to the first page, determine the corresponding fisrt feature of the first object word to
Amount；
The attribute information that each page is concentrated according to the second page, determine the corresponding second feature of the first object word to
Amount.
4. method as claimed in claim 3, which is characterized in that the attribute information of each page, including：The type of each page
Or the type of each affiliated website of the page.
5. the method as described in claim 1, which is characterized in that the corresponding fisrt feature of the determination first object word to
Amount and second feature vector, including：
It is the first pictures and second picture collection to obtain corresponding with the first object word, wherein at least one in the first pictures
Picture in a picture pictures corresponding with the second target word is identical；
According to the content of each picture in first pictures, the corresponding first eigenvector of the first object word is determined；
The content that each picture is concentrated according to the second picture determines the corresponding second feature vector of the first object word.
6. method as described in any one in claim 1-5, which is characterized in that it is described obtain first object word to be identified and
Second target word, including：
According to query statement and candidate result, the first object word to be identified and the second target word are determined.
7. method as claimed in claim 6, which is characterized in that the determination first object word and second target
After escape probability when word combination, further include：
Escape probability when according to the first object word and the second target word combination, determines the candidate result
Display order.
8. method as described in any one in claim 1-5, which is characterized in that the determination first object word with it is described
Escape probability when the second target word combination, including：
If the first eigenvector between the second feature vector at a distance from be less than first threshold, and the third feature with
Distance between the fourth feature vector is more than second threshold, it is determined that the first object word and the second target word
Escape probability when combination is more than third threshold value；
Alternatively,
If the first eigenvector between the second feature vector at a distance from be more than second threshold, and the third feature with
Distance between the fourth feature vector is less than first threshold, it is determined that the first object word and the second target word
Escape probability when combination is more than third threshold value；
Wherein, the first threshold is less than or equal to the second threshold.
9. a kind of escape identification device, which is characterized in that including：
Acquisition module, for obtaining first object word to be identified and the second target word；
Determining module, for determining that the corresponding first eigenvector of the first object word and second feature are vectorial and described
The corresponding third feature vector of second target word and fourth feature vector；Wherein, the first eigenvector and described second
Target word is related, and second feature vector is unrelated with the second target word, third feature vector and the first object word
Language is related, and fourth feature vector is unrelated with the first object word；
Escape probability determination module, for according to the first eigenvector between the second feature vector at a distance from and institute
State third feature between the fourth feature vector at a distance from, determine the first object word and the second target word group
Escape probability when conjunction.
10. a kind of computer equipment, which is characterized in that including processor and memory；
Wherein, the processor can perform to run with described by reading the executable program code stored in the memory
The corresponding program of program code, for realizing the escape recognition methods as described in any one of claim 1-8.
11. a kind of non-transitorycomputer readable storage medium, is stored thereon with computer program, which is characterized in that the program
The escape recognition methods as described in any one of claim 1-8 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810367116.7A CN108595620B (en) | 2018-04-23 | 2018-04-23 | Escape identification method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810367116.7A CN108595620B (en) | 2018-04-23 | 2018-04-23 | Escape identification method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108595620A true CN108595620A (en) | 2018-09-28 |
| CN108595620B CN108595620B (en) | 2022-04-26 |
Family
ID=63614014
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810367116.7A Active CN108595620B (en) | 2018-04-23 | 2018-04-23 | Escape identification method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108595620B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150172714A1 (en) * | 2012-07-09 | 2015-06-18 | Mediatek Inc. | METHOD AND APPARATUS of INTER-VIEW SUB-PARTITION PREDICTION in 3D VIDEO CODING |
| CN106328147A (en) * | 2016-08-31 | 2017-01-11 | 中国科学技术大学 | Speech recognition method and device |
| CN106547871A (en) * | 2016-10-31 | 2017-03-29 | 北京百度网讯科技有限公司 | Method and apparatus is recalled based on the Search Results of neutral net |
| CN106815252A (en) * | 2015-12-01 | 2017-06-09 | 阿里巴巴集团控股有限公司 | A kind of searching method and equipment |
| CN107229684A (en) * | 2017-05-11 | 2017-10-03 | 合肥美的智能科技有限公司 | Statement classification method, system, electronic equipment, refrigerator and storage medium |
-
2018
- 2018-04-23 CN CN201810367116.7A patent/CN108595620B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150172714A1 (en) * | 2012-07-09 | 2015-06-18 | Mediatek Inc. | METHOD AND APPARATUS of INTER-VIEW SUB-PARTITION PREDICTION in 3D VIDEO CODING |
| CN106815252A (en) * | 2015-12-01 | 2017-06-09 | 阿里巴巴集团控股有限公司 | A kind of searching method and equipment |
| CN106328147A (en) * | 2016-08-31 | 2017-01-11 | 中国科学技术大学 | Speech recognition method and device |
| CN106547871A (en) * | 2016-10-31 | 2017-03-29 | 北京百度网讯科技有限公司 | Method and apparatus is recalled based on the Search Results of neutral net |
| CN107229684A (en) * | 2017-05-11 | 2017-10-03 | 合肥美的智能科技有限公司 | Statement classification method, system, electronic equipment, refrigerator and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108595620B (en) | 2022-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Katsurai et al. | Image sentiment analysis using latent correlations among visual, textual, and sentiment views | |
| Unar et al. | Visual and textual information fusion using Kernel method for content based image retrieval | |
| CN108009293A (en) | Video tab generation method, device, computer equipment and storage medium | |
| WO2017139764A1 (en) | Zero-shot event detection using semantic embedding | |
| Lee et al. | Tag refinement in an image folksonomy using visual similarity and tag co-occurrence statistics | |
| CN108563655A (en) | Text based event recognition method and device | |
| CN110458078A (en) | A kind of face image data clustering method, system and equipment | |
| CN111241230A (en) | Method and system for identifying string mark risk based on text mining | |
| CN108460098A (en) | Information recommendation method, device and computer equipment | |
| CN107861948B (en) | Label extraction method, device, equipment and medium | |
| Wang et al. | Task-generic semantic convolutional neural network for web text-aided image classification | |
| CN111475603A (en) | Enterprise identifier identification method and device, computer equipment and storage medium | |
| CN109815500A (en) | Management method, device, computer equipment and the storage medium of unstructured official document | |
| CN111783903A (en) | Text processing method, text model processing method and device and computer equipment | |
| CN112015928A (en) | Information extraction method and device of multimedia resource, electronic equipment and storage medium | |
| Zayed et al. | Phrase-level metaphor identification using distributed representations of word meaning | |
| Tian et al. | Image classification based on the combination of text features and visual features | |
| Nemade et al. | Image segmentation using convolutional neural network for image annotation | |
| Li et al. | Social context-aware person search in videos via multi-modal cues | |
| Yao | Key frame extraction method of music and dance video based on multicore learning feature fusion | |
| Amorim et al. | Novelty detection in social media by fusing text and image into a single structure | |
| Tüselmann et al. | Recognition-free question answering on handwritten document collections | |
| Verma et al. | Script identification in natural scene images: a dataset and texture-feature based performance evaluation | |
| Qiu et al. | Improving news video annotation with semantic context | |
| CN111859898B (en) | Hidden association network-based multi-domain text implicit feature extraction method and computer storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |