WO2017107651A1 - 确定新闻之间相关性、多新闻之间相关性计算方法和装置 - Google Patents

确定新闻之间相关性、多新闻之间相关性计算方法和装置 Download PDF

Info

Publication number
WO2017107651A1
WO2017107651A1 PCT/CN2016/103397 CN2016103397W WO2017107651A1 WO 2017107651 A1 WO2017107651 A1 WO 2017107651A1 CN 2016103397 W CN2016103397 W CN 2016103397W WO 2017107651 A1 WO2017107651 A1 WO 2017107651A1
Authority
WO
WIPO (PCT)
Prior art keywords
news
benchmark
vector
association relationship
distance
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.)
Ceased
Application number
PCT/CN2016/103397
Other languages
English (en)
French (fr)
Inventor
张伸正
魏少俊
陈培军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Qihoo Technology Co Ltd
Qizhi Software Beijing Co Ltd
Original Assignee
Beijing Qihoo Technology Co Ltd
Qizhi Software Beijing Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Priority claimed from CN201510974431.2A external-priority patent/CN105630766B/zh
Priority claimed from CN201510974316.5A external-priority patent/CN105528335B/zh
Application filed by Beijing Qihoo Technology Co Ltd, Qizhi Software Beijing Co Ltd filed Critical Beijing Qihoo Technology Co Ltd
Priority to US15/744,688 priority Critical patent/US10217025B2/en
Publication of WO2017107651A1 publication Critical patent/WO2017107651A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/30Semantic analysis
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/3331Query processing
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/22Matching criteria, e.g. proximity measures
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/40Document-oriented image-based pattern recognition
    • G06V30/41Analysis of document content
    • G06V30/414Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/40Document-oriented image-based pattern recognition
    • G06V30/41Analysis of document content
    • G06V30/416Extracting the logical structure, e.g. chapters, sections or page numbers; Identifying elements of the document, e.g. authors

Definitions

  • the present invention relates to the field of computer technology, and in particular, to a method and apparatus for determining correlation between news, and a method and apparatus for calculating correlation between multiple news.
  • the present invention has been made in order to provide a method and apparatus for determining correlation between news, which overcomes the above problems or at least partially solves the above problems, and a method and apparatus for calculating correlation between multiple news.
  • a method for determining correlation between news items includes: comparing a first news with a benchmark news to obtain a distance between the first news and the benchmark news; and a second news Comparing with the benchmark news, obtaining a distance between the second news and the benchmark news; calculating a distance between the first news and the benchmark news and a distance between the first news and the benchmark news a distance difference, and determining a correlation between the first news and the second news according to the distance difference.
  • an apparatus for determining a correlation between news comprising: a first distance calculation module, configured to compare a first news with a benchmark news, to obtain the first news and the a distance of the benchmark news; a second distance calculation module, configured to compare the second news with the benchmark news, to obtain a distance between the second news and the benchmark news; and a correlation calculation module, configured to calculate the first a distance difference between a news and a distance of the benchmark news and a distance between the first news and the benchmark news, and determining a correlation between the first news and the second news according to the distance difference .
  • the method and apparatus for determining the correlation between news of the present invention have at least the following advantages:
  • the technical solution of the present invention when it is necessary to analyze the correlation between different news, it is not necessary to perform multiple The contrast between news, but the comparison between multiple news and benchmarking news, if the distance between the two news and the benchmark is similar, it means that there is a certain degree of similarity between the two news; because there is only one benchmark news However, other news does not need to be compared with each other, and only needs to be compared with benchmark news to determine the correlation between multiple news, so the efficiency of obtaining related news according to the technical solution of the present invention is very high.
  • a multi-news correlation calculation method including: determining a first-level association relationship between a first news and a second news; and determining a level of the second news and the third news. Correlating relationship; establishing the first news and the third news according to the first-level association relationship between the first news and the second news, the first-level association relationship between the second news and the third news A secondary relationship between the two.
  • a multi-news correlation calculation device including: a first association determination module, configured to determine a first-level association relationship between the first news and the second news; and a second association relationship Determining a module, determining a first-level association relationship between the second news and the third news; a third association relationship determining module, according to the first-level association relationship between the first news and the second news, the second news and a first-level association relationship of the third news, establishing a secondary association relationship between the first news and the third news.
  • the multi-news correlation calculation method and apparatus of the present invention have at least the following advantages:
  • the first-level association relationship between the first news and the second news reflects the degree of similarity between the first news and the second news, and between the second news and the third news.
  • the first-level relationship reflects the degree of similarity between the first news and the second news, but the calculated second-level relationship between the first news and the third news is no longer limited by the literal content, because The second news is similar to the first news literally, and the second news and the third news literally different parts, which actually reflect the semantic similarity between the first news and the third news, so the establishment according to the present invention Secondary associations can find news that is literal but similar in semantics.
  • a computer program comprising computer readable code, when the computer readable code is run on a computing device, causing the computing device to perform a determination of news according to any of the above Inter-correlation methods, and/or multi-news correlation calculation methods.
  • a computer readable medium storing a computer program as described above is provided.
  • FIG. 1 shows a flow chart of a method of determining correlation between news in accordance with one embodiment of the present invention
  • FIG. 2 shows a block diagram of an apparatus for determining correlation between news in accordance with one embodiment of the present invention
  • FIG. 3 shows a block diagram of an apparatus for determining correlation between news items, in accordance with one embodiment of the present invention
  • FIG. 4 is a flow chart showing a method of calculating a correlation between multiple news according to an embodiment of the present invention
  • FIG. 5 is a flow chart showing a method of calculating a correlation between multiple news according to an embodiment of the present invention
  • FIG. 6 is a schematic diagram showing a method of calculating a correlation between multiple news according to an embodiment of the present invention
  • FIG. 7 is a block diagram showing a multi-news correlation calculation device according to an embodiment of the present invention.
  • FIG. 8 is a block diagram showing a multi-news correlation calculation device according to an embodiment of the present invention.
  • FIG. 9 is a block diagram schematically showing a computing device for performing a method of determining correlation between news according to the present invention, and/or a multi-news correlation calculation method;
  • Fig. 10 schematically shows a storage unit for holding or carrying a program code for realizing a correlation between news according to the present invention, and/or a multi-news correlation calculation method.
  • an embodiment of the present invention provides a method for determining correlation between news, including:
  • step 110 the first news is compared with the benchmark news to obtain the distance between the first news and the benchmark news.
  • the benchmark news there is no restriction on the benchmark news, and any news can be selected as the benchmark news.
  • step 120 the second news is compared with the benchmark news to obtain the distance between the second news and the benchmark news.
  • Step 130 Calculate a distance difference between the distance between the first news and the benchmark news and the distance between the first news and the benchmark news, and determine the correlation between the first news and the second news according to the distance difference.
  • the distance reflects the difference between the news, and the embodiment does not limit the manner of calculating the distance; according to the technical solution of the embodiment, it is easy to understand, when the distance between the first news and the benchmark news, When the distance between the second news and the benchmark news is similar, it indicates that the distance between the first news and the second news is relatively close, that is, the correlation between the two news is strong; when the first news and the benchmark news When the distance between the second news and the benchmark news is relatively long, the distance between the first news and the second news is far. The correlation between the two news is weak.
  • a method for determining a correlation between news items is provided.
  • the method for determining the correlation between news items according to the foregoing embodiment further includes, before step 110, identifying: The type of the first news, and select the benchmark news with the corresponding type from the benchmark news collection.
  • this embodiment if the distance between the first news, the second news, and the benchmark news is too large, it can only indicate that the first news, the second news, and the benchmark news are all very different, but it is difficult to explain the first news, What is the correlation between the second news. While there is a high correlation between the same types of news, this embodiment makes the distance between the first news and the benchmark news smaller, indicating that the first news has a higher relevance to the benchmark news, and the second news and the benchmark The large news distance is equivalent to the distance from the first news, that is, the first news and the second news are weakly related. The second news and the benchmark news are smaller than the first news, that is, the first news and the first news. Second, the news is more relevant. For example, if the first news is sports news, the selected benchmark news is sports news.
  • a method for determining the correlation between the news is further provided.
  • this embodiment if the distance between the first news, the second news, and the benchmark news is too large, it can only indicate that the first news, the second news, and the benchmark news are all very different, but it is difficult to explain the first news, What is the correlation between the second news. If there is a high correlation between the news with the same keywords, this embodiment makes the distance between the first news and the benchmark news smaller, indicating that the first news has a higher relevance to the benchmark news, and the second news.
  • the distance from the benchmark news is equivalent to the distance from the first news, that is, the correlation between the first news and the second news is weak, and the distance between the second news and the benchmark news is equivalent to the distance from the first news, that is, the first news. Strong correlation with the second news. For example, if the first news headline is "Star A Award", the selected benchmark news may be "Star A Full Record” and the keyword is Star A.
  • a method for determining the correlation between the news is further provided.
  • the step 110 of the method for determining the correlation between the news in the embodiment includes: acquiring the first A feature attribute of the news, and generating a vector corresponding to the first news according to the feature attribute of the first news, and comparing the vector corresponding to the first news with the vector corresponding to the benchmark news.
  • the step 120 in the method for determining the correlation between the news in the embodiment includes: acquiring the feature attribute of the second news, and generating a vector corresponding to the second news according to the feature attribute of the second news, and corresponding to the second news The vector is compared to the vector corresponding to the benchmark news.
  • the feature attribute is not limited, and the news is easily quantized by using one or more feature attributes of the news, and the distance between the news can be calculated more easily and accurately.
  • An embodiment of the present invention also provides a method of determining correlation between news, as compared to the foregoing
  • the step 110 in the method for determining the correlation between the news specifically includes: segmenting the first news to obtain a plurality of words, and calculating a word frequency of the plurality of words of the first news, as the first news.
  • Feature attribute The step 120 in the method for determining the correlation between the news in the embodiment specifically includes: segmenting the second news to obtain a plurality of words, and calculating a word frequency of the plurality of words of the second news as the feature attribute of the second news.
  • a random number may be assigned to the first news, and an article vector is constructed for the first news according to the random number and the calculated word frequency; similarly, the second news and the benchmark news may also construct corresponding articles.
  • the minimum hash value of the article vector of the first news, the second news, and the benchmark news is calculated, and the specific distance value between the first news and the benchmark news, the second news and The distance between the benchmark news.
  • the distance between News A, News B and benchmark news is 4 and 3 respectively.
  • a small difference of 1 means that News A and News B are highly correlated, and News C and Benchmark News have a difference of 1. The difference between the distances from News A and News B is large, and the correlation is low.
  • a method for determining the correlation between news items includes: When the difference is in the preset interval, the second news is set as the related news of the first news for pushing the second news when the related news of the first news needs to be pushed.
  • the preset interval is 0-1
  • the news A is the news related to the news B
  • the user can push the news B for the user when viewing the news A.
  • an embodiment of the present invention provides an apparatus for determining correlation between news, including:
  • the first distance calculation module 210 is adapted to compare the first news with the benchmark news to obtain the distance between the first news and the benchmark news. In this embodiment, there is no restriction on the benchmark news, and any news can be selected as the benchmark news.
  • the second distance calculation module 220 is adapted to compare the second news with the benchmark news to obtain the distance between the second news and the benchmark news.
  • the correlation calculation module 230 is adapted to calculate a distance difference between the distance between the first news and the benchmark news and the distance between the first news and the benchmark news, and determine the correlation between the first news and the second news according to the distance difference.
  • the distance reflects the difference between the news, and the embodiment does not limit the manner of calculating the distance; according to the technical solution of the embodiment, it is easy to understand, when the distance between the first news and the benchmark news, When the distance between the second news and the benchmark news is similar, it indicates that the distance between the first news and the second news is relatively close, that is, the correlation between the two news is strong; when the first news and the benchmark news The distance between the second news and the benchmark news is relatively far, indicating that the distance between the first news and the second news is far, and the correlation between the two news is weak.
  • an embodiment of the present invention further provides an apparatus for determining correlation between news
  • the apparatus for determining the correlation between news in the embodiment further includes:
  • the first benchmark news acquisition module 310 is adapted to identify the type of the first news and select a benchmark news with a corresponding type from the set of benchmark news.
  • this embodiment if the distance between the first news, the second news, and the benchmark news is too large, it can only indicate that the first news, the second news, and the benchmark news are all very different, but it is difficult to explain the first news, What is the correlation between the second news. While there is a high correlation between the same types of news, this embodiment makes the distance between the first news and the benchmark news smaller, indicating that the first news has a higher relevance to the benchmark news, and the second news and the benchmark The large news distance is equivalent to the distance from the first news, that is, the first news and the second news are weakly related. The second news and the benchmark news are smaller than the first news, that is, the first news and the first news. Second, the news is more relevant. For example, if the first news is sports news, the selected benchmark news is sports news.
  • an apparatus for determining correlation between news items is provided in an embodiment of the present invention.
  • the device for determining the correlation between news items according to the foregoing embodiment further includes:
  • the second benchmark news acquisition module 320 is adapted to obtain keywords in the first news, and select benchmark news with keywords from the set of benchmark news.
  • this embodiment if the distance between the first news, the second news, and the benchmark news is too large, it can only indicate that the first news, the second news, and the benchmark news are all very different, but it is difficult to explain the first news, What is the correlation between the second news. If there is a high correlation between the news with the same keywords, this embodiment makes the distance between the first news and the benchmark news smaller, indicating that the first news has a higher relevance to the benchmark news, and the second news.
  • the distance from the benchmark news is equivalent to the distance from the first news, that is, the correlation between the first news and the second news is weak, and the distance between the second news and the benchmark news is equivalent to the distance from the first news, that is, the first news. Strong correlation with the second news. For example, if the first news headline is "Star A Award", the selected benchmark news may be "Star A Full Record” and the keyword is Star A.
  • an apparatus for determining correlation between news items is provided in an embodiment of the present invention.
  • the device for determining the correlation between news items according to the foregoing embodiment further includes:
  • the first vector generating module 330 is adapted to acquire a feature attribute of the first news, and generate a vector corresponding to the first news according to the feature attribute of the first news;
  • the first benchmark news acquisition module 310 is further adapted to compare the vector corresponding to the first news with the vector corresponding to the benchmark news.
  • the second vector generating module 340 is adapted to acquire a feature attribute of the second news, and generate a vector corresponding to the second news according to the feature attribute of the second news;
  • the first benchmark news acquisition module 320 is further adapted to compare the vector corresponding to the second news with the vector corresponding to the benchmark news.
  • the feature attributes are not limited; by using one or more feature attributes of the news, it is easy to quantify the news as a number, and the distance between the news can be calculated more easily and accurately.
  • the first vector generating module 330 is further adapted to determine the correlation between news according to the foregoing embodiment.
  • the first news is segmented to obtain a plurality of words, and the word frequency of the plurality of words of the first news is calculated as a feature attribute of the first news.
  • the second vector generating module 340 is further adapted to perform segmentation on the second news to obtain a plurality of words, and calculate a word frequency of the plurality of words of the second news as a feature attribute of the second news.
  • a random number may be assigned to the first news, and an article vector is constructed for the first news according to the random number and the calculated word frequency; similarly, the second news and the benchmark news may also construct corresponding articles.
  • the minimum hash value of the article vector of the first news, the second news, and the benchmark news is calculated, and the specific distance value between the first news and the benchmark news, the second news and The distance between the benchmark news.
  • the distance between News A, News B and benchmark news is 4 and 3 respectively.
  • a small difference of 1 means that News A and News B are highly correlated, and News C and Benchmark News have a difference of 1. The difference between the distances from News A and News B is large, and the correlation is low.
  • An embodiment of the present invention further provides an apparatus for determining correlation between news.
  • the apparatus for determining correlation between news according to the embodiment, the correlation calculation module 310 is further adapted to When the distance difference is in the preset interval, the second news is set as the related news of the first news for pushing the second news when the related news of the first news needs to be pushed.
  • the preset interval is 0-1
  • the news A is the news related to the news B
  • the user can push the news B for the user when viewing the news A.
  • an embodiment of the present invention provides a method for calculating correlation between multiple news, including:
  • Step 410 Determine a first-level association relationship between the first news and the second news.
  • the first-level association relationship between the first news and the second news reflects the degree of similarity between the first news and the literal content of the second news.
  • Step 420 Determine a first-level association relationship between the second news and the third news.
  • Second news and third news The first-level relationship reflects the degree of similarity between the first news and the second news.
  • Step 430 Establish a secondary association relationship between the first news and the third news according to the first-level association relationship between the first news and the second news, and the first-level association relationship between the second news and the third news.
  • the second-level association relationship between the calculated first news and the third news is no longer limited by the literal content, because the second news is similar to the first news, and the second news and The third news literally similar parts may be different, which actually reflects the semantic similarity between the first news and the third news, so according to the present invention establishing the second-level association relationship, it is possible to find news with different literals but similar semantics.
  • An embodiment of the present invention further provides a method for calculating a correlation between multiple news items.
  • the step 410 of the multi-news correlation calculation method of the present embodiment includes:
  • the first vector is set according to the feature attribute of the first news
  • the second vector is set according to the feature attribute of the second news
  • the cosine similarity of the first vector and the second vector is calculated as a first-level association relationship between the first news and the second news.
  • the feature attributes are not limited, and feature attributes such as the word frequency of the word segmentation and the time of the release may be used to generate the vector.
  • the obtained first association relationship can reflect the degree of similarity between the first news and the second news. For example, for news A and news B, the first-level relationship between news A and news B is:
  • i is the dimension of the vector and the similarity is the pre-similarity.
  • the step 420 in the multi-news correlation calculation method of the embodiment specifically includes: setting a third vector according to the feature attribute of the third news, and calculating a cosine similarity of the second vector and the third vector as the second news and the third The first level of association of news. Similarly, the first-order relationship between the second news and the third news can also be calculated using the cosine similarity.
  • an embodiment of the present invention further provides a multi-news correlation calculation method, and steps in the multi-news correlation calculation method of the present embodiment compared to the foregoing embodiments.
  • specific package include:
  • Step 510 Calculate information entropy of the first news and the third news according to the first-level association relationship between the first news and the second news, and the first-level association relationship between the second news and the third news.
  • information entropy is used to determine the second-level association relationship between the first news and the third news, and the obtained second association relationship can reflect the semantic similarity between the first news and the third news, without being literally Similar effects.
  • news A and news B are first-level associations
  • news B and news C are first-level associations.
  • the information entropy of news A and news C is:
  • p(A, Bi, C) is the path probability from news A to news C through news Bi, and i means news B includes multiple news.
  • W1 cos(A,B1) ⁇ 2,
  • ... 1. Based on the above formula, information entropy can be calculated.
  • the step 430 in the multi-news correlation calculation method of the embodiment further includes: step 520, determining the second-level association relationship between the first news and the third news according to the information entropy of the first news and the third news.
  • a multi-news correlation calculation method is further provided.
  • the multi-news correlation calculation method of the embodiment when the second news is multiple, before step 420, the method further includes: searching for a corresponding third news for each second news according to a preset rule.
  • the step 130 of the multi-news correlation calculation method in this embodiment specifically includes:
  • a multi-news correlation calculation method is further provided.
  • the method further includes: sorting the plurality of third news according to the second-level association relationship between the plurality of third news and the first news.
  • the calculated second association relationship can actually reflect the semantic similarity between the news, so according to the second association relationship, more similar third news can be provided to the user.
  • an embodiment of the present invention provides a multi-news correlation calculation device, including:
  • the first association determining module 710 is adapted to determine a first-level association relationship between the first news and the second news.
  • the first-level association relationship between the first news and the second news reflects the degree of similarity between the first news and the literal content of the second news.
  • the second association determining module 720 is adapted to determine a first-level association relationship between the second news and the third news.
  • the first-level relationship between the second news and the third news reflects the degree of similarity between the first news and the second news.
  • the third association determining module 730 is configured to establish a level between the first news and the third news according to the first-level association relationship between the first news and the second news, and the first-level association relationship between the second news and the third news. connection relation.
  • the second-level association relationship between the calculated first news and the third news is no longer limited by the literal content, because the second news is similar to the first news, and the second news and The third news literally similar parts may be different, which actually reflects the semantic similarity between the first news and the third news, so according to the present invention establishing the second-level association relationship, it is possible to find news with different literals but similar semantics.
  • a multi-news correlation calculation device is further provided.
  • the first association determination module 710 further Suitable for setting a first vector according to a feature attribute of the first news, setting a second vector according to a feature attribute of the second news, and calculating a cosine similarity of the first vector and the second vector as a first-level association between the first news and the second news relationship.
  • the feature attributes are not limited, and feature attributes such as the word frequency of the word segmentation and the time of the release may be used to generate the vector.
  • the cosine similarity as the first association relationship, the obtained first association relationship can reflect the degree of similarity between the first news and the second news. For example, for news A and news B, the first-level relationship between news A and news B is:
  • i is the dimension of the vector and the similarity is the pre-similarity.
  • the second association determining module 720 is further adapted to set a third vector according to the feature attribute of the third news, and calculate a cosine similarity of the second vector and the third vector as a first-level association relationship between the second news and the third news. Same In the sample, the first-order relationship between the second news and the third news can also be calculated by using the cosine similarity.
  • a multi-news correlation calculation device is further provided.
  • the third association determination module 730 is compared with the foregoing embodiment. It is further adapted to calculate an information entropy of the first news and the third news according to the first-level association relationship between the first news and the second news, the first-level association relationship between the second news and the third news; and according to the first news and the third news The information entropy of the news determines the secondary relationship between the first news and the third news.
  • information entropy is used to determine the second-level association relationship between the first news and the third news, and the obtained second association relationship can reflect the semantic similarity between the first news and the third news, without being literally Similar effects.
  • news A and news B are first-level associations
  • news B and news C are first-level associations.
  • the information entropy of news A and news C is:
  • p(A, Bi, C) is the path probability from news A to news C through news Bi, and i means news B includes multiple news.
  • W1 cos(A,B1) ⁇ 2,
  • ... 1. Based on the above formula, information entropy can be calculated.
  • an embodiment of the present invention further provides a multi-news correlation calculation device.
  • the multi-news correlation calculation device of the present embodiment is second.
  • the device also includes:
  • the searching module 810 is adapted to search for a corresponding third news for each second news according to a preset rule.
  • the third association determination 730 module is further adapted to calculate a weight value of each third news according to a proportion of the second news corresponding to each third news in all the second news; and a weight according to each third news
  • the value, the information entropy corresponding to each third news calculates the second-level association relationship between each third news and the first news.
  • an embodiment of the present invention further provides a multi-news correlation calculation device.
  • the multi-news correlation calculation device of the present embodiment is third.
  • the device also includes:
  • a sorting module 820 configured to perform multiple levels according to a second level relationship between the plurality of third news and the first news Three news is sorted.
  • the calculated second association relationship can actually reflect the semantic similarity between the news, so according to the second association relationship, more similar third news can be provided to the user.
  • modules in the devices of the embodiments can be adaptively changed and placed in one or more devices different from the embodiment.
  • the modules or units or components of the embodiments may be combined into one module or unit or component, and further they may be divided into a plurality of sub-modules or sub-units or sub-components.
  • any combination of the features disclosed in the specification, including the accompanying claims, the abstract and the drawings, and any methods so disclosed, or All processes or units of the device are combined.
  • Each feature disclosed in this specification (including the accompanying claims, the abstract and the drawings) may be replaced by alternative features that provide the same, equivalent or similar purpose.
  • the various component embodiments of the present invention may be implemented in hardware, or in a software module running on one or more processors, or in a combination thereof.
  • a microprocessor or digital signal processor can be used in practice to implement correlation determination between devices and/or multiple news that determine correlation between news in accordance with embodiments of the present invention.
  • the invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for performing some or all of the methods described herein.
  • a program implementing the invention may be stored on a computer readable medium or may be in the form of one or more signals. This Such signals can be downloaded from the Internet website, or provided on a carrier signal, or provided in any other form.
  • FIG. 9 illustrates a computing device that can implement a method of determining correlation between news and/or a method of calculating correlation between multiple news in accordance with the present invention.
  • the computing device conventionally includes a processor 910 and a computer program product or computer readable medium in the form of a memory 920.
  • the memory 920 may be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), an EPROM, a hard disk, or a ROM.
  • Memory 920 has a memory space 930 for program code 931 for performing any of the method steps described above.
  • storage space 930 for program code may include various program code 931 for implementing various steps in the above methods, respectively.
  • the program code can be read from or written to one or more computer program products.
  • These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks.
  • Such a computer program product is typically a portable or fixed storage unit as described with reference to FIG.
  • the storage unit may have storage segments, storage spaces, and the like that are similarly arranged to memory 920 in the computing device of FIG.
  • the program code can be compressed, for example, in an appropriate form.
  • the storage unit includes computer readable code 931', ie, code that can be read by a processor, such as 910, that when executed by a computing device causes the computing device to perform each of the methods described above step.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Databases & Information Systems (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Multimedia (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Computer Graphics (AREA)
  • Geometry (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本发明提供一种确定新闻之间相关性、多新闻之间相关性计算方法和装置。确定新闻之间相关性的方法包括:将第一新闻与标杆新闻进行比较,得到第一新闻与标杆新闻的距离;将第二新闻与标杆新闻进行比较,得到第二新闻与标杆新闻的距离;计算第一新闻与标杆新闻的距离和第一新闻与标杆新闻的距离之间的距离差,根据距离差确定第一新闻与第二新闻之间的相关性。多新闻之间相关性计算方法包括:确定第一新闻与第二新闻的一级关联关系;确定第二新闻与第三新闻的一级关联关系;根据第一新闻与第二新闻的一级关联关系、第二新闻与第三新闻的一级关联关系,建立第一新闻与第三新闻之间的二级关联关系。

Description

确定新闻之间相关性、多新闻之间相关性计算方法和装置 技术领域
本发明涉及计算机技术领域,具体而言,涉及一种确定新闻之间相关性的方法和装置,以及多新闻之间相关性计算方法和装置。
背景技术
在互联网领域内,当新的新闻出现时,需要将其和已有的新闻进行比较,确定新的新闻和已有的哪些新闻是相关新闻关系,以便于在用户查看新闻时将相关新闻一起推荐给用户。
由于已有新闻的数量庞大,而每个新的新闻都需要与所有已有新闻进行比较,导致计算量非常巨大,计算新闻相关性的效率非常低。
发明内容
鉴于上述问题,提出了本发明以便提供一种克服上述问题或者至少部分地解决上述问题的确定新闻之间相关性的方法和装置,以及多新闻之间相关性计算方法和装置。
依据本发明的一个方面,提供了一种确定新闻之间相关性的方法,包括:将第一新闻与标杆新闻进行比较,得到所述第一新闻与所述标杆新闻的距离;将第二新闻与所述标杆新闻进行比较,得到所述第二新闻与所述标杆新闻的距离;计算所述第一新闻与所述标杆新闻的距离和所述第一新闻与所述标杆新闻的距离之间的距离差,根据所述距离差确定所述第一新闻与所述第二新闻之间的相关性。
依据本发明的另一个方面,提供了一种确定新闻之间相关性的装置,包括:第一距离计算模块,用于将第一新闻与标杆新闻进行比较,得到所述第一新闻与所述标杆新闻的距离;第二距离计算模块,用于将第二新闻与所述标杆新闻进行比较,得到所述第二新闻与所述标杆新闻的距离;相关性计算模块,用于计算所述第一新闻和所述标杆新闻的距离与所述第一新闻与所述标杆新闻的距离之间的距离差,根据所述距离差确定所述第一新闻与所述第二新闻之间的相关性。
根据以上技术方案,本发明的确定新闻之间相关性的方法和装置至少具有以下优点:
根据本发明的技术方案,当需要分析不同新闻之间的相关性时,不必进行多个 新闻之间的对比,而是进行多个新闻与标杆新闻之间的比较,如果两个新闻与标杆之间的距离相似,则说明两个新闻之间具有一定的类似程度;由于标杆新闻只有一个,而其他新闻不需要进行互相之间的对比,只需要进行与标杆新闻的对比,即可确定多个新闻之间的相关性,所以根据本发明的技术方案获取相关新闻的效率非常高。
依据本发明的再一个方面,提供了一种多新闻之间相关性计算方法,包括:确定第一新闻与第二新闻的一级关联关系;确定所述第二新闻与第三新闻的一级关联关系;根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,建立所述第一新闻与所述第三新闻之间的二级关联关系。
依据本发明的又一个方面,提供了一种多新闻之间相关性计算装置,包括:第一关联关系确定模块,用于确定第一新闻与第二新闻的一级关联关系;第二关联关系确定模块,确定所述第二新闻与第三新闻的一级关联关系;第三关联关系确定模块,根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,建立所述第一新闻与所述第三新闻之间的二级关联关系。
根据以上技术方案,本发明的多新闻之间相关性计算方法和装置至少具有以下优点:
在本发明的技术方案中,第一新闻与第二新闻之间的一级关联关系反映了第一新闻与第二新闻的字面内容之间的相似程度,第二新闻与第三新闻之间的一级关联关系反映了第一新闻与第二新闻的字面内容之间的相似程度,但计算得到的第一新闻与第三新闻之间的二级关联关系不再受到字面内容的限制,因为第二新闻与第一新闻字面相似的部分、和第二新闻与第三新闻字面相似的部分可以不同,其实际上反映的是第一新闻与第三新闻的语义相似程度,所以根据本发明的建立二级关联关系,可以查找到字面不同但语义相似的新闻。
根据本发明的又一个方面,提供了一种计算机程序,其包括计算机可读代码,当所述计算机可读代码在计算设备上运行时,导致所述计算设备执行根据上述任一个的确定新闻之间相关性的方法,和/或,多新闻之间相关性计算方法。
根据本发明的再一个方面,提供了一种计算机可读介质,其中存储了如上所述的计算机程序。
上述说明仅是本发明技术方案的概述,为了能够更清楚了解本发明的技术手段,而可依照说明书的内容予以实施,并且为了让本发明的上述和其它目的、特征和优点能够更明显易懂,以下特举本发明的具体实施方式。
附图说明
通过阅读下文优选实施方式的详细描述,各种其他的优点和益处对于本领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的,而并不认为是对本发明的限制。而且在整个附图中,用相同的参考符号表示相同的部件。在附图中:
图1示出了根据本发明的一个实施例的确定新闻之间相关性的方法的流程图;
图2示出了根据本发明的一个实施例的确定新闻之间相关性的装置的框图;
图3示出了根据本发明的一个实施例的确定新闻之间相关性的装置的框图;
图4示出了根据本发明的一个实施例的多新闻之间相关性计算方法的流程图;
图5示出了根据本发明的一个实施例的多新闻之间相关性计算方法的流程图;
图6示出了根据本发明的一个实施例的多新闻之间相关性计算方法的原理图;
图7示出了根据本发明的一个实施例的多新闻之间相关性计算装置的框图;
图8示出了根据本发明的一个实施例的多新闻之间相关性计算装置的框图;
图9示意性地示出了用于执行根据本发明的确定新闻之间相关性的方法,和/或,多新闻之间相关性计算方法的计算设备的框图;以及
图10示意性地示出了用于保持或者携带实现根据本发明的确定新闻之间相关性的方法,和/或,多新闻之间相关性计算方法的程序代码的存储单元。
具体实施方式
下面结合附图和具体的实施方式对本发明作进一步的描述。
如图1所示,本发明的一个实施例中提供了一种确定新闻之间相关性的方法,包括:
步骤110,将第一新闻与标杆新闻进行比较,得到第一新闻与标杆新闻的距离。在本实施例中,对标杆新闻不进行限制,任何一篇新闻都可以选定作为标杆新闻。
步骤120,将第二新闻与标杆新闻进行比较,得到第二新闻与标杆新闻的距离。
步骤130,计算第一新闻与标杆新闻的距离和第一新闻与标杆新闻的距离之间的距离差,根据距离差确定第一新闻与第二新闻之间的相关性。
在本实施例中,距离反映了新闻之间的差别,本实施例对计算距离的方式不进行限制;根据本实施例的技术方案,容易理解,当第一新闻与标杆新闻之间的距离,和第二新闻与标杆新闻之间的距离相似时,说明第一新闻与第二新闻之间的距离较近,也就是两个新闻之间的相关性较强;当第一新闻与标杆新闻之间距离较近,第二新闻与标杆新闻之间的距离较远时,说明第一新闻与第二新闻之间的距离较远, 两个新闻之间的相关性较弱。
本发明的一个实施例中还提供一种确定新闻之间相关性的方法,相比于前述的实施例,本实施例的确定新闻之间相关性的方法,在步骤110之前,还包括:识别第一新闻的类型,并从标杆新闻集合中选择具有相对应类型的标杆新闻。
在本实施例中,如果第一新闻、第二新闻与标杆新闻之间的距离过大,只能说明第一新闻、第二新闻与标杆新闻均有很大不同,但难以说明第一新闻、第二新闻之间的相关性如何。而同类型的新闻之间具有较高的相关性,则本实施例使得第一新闻与标杆新闻之间的距离较小,说明第一新闻与标杆新闻相关性较高,则第二新闻与标杆新闻距离大则相当于与第一新闻距离大,即第一新闻与第二新闻相关性较弱,第二新闻与标杆新闻距离小则相当于与第一新闻距离小,即第一新闻与第二新闻相关性较强。例如,如果第一新闻为体育新闻,则选取的标杆新闻为体育新闻。
本发明的一个实施例中还提供一种确定新闻之间相关性的方法,相比于前述的实施例,本实施例的确定新闻之间相关性的方法,在步骤110之前,还包括:获取第一新闻中的关键词,并从标杆新闻集合中选择具有关键词的标杆新闻。
在本实施例中,如果第一新闻、第二新闻与标杆新闻之间的距离过大,只能说明第一新闻、第二新闻与标杆新闻均有很大不同,但难以说明第一新闻、第二新闻之间的相关性如何。而具有相同关键词的新闻之间具有较高的相关性,则本实施例使得第一新闻与标杆新闻之间的距离较小,说明第一新闻与标杆新闻相关性较高,则第二新闻与标杆新闻距离大则相当于与第一新闻距离大,即第一新闻与第二新闻相关性较弱,第二新闻与标杆新闻距离小则相当于与第一新闻距离小,即第一新闻与第二新闻相关性较强。例如,如果第一新闻标题为《明星A获奖》,则选取的标杆新闻可以是《明星A全纪录》,关键词是明星A。
本发明的一个实施例中还提供一种确定新闻之间相关性的方法,相比于前述的实施例,本实施例的确定新闻之间相关性的方法中的步骤110,具体包括:获取第一新闻的特征属性,并根据述第一新闻的特征属性生成第一新闻对应的向量,将第一新闻对应的向量与标杆新闻对应的向量进行比较。本实施例的确定新闻之间相关性的方法中的步骤120具体包括:获取第二新闻的特征属性,并根据述第二新闻的特征属性生成第二新闻对应的向量,并将第二新闻对应的向量与标杆新闻对应的向量进行比较。
在本实施例中,对特征属性不进行限制,利用新闻的一个或多个特征属性,容易将新闻量化称谓数字,能够更容易、更精确地计算新闻之间的距离。
本发明的一个实施例中还提供一种确定新闻之间相关性的方法,相比于前述的 实施例,本实施例的确定新闻之间相关性的方法中的步骤110,具体包括:对第一新闻进行分词得到多个词,计算第一新闻的多个词的词频,作为第一新闻的特征属性。本实施例的确定新闻之间相关性的方法中的步骤120,具体包括:对第二新闻进行分词得到多个词,计算第二新闻的多个词的词频,作为第二新闻的特征属性。
在本实施例中,可以为第一新闻分配一个随机数,根据随机数和计算得到的词频,为第一新闻构造出一个文章向量;同样地,第二新闻、标杆新闻也可以构造相应的文章向量;在本实施例中,对第一新闻、第二新闻、标杆新闻的文章向量取最小哈希值再进行计算,能够得到第一新闻与标杆新闻之间的具体距离值、第二新闻与标杆新闻之间的距离值。例如,新闻A、新闻B与标杆新闻之间的距离分别为4、3,差值较小为1表示新闻A、新闻B相关性较高,而新闻C与标杆新闻的差值为1,其与新闻A、新闻B的距离差值较大,则相关性较低。
本发明的一个实施例中还提供一种确定新闻之间相关性的方法,相比于前述的实施例,本实施例的确定新闻之间相关性的方法中的步骤130,具体包括:在距离差位于预设区间时,将第二新闻设置为第一新闻的相关新闻,以用于在需推送第一新闻的相关新闻时推送第二新闻。例如,结合前述的实施例内容,假设预设区间为0-1,则新闻A为新闻B的相关新闻,用户查看新闻A时可以为用户推送新闻B。
如图2所示,本发明的一个实施例中提供了一种确定新闻之间相关性的装置,包括:
第一距离计算模块210,适于将第一新闻与标杆新闻进行比较,得到第一新闻与标杆新闻的距离。在本实施例中,对标杆新闻不进行限制,任何一篇新闻都可以选定作为标杆新闻。
第二距离计算模块220,适于将第二新闻与标杆新闻进行比较,得到第二新闻与标杆新闻的距离。
相关性计算模块230,适于计算第一新闻和标杆新闻的距离与第一新闻与标杆新闻的距离之间的距离差,根据距离差确定第一新闻与第二新闻之间的相关性。
在本实施例中,距离反映了新闻之间的差别,本实施例对计算距离的方式不进行限制;根据本实施例的技术方案,容易理解,当第一新闻与标杆新闻之间的距离,和第二新闻与标杆新闻之间的距离相似时,说明第一新闻与第二新闻之间的距离较近,也就是两个新闻之间的相关性较强;当第一新闻与标杆新闻之间距离较近,第二新闻与标杆新闻之间的距离较远时,说明第一新闻与第二新闻之间的距离较远,两个新闻之间的相关性较弱。
如图3所示,本发明的一个实施例中还提供一种确定新闻之间相关性的装置, 相比于前述的实施例,本实施例的确定新闻之间相关性的装置,还包括:
第一标杆新闻获取模块310,适于识别第一新闻的类型,并从标杆新闻集合中选择具有相对应类型的标杆新闻。
在本实施例中,如果第一新闻、第二新闻与标杆新闻之间的距离过大,只能说明第一新闻、第二新闻与标杆新闻均有很大不同,但难以说明第一新闻、第二新闻之间的相关性如何。而同类型的新闻之间具有较高的相关性,则本实施例使得第一新闻与标杆新闻之间的距离较小,说明第一新闻与标杆新闻相关性较高,则第二新闻与标杆新闻距离大则相当于与第一新闻距离大,即第一新闻与第二新闻相关性较弱,第二新闻与标杆新闻距离小则相当于与第一新闻距离小,即第一新闻与第二新闻相关性较强。例如,如果第一新闻为体育新闻,则选取的标杆新闻为体育新闻。
如图3所示,本发明的一个实施例中还提供一种确定新闻之间相关性的装置,相比于前述的实施例,本实施例的确定新闻之间相关性的装置,还包括:
第二标杆新闻获取模块320,适于获取第一新闻中的关键词,并从标杆新闻集合中选择具有关键词的标杆新闻。
在本实施例中,如果第一新闻、第二新闻与标杆新闻之间的距离过大,只能说明第一新闻、第二新闻与标杆新闻均有很大不同,但难以说明第一新闻、第二新闻之间的相关性如何。而具有相同关键词的新闻之间具有较高的相关性,则本实施例使得第一新闻与标杆新闻之间的距离较小,说明第一新闻与标杆新闻相关性较高,则第二新闻与标杆新闻距离大则相当于与第一新闻距离大,即第一新闻与第二新闻相关性较弱,第二新闻与标杆新闻距离小则相当于与第一新闻距离小,即第一新闻与第二新闻相关性较强。例如,如果第一新闻标题为《明星A获奖》,则选取的标杆新闻可以是《明星A全纪录》,关键词是明星A。
如图3所示,本发明的一个实施例中还提供一种确定新闻之间相关性的装置,相比于前述的实施例,本实施例的确定新闻之间相关性的装置,还包括:
第一向量生成模块330,适于获取第一新闻的特征属性,并根据第一新闻的特征属性生成第一新闻对应的向量;
第一标杆新闻获取模块310还适于将第一新闻对应的向量与标杆新闻对应的向量进行比较。
第二向量生成模块340,适于获取第二新闻的特征属性,并根据述第二新闻的特征属性生成第二新闻对应的向量;
第一标杆新闻获取模块320还适于将第二新闻对应的向量与标杆新闻对应的向量进行比较。
在本实施例中,对特征属性不进行限制;利用新闻的一个或多个特征属性,容易将新闻量化称谓数字,能够更容易、更精确地计算新闻之间的距离。
本发明的一个实施例中还提供一种确定新闻之间相关性的装置,相比于前述的实施例,本实施例的确定新闻之间相关性的装置,第一向量生成模块330还适于对第一新闻进行分词得到多个词,计算第一新闻的多个词的词频,作为第一新闻的特征属性。
第二向量生成模块340还适于对第二新闻进行分词得到多个词,计算第二新闻的多个词的词频,作为第二新闻的特征属性。
在本实施例中,可以为第一新闻分配一个随机数,根据随机数和计算得到的词频,为第一新闻构造出一个文章向量;同样地,第二新闻、标杆新闻也可以构造相应的文章向量;在本实施例中,对第一新闻、第二新闻、标杆新闻的文章向量取最小哈希值再进行计算,能够得到第一新闻与标杆新闻之间的具体距离值、第二新闻与标杆新闻之间的距离值。例如,新闻A、新闻B与标杆新闻之间的距离分别为4、3,差值较小为1表示新闻A、新闻B相关性较高,而新闻C与标杆新闻的差值为1,其与新闻A、新闻B的距离差值较大,则相关性较低。
本发明的一个实施例中还提供一种确定新闻之间相关性的装置,相比于前述的实施例,本实施例的确定新闻之间相关性的装置,相关性计算模块310还适于在距离差位于预设区间时,将第二新闻设置为第一新闻的相关新闻,以用于在需推送第一新闻的相关新闻时推送第二新闻。例如,结合前述的实施例内容,假设预设区间为0-1,则新闻A为新闻B的相关新闻,用户查看新闻A时可以为用户推送新闻B。
进一步,在互联网领域,当为一个新闻出现时,通常需要为该新闻查找相关的新闻,以一并推送给用户。
目前的现有技术中,通常是按照一个新闻中的字面内容来查找相关新闻。这种查找相关新闻的方式中存在较大的缺陷:因为相关的两个新闻之间可能语义相同而字面内容不同,则根据现有技术无法找到语义相同而字面内容不同的新闻作为相关新闻。
为解决这一技术问题,如图4所示,本发明的一个实施例中提供了一种多新闻之间相关性计算方法,包括:
步骤410,确定第一新闻与第二新闻的一级关联关系。在本实施例中,第一新闻与第二新闻之间的一级关联关系反映了第一新闻与第二新闻的字面内容之间的相似程度。
步骤420,确定第二新闻与第三新闻的一级关联关系。第二新闻与第三新闻之 间的一级关联关系反映了第一新闻与第二新闻的字面内容之间的相似程度。
步骤430,根据第一新闻与第二新闻的一级关联关系、第二新闻与第三新闻的一级关联关系,建立第一新闻与第三新闻之间的二级关联关系。
根据本实施的技术方案,计算得到的第一新闻与第三新闻之间的二级关联关系不再受到字面内容的限制,因为第二新闻与第一新闻字面相似的部分、和第二新闻与第三新闻字面相似的部分可以不同,其实际上反映的是第一新闻与第三新闻的语义相似程度,所以根据本发明的建立二级关联关系,可以查找到字面不同但语义相似的新闻。
本发明的一个实施例中还提供了一种多新闻之间相关性计算方法,相比于前述的实施例,本实施例的多新闻之间相关性计算方法中的步骤410,具体包括:
根据第一新闻的特征属性设置第一向量,根据第二新闻的特征属性设置第二向量,计算第一向量和第二向量的余弦相似度作为第一新闻与第二新闻的一级关联关系。
在本实施例中,对特征属性不进行限制,例如其分词的词频、发布的时间等特征属性均可用于生成向量。利用余弦相似度作为第一关联关系,则得到的第一关联关系能够反映第一新闻与第二新闻的相似程度。例如对于新闻A、新闻B,则新闻A与新闻B的一级关联关系为:
Figure PCTCN2016103397-appb-000001
其中,i为向量的维度,similarity即为预先相似度。
本实施例的多新闻之间相关性计算方法中的步骤420具体包括:根据第三新闻的特征属性设置第三向量,计算第二向量和第三向量的余弦相似度作为第二新闻与第三新闻的一级关联关系。同样地,利用余弦相似度也可以计算出第二新闻与第三新闻的一级关联关系。
如图5所示,本发明的一个实施例中还提供了一种多新闻之间相关性计算方法,相比于前述的实施例,本实施例的多新闻之间相关性计算方法中的步骤430,具体包 括:
步骤510,根据第一新闻与第二新闻的一级关联关系、第二新闻与第三新闻的一级关联关系,计算第一新闻与第三新闻的信息熵。在本实施例中,采用信息熵来确定第一新闻与第三新闻的二级关联关系,则得到的第二关联关系能够体现第一新闻与第三新闻的语义相似度,而不受到字面是否相似的影响。例如,新闻A、新闻B之间为一级关联关系,新闻B、新闻C之间为一级关联关系,则新闻A、新闻C的信息熵为:
Figure PCTCN2016103397-appb-000002
这里的p(A,Bi,C)为从新闻A通过新闻Bi到达新闻C的路径概率,i表示新闻B包括多个新闻。
如图6所示,计算通过新闻A通过新闻B1到达新闻C1的概率为:P(A,B1,C1)=|W1|*|V11|,这里||为向量余弦值的平方标准化后的权重。
W1=cos(A,B1)^2,|W1|+|W2|+|W3|+…=1;V11=cos(B1,C1)^2,|V11|+|V12|+|V13|…=1。基于以上公式,可以计算得到信息熵。
本实施例的多新闻之间相关性计算方法中的步骤430还包括:步骤520,根据第一新闻与第三新闻的信息熵,确定第一新闻与第三新闻的二级关联关系。
本发明的一个实施例中还提供了一种多新闻之间相关性计算方法,相比于前述的实施例,本实施例的多新闻之间相关性计算方法,第二新闻为多个时,在步骤420之前,还包括:根据预设规则,为每个第二新闻查找相应的第三新闻。
本实施例的多新闻之间相关性计算方法的步骤130具体包括:
根据每个第三新闻对应的第二新闻在全部第二新闻中的占比,计算每个第三新闻的权重值;根据每个第三新闻的权重值、每个第三新闻对应的信息熵,计算每个第三新闻与第一新闻的二级关联关系。在本实施例中,假设第二新闻共有N条,某第三新闻关联到其中的nj个第二新闻,则该第三新闻的权重值IDF=logN/nj,该权值能够反映出第三新闻的重要性程度。
本发明的一个实施例中还提供了一种多新闻之间相关性计算方法,相比于前述的实施例,本实施例的多新闻之间相关性计算方法,第三新闻为多个时,方法还包括:根据多个第三新闻与第一新闻的二级关联关系,对多个第三新闻进行排序。
在本实施例的技术方案中,计算出的第二关联关系实际上能够反映新闻之间的语义相似度高低,所以依照第二关联关系排序,能够把更相似的第三新闻提供给用户。
如图7所示,本发明的一个实施例中提供了一种多新闻之间相关性计算装置,包括:
第一关联关系确定模块710,适于确定第一新闻与第二新闻的一级关联关系。在本实施例中,第一新闻与第二新闻之间的一级关联关系反映了第一新闻与第二新闻的字面内容之间的相似程度。
第二关联关系确定模块720,适于确定第二新闻与第三新闻的一级关联关系。第二新闻与第三新闻之间的一级关联关系反映了第一新闻与第二新闻的字面内容之间的相似程度。
第三关联关系确定模块730,适于根据第一新闻与第二新闻的一级关联关系、第二新闻与第三新闻的一级关联关系,建立第一新闻与第三新闻之间的二级关联关系。
根据本实施的技术方案,计算得到的第一新闻与第三新闻之间的二级关联关系不再受到字面内容的限制,因为第二新闻与第一新闻字面相似的部分、和第二新闻与第三新闻字面相似的部分可以不同,其实际上反映的是第一新闻与第三新闻的语义相似程度,所以根据本发明的建立二级关联关系,可以查找到字面不同但语义相似的新闻。
本发明的一个实施例中还提供了一种多新闻之间相关性计算装置,相比于前述的实施例,本实施例的多新闻之间相关性计算装置,第一关联关系确定模块710还适于根据第一新闻的特征属性设置第一向量,根据第二新闻的特征属性设置第二向量,计算第一向量和第二向量的余弦相似度作为第一新闻与第二新闻的一级关联关系。在本实施例中,对特征属性不进行限制,例如其分词的词频、发布的时间等特征属性均可用于生成向量。利用余弦相似度作为第一关联关系,则得到的第一关联关系能够反映第一新闻与第二新闻的相似程度。例如对于新闻A、新闻B,则新闻A与新闻B的一级关联关系为:
Figure PCTCN2016103397-appb-000003
其中,i为向量的维度,similarity即为预先相似度。
第二关联关系确定模块720还适于根据第三新闻的特征属性设置第三向量,计算第二向量和第三向量的余弦相似度作为第二新闻与第三新闻的一级关联关系。同 样地,利用余弦相似度也可以计算出第二新闻与第三新闻的一级关联关系。
本发明的一个实施例中还提供了一种多新闻之间相关性计算装置,相比于前述的实施例,本实施例的多新闻之间相关性计算装置中,第三关联关系确定模块730还适于根据第一新闻与第二新闻的一级关联关系、第二新闻与第三新闻的一级关联关系,计算第一新闻与第三新闻的信息熵;以及根据第一新闻与第三新闻的信息熵,确定第一新闻与第三新闻的二级关联关系。
在本实施例中,采用信息熵来确定第一新闻与第三新闻的二级关联关系,则得到的第二关联关系能够体现第一新闻与第三新闻的语义相似度,而不受到字面是否相似的影响。例如,新闻A、新闻B之间为一级关联关系,新闻B、新闻C之间为一级关联关系,则新闻A、新闻C的信息熵为:
Figure PCTCN2016103397-appb-000004
这里的p(A,Bi,C)为从新闻A通过新闻Bi到达新闻C的路径概率,i表示新闻B包括多个新闻。
如图6所示,计算通过新闻A通过新闻B1到达新闻C1的概率为:P(A,B1,C1)=|W1|*|V11|,这里||为向量余弦值的平方标准化后的权重。
W1=cos(A,B1)^2,|W1|+|W2|+|W3|+…=1;V11=cos(B1,C1)^2,|V11|+|V12|+|V13|…=1。基于以上公式,可以计算得到信息熵。
如图8所示,本发明的一个实施例中还提供了一种多新闻之间相关性计算装置,相比于前述的实施例,本实施例的多新闻之间相关性计算装置,第二新闻为多个时,装置还包括:
查找模块810,适于根据预设规则,为每个第二新闻查找相应的第三新闻。
第三关联关系确定730模块还适于根据每个第三新闻对应的第二新闻在全部第二新闻中的占比,计算每个第三新闻的权重值;以及根据每个第三新闻的权重值、每个第三新闻对应的信息熵,计算每个第三新闻与第一新闻的二级关联关系。在本实施例中,假设第二新闻共有N条,某第三新闻关联到其中的nj个第二新闻,则该第三新闻的权重值IDF=logN/nj,该权值能够反映出第三新闻的重要性程度。
如图8所示,本发明的一个实施例中还提供了一种多新闻之间相关性计算装置,相比于前述的实施例,本实施例的多新闻之间相关性计算装置,第三新闻为多个时,装置还包括:
排序模块820,用于根据多个第三新闻与第一新闻的二级关联关系,对多个第 三新闻进行排序。在本实施例的技术方案中,计算出的第二关联关系实际上能够反映新闻之间的语义相似度高低,所以依照第二关联关系排序,能够把更相似的第三新闻提供给用户。
在此处所提供的说明书中,说明了大量具体细节。然而,能够理解,本发明的实施例可以在没有这些具体细节的情况下实践。在一些实例中,并未详细示出公知的方法、结构和技术,以便不模糊对本说明书的理解。
类似地,应当理解,为了精简本公开并帮助理解各个发明方面中的一个或多个,在上面对本发明的示例性实施例的描述中,本发明的各个特征有时被一起分组到单个实施例、图、或者对其的描述中。然而,并不应将该公开的方法解释成反映如下意图:即所要求保护的本发明要求比在每个权利要求中所明确记载的特征更多的特征。更确切地说,如下面的权利要求书所反映的那样,发明方面在于少于前面公开的单个实施例的所有特征。因此,遵循具体实施方式的权利要求书由此明确地并入该具体实施方式,其中每个权利要求本身都作为本发明的单独实施例。
本领域那些技术人员可以理解,可以对实施例中的设备中的模块进行自适应性地改变并且把它们设置在与该实施例不同的一个或多个设备中。可以把实施例中的模块或单元或组件组合成一个模块或单元或组件,以及此外可以把它们分成多个子模块或子单元或子组件。除了这样的特征和/或过程或者单元中的至少一些是相互排斥之外,可以采用任何组合对本说明书(包括伴随的权利要求、摘要和附图)中公开的所有特征以及如此公开的任何方法或者设备的所有过程或单元进行组合。除非另外明确陈述,本说明书(包括伴随的权利要求、摘要和附图)中公开的每个特征可以由提供相同、等同或相似目的的替代特征来代替。
此外,本领域的技术人员能够理解,尽管在此所述的一些实施例包括其它实施例中所包括的某些特征而不是其它特征,但是不同实施例的特征的组合意味着处于本发明的范围之内并且形成不同的实施例。例如,在下面的权利要求书中,所要求保护的实施例的任意之一都可以以任意的组合方式来使用。
本发明的各个部件实施例可以以硬件实现,或者以在一个或者多个处理器上运行的软件模块实现,或者以它们的组合实现。本领域的技术人员应当理解,可以在实践中使用微处理器或者数字信号处理器(DSP)来实现根据本发明实施例的确定新闻之间相关性的装置和/或多新闻之间相关性计算装置中的一些或者全部部件的一些或者全部功能。本发明还可以实现为用于执行这里所描述的方法的一部分或者全部的设备或者装置程序(例如,计算机程序和计算机程序产品)。这样的实现本发明的程序可以存储在计算机可读介质上,或者可以具有一个或者多个信号的形式。这 样的信号可以从因特网网站上下载得到,或者在载体信号上提供,或者以任何其他形式提供。
例如,图9示出了可以实现根据本发明的确定新闻之间相关性的方法和/或多新闻之间相关性计算方法的计算设备。该计算设备传统上包括处理器910和以存储器920形式的计算机程序产品或者计算机可读介质。存储器920可以是诸如闪存、EEPROM(电可擦除可编程只读存储器)、EPROM、硬盘或者ROM之类的电子存储器。存储器920具有用于执行上述方法中的任何方法步骤的程序代码931的存储空间930。例如,用于程序代码的存储空间930可以包括分别用于实现上面的方法中的各种步骤的各个程序代码931。这些程序代码可以从一个或者多个计算机程序产品中读出或者写入到这一个或者多个计算机程序产品中。这些计算机程序产品包括诸如硬盘,紧致盘(CD)、存储卡或者软盘之类的程序代码载体。这样的计算机程序产品通常为如参考图10所述的便携式或者固定存储单元。该存储单元可以具有与图9的计算设备中的存储器920类似布置的存储段、存储空间等。程序代码可以例如以适当形式进行压缩。通常,存储单元包括计算机可读代码931’,即可以由例如诸如910之类的处理器读取的代码,这些代码当由计算设备运行时,导致该计算设备执行上面所描述的方法中的各个步骤。
本文中所称的“一个实施例”、“实施例”或者“一个或者多个实施例”意味着,结合实施例描述的特定特征、结构或者特性包括在本发明的至少一个实施例中。此外,请注意,这里“在一个实施例中”的词语例子不一定全指同一个实施例。
应该注意的是上述实施例对本发明进行说明而不是对本发明进行限制,并且本领域技术人员在不脱离所附权利要求的范围的情况下可设计出替换实施例。在权利要求中,不应将位于括号之间的任何参考符号构造成对权利要求的限制。单词“包含”不排除存在未列在权利要求中的元件或步骤。位于元件之前的单词“一”或“一个”不排除存在多个这样的元件。本发明可以借助于包括有若干不同元件的硬件以及借助于适当编程的计算机来实现。在列举了若干装置的单元权利要求中,这些装置中的若干个可以是通过同一个硬件项来具体体现。单词第一、第二、以及第三等的使用不表示任何顺序。可将这些单词解释为名称。
此外,还应当注意,本说明书中使用的语言主要是为了可读性和教导的目的而选择的,而不是为了解释或者限定本发明的主题而选择的。因此,在不偏离所附权利要求书的范围和精神的情况下,对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。对于本发明的范围,对本发明所做的公开是说明性的,而非限制性的,本发明的范围由所附权利要求书限定。

Claims (24)

  1. 一种确定新闻之间相关性的方法,包括:
    将第一新闻与标杆新闻进行比较,得到所述第一新闻与所述标杆新闻的距离;
    将第二新闻与所述标杆新闻进行比较,得到所述第二新闻与所述标杆新闻的距离;
    计算所述第一新闻与所述标杆新闻的距离和所述第一新闻与所述标杆新闻的距离之间的距离差,根据所述距离差确定所述第一新闻与所述第二新闻之间的相关性。
  2. 根据权利要求1所述的方法,其中,在将第一新闻与标杆新闻进行比较之前,还包括:
    识别所述第一新闻的类型,并从标杆新闻集合中选择具有相对应类型的所述标杆新闻。
  3. 根据权利要求1-2任一项所述的方法,其中,在将第一新闻与标杆新闻进行比较之前,还包括:
    获取所述第一新闻中的关键词,并从标杆新闻集合中选择具有所述关键词的所述标杆新闻。
  4. 根据权利要求1-3任一项所述的方法,其中,
    将第一新闻与标杆新闻进行比较,得到所述第一新闻与所述标杆新闻的距离,具体包括:
    获取所述第一新闻的特征属性,并根据述第一新闻的特征属性生成所述第一新闻对应的向量,将所述第一新闻对应的向量与所述标杆新闻对应的向量进行比较;
    将第二新闻与所述标杆新闻进行比较,得到所述第一新闻与所述标杆新闻的距离,具体包括:
    获取所述第二新闻的特征属性,并根据述第二新闻的特征属性生成所述第二新闻对应的向量,并将所述第二新闻对应的向量与所述标杆新闻对应的向量进行比较。
  5. 根据权利要求1-4任一项所述的方法,其中,
    获取所述第一新闻的特征属性,具体包括:
    对所述第一新闻进行分词得到多个词,计算所述第一新闻的多个词的词频,作为所述第一新闻的特征属性;
    获取所述第二新闻的特征属性,具体包括:
    对所述第二新闻进行分词得到多个词,计算所述第二新闻的多个词的词频,作为所述第二新闻的特征属性。
  6. 根据权利要求1-5中任一项所述的方法,其中,根据所述距离差确定所述第一新闻与所述第二新闻之间的相关性,具体包括:
    在所述距离差位于预设区间时,将所述第二新闻设置为所述第一新闻的相关新闻,以用于在需推送所述第一新闻的相关新闻时推送所述第二新闻。
  7. 一种确定新闻之间相关性的装置,包括:
    第一距离计算模块,适于将第一新闻与标杆新闻进行比较,得到所述第一新闻与所述标杆新闻的距离;
    第二距离计算模块,适于将第二新闻与所述标杆新闻进行比较,得到所述第二新闻与所述标杆新闻的距离;
    相关性计算模块,适于计算所述第一新闻和所述标杆新闻的距离与所述第一新闻与所述标杆新闻的距离之间的距离差,根据所述距离差确定所述第一新闻与所述第二新闻之间的相关性。
  8. 根据权利要求7所述的装置,其中,还包括:
    第一标杆新闻获取模块,适于识别所述第一新闻的类型,并从标杆新闻集合中选择具有相对应类型的所述标杆新闻。
  9. 根据权利要求7-8任一项所述的装置,其中,还包括:
    第二标杆新闻获取模块,适于获取所述第一新闻中的关键词,并从标杆新闻集合中选择具有所述关键词的所述标杆新闻。
  10. 根据权利要求7-9任一项所述的装置,其中,还包括:
    第一向量生成模块,适于获取所述第一新闻的特征属性,并根据述第一新闻的特征属性生成所述第一新闻对应的向量;所述第一标杆新闻获取模块将所述第一新闻对应的向量与所述标杆新闻对应的向量进行比较;
    第二向量生成模块,适于获取所述第二新闻的特征属性,并根据述第二新闻的特征属性生成所述第二新闻对应的向量;所述第一标杆新闻获取模块将所述第二新闻对应的向量与所述标杆新闻对应的向量进行比较。
  11. 根据权利要求10所述的装置,其中,所述第一向量生成模块对所述第一新闻进行分词得到多个词,计算所述第一新闻的多个词的词频,作为所述第一新闻的特征属性;所述第二向量生成模块对所述第二新闻进行分词得到多个词,计算所述第二新闻的多个词的词频,作为所述第二新闻的特征属性。
  12. 根据权利要求7-11任一项所述的装置,其中,所述相关性计算模块在所述距离差位于预设区间时,将所述第二新闻设置为所述第一新闻的相关新闻,以用于在需推送所述第一新闻的相关新闻时推送所述第二新闻。
  13. 一种多新闻之间相关性计算方法,包括:
    确定第一新闻与第二新闻的一级关联关系;
    确定所述第二新闻与第三新闻的一级关联关系;
    根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,建立所述第一新闻与所述第三新闻之间的二级关联关系。
  14. 根据权利要求13所述的方法,其中,
    确定第一新闻与第二新闻的一级关联关系,具体包括:
    根据所述第一新闻的特征属性设置第一向量,根据所述第二新闻的特征属性设置第二向量,计算所述第一向量和所述第二向量的余弦相似度作为所述第一新闻与所述第二新闻的一级关联关系;
    确定所述第二新闻与第三新闻的一级关联关系,具体包括:
    根据所述第三新闻的特征属性设置第三向量,计算所述第二向量和所述第三向量的余弦相似度作为所述第二新闻与所述第三新闻的一级关联关系。
  15. 根据权利要求13-14任一项所述的方法,其中,根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,建立所述第一新闻与所述第三新闻之间的二级关联关系,具体包括:
    根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,计算所述第一新闻与所述第三新闻的信息熵;
    根据所述第一新闻与所述第三新闻的信息熵,确定所述第一新闻与所述第三新闻的二级关联关系。
  16. 根据权利要求13-15任一项所述的方法,其中,所述第二新闻为多个;
    在确定所述第二新闻与第三新闻的一级关联关系之前,还包括:
    根据预设规则,为每个第二新闻查找相应的第三新闻;
    根据所述第一新闻与所述第三新闻的信息熵,确定所述第一新闻与所述第三新闻的二级关联关系,具体包括:
    根据每个第三新闻对应的第二新闻在全部第二新闻中的占比,计算所述每个第三新闻的权重值;
    根据所述每个第三新闻的权重值、所述每个第三新闻对应的信息熵,计算所述每个第三新闻与所述第一新闻的二级关联关系。
  17. 根据权利要求13-16任一项所述的方法,其中,所述第三新闻为多个;所述方法还包括:
    根据所述多个第三新闻与所述第一新闻的二级关联关系,对所述多个第三新闻 进行排序。
  18. 一种多新闻之间相关性计算装置,包括:
    第一关联关系确定模块,适于确定第一新闻与第二新闻的一级关联关系;
    第二关联关系确定模块,适于确定所述第二新闻与第三新闻的一级关联关系;
    第三关联关系确定模块,适于根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,建立所述第一新闻与所述第三新闻之间的二级关联关系。
  19. 根据权利要求17所述的装置,其中,
    所述第一关联关系确定模块还适于根据所述第一新闻的特征属性设置第一向量,根据所述第二新闻的特征属性设置第二向量,计算所述第一向量和所述第二向量的余弦相似度作为所述第一新闻与所述第二新闻的一级关联关系;
    所述第二关联关系确定模块还适于根据所述第三新闻的特征属性设置第三向量,计算所述第二向量和所述第三向量的余弦相似度作为所述第二新闻与所述第三新闻的一级关联关系。
  20. 根据权利要求18-19任一项所述的装置,其中,
    所述第三关联关系确定模块还适于根据所述第一新闻与所述第二新闻的一级关联关系、所述第二新闻与所述第三新闻的一级关联关系,计算所述第一新闻与所述第三新闻的信息熵;以及根据所述第一新闻与所述第三新闻的信息熵,确定所述第一新闻与所述第三新闻的二级关联关系。
  21. 根据权利要求18-20任一项所述的装置,其中,所述第二新闻为多个;所述装置还包括:
    查找模块,用于根据预设规则,为每个第二新闻查找相应的第三新闻;
    所述第三关联关系确定模块还适于根据每个第三新闻对应的第二新闻在全部第二新闻中的占比,计算所述每个第三新闻的权重值;以及根据所述每个第三新闻的权重值、所述每个第三新闻对应的信息熵,计算所述每个第三新闻与所述第一新闻的二级关联关系。
  22. 根据权利要求18-21任一项所述的装置,其中,所述第三新闻为多个;所述装置还包括:
    排序模块,用于根据所述多个第三新闻与所述第一新闻的二级关联关系,对所述多个第三新闻进行排序。
  23. 一种计算机程序,包括计算机可读代码,当所述计算机可读代码在计算设备上运行时,导致所述计算设备执行根据权利要求1-6中的任一个所述的确定新闻之 间相关性的方法,和/或,权利要求13-17中的任一个所述的多新闻之间相关性计算方法。
  24. 一种计算机可读介质,其中存储了如权利要求23所述的计算机程序。
PCT/CN2016/103397 2015-12-22 2016-10-26 确定新闻之间相关性、多新闻之间相关性计算方法和装置 Ceased WO2017107651A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US15/744,688 US10217025B2 (en) 2015-12-22 2016-10-26 Method and apparatus for determining relevance between news and for calculating relevance among multiple pieces of news

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN201510974431.2A CN105630766B (zh) 2015-12-22 2015-12-22 多新闻之间相关性计算方法和装置
CN201510974316.5 2015-12-22
CN201510974431.2 2015-12-22
CN201510974316.5A CN105528335B (zh) 2015-12-22 2015-12-22 确定新闻之间相关性的方法和装置

Publications (1)

Publication Number Publication Date
WO2017107651A1 true WO2017107651A1 (zh) 2017-06-29

Family

ID=59088938

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/103397 Ceased WO2017107651A1 (zh) 2015-12-22 2016-10-26 确定新闻之间相关性、多新闻之间相关性计算方法和装置

Country Status (2)

Country Link
US (1) US10217025B2 (zh)
WO (1) WO2017107651A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017107651A1 (zh) * 2015-12-22 2017-06-29 北京奇虎科技有限公司 确定新闻之间相关性、多新闻之间相关性计算方法和装置
US10860396B1 (en) 2020-01-30 2020-12-08 PagerDuty, Inc. Inline categorizing of events
JP2024075257A (ja) * 2022-11-22 2024-06-03 株式会社リコー 情報処理装置、情報処理システム、情報処理方法及びプログラム

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1959671A (zh) * 2005-10-31 2007-05-09 北大方正集团有限公司 基于文档结构的文档相似性度量方法
US20090265160A1 (en) * 2005-05-13 2009-10-22 Curtin University Of Technology Comparing text based documents
CN103164427A (zh) * 2011-12-13 2013-06-19 中国移动通信集团公司 新闻聚合方法及装置
CN103257957A (zh) * 2012-02-15 2013-08-21 深圳市腾讯计算机系统有限公司 一种基于中文分词的文本相似性识别方法及装置
CN103389975A (zh) * 2012-05-07 2013-11-13 腾讯科技(深圳)有限公司 一种新闻推荐方法及系统
CN103678645A (zh) * 2013-12-20 2014-03-26 中电长城网际系统应用有限公司 文档匹配方法和文档匹配装置
CN105528335A (zh) * 2015-12-22 2016-04-27 北京奇虎科技有限公司 确定新闻之间相关性的方法和装置
CN105630766A (zh) * 2015-12-22 2016-06-01 北京奇虎科技有限公司 多新闻之间相关性计算方法和装置

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6182008B1 (en) * 1999-06-25 2001-01-30 Microsoft Corporation Ordering destinations along a route using a shortest line approximation
US7209973B2 (en) * 2001-04-09 2007-04-24 Swsoft Holdings, Ltd. Distributed network data storage system and method
US20030163785A1 (en) * 2002-02-28 2003-08-28 Hui Chao Composing unique document layout for document differentiation
US8209278B1 (en) * 2007-03-23 2012-06-26 Jay Bradley Straus Computer editing system for common textual patterns in legal documents
US10762080B2 (en) * 2007-08-14 2020-09-01 John Nicholas and Kristin Gross Trust Temporal document sorter and method
US20120254333A1 (en) * 2010-01-07 2012-10-04 Rajarathnam Chandramouli Automated detection of deception in short and multilingual electronic messages
US20130238720A1 (en) * 2012-03-09 2013-09-12 Research In Motion Limited Method for sharing a file when multiple versions exist
CN103324666A (zh) 2013-05-14 2013-09-25 亿赞普(北京)科技有限公司 一种基于微博数据的话题跟踪方法及装置
CN104424279B (zh) 2013-08-30 2018-11-20 腾讯科技(深圳)有限公司 一种文本的相关性计算方法和装置
CN104090890B (zh) 2013-12-12 2016-05-04 深圳市腾讯计算机系统有限公司 关键词相似度获取方法、装置及服务器
CN104361062B (zh) 2014-11-03 2017-10-31 百度在线网络技术(北京)有限公司 一种关联信息的推荐方法及装置
CN104346476B (zh) 2014-11-20 2017-07-04 西安电子科技大学 基于物品相似度和网络结构的个性化物品推荐方法
CN104462323B (zh) 2014-12-02 2018-02-27 百度在线网络技术(北京)有限公司 语义相似度计算方法、搜索结果处理方法和装置
CN105022840B (zh) 2015-08-18 2018-06-05 新华网股份有限公司 一种新闻信息处理方法、新闻推荐方法和相关装置
WO2017107651A1 (zh) * 2015-12-22 2017-06-29 北京奇虎科技有限公司 确定新闻之间相关性、多新闻之间相关性计算方法和装置

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090265160A1 (en) * 2005-05-13 2009-10-22 Curtin University Of Technology Comparing text based documents
CN1959671A (zh) * 2005-10-31 2007-05-09 北大方正集团有限公司 基于文档结构的文档相似性度量方法
CN103164427A (zh) * 2011-12-13 2013-06-19 中国移动通信集团公司 新闻聚合方法及装置
CN103257957A (zh) * 2012-02-15 2013-08-21 深圳市腾讯计算机系统有限公司 一种基于中文分词的文本相似性识别方法及装置
CN103389975A (zh) * 2012-05-07 2013-11-13 腾讯科技(深圳)有限公司 一种新闻推荐方法及系统
CN103678645A (zh) * 2013-12-20 2014-03-26 中电长城网际系统应用有限公司 文档匹配方法和文档匹配装置
CN105528335A (zh) * 2015-12-22 2016-04-27 北京奇虎科技有限公司 确定新闻之间相关性的方法和装置
CN105630766A (zh) * 2015-12-22 2016-06-01 北京奇虎科技有限公司 多新闻之间相关性计算方法和装置

Also Published As

Publication number Publication date
US10217025B2 (en) 2019-02-26
US20180197045A1 (en) 2018-07-12

Similar Documents

Publication Publication Date Title
CN103699625B (zh) 基于关键词进行检索的方法及装置
CN107577687B (zh) 图像检索方法及装置
CN108920611B (zh) 文章生成方法、装置、设备及存储介质
US20160188633A1 (en) A method and apparatus for tracking microblog messages for relevancy to an entity identifiable by an associated text and an image
CN105404627B (zh) 一种用于确定搜索结果的方法与设备
CN101826099B (zh) 一种相似文档识别、文档扩散度确定的方法及系统
WO2015184992A1 (zh) 一种识别重复图片的方法、图片搜索去重方法及其装置
Letessier et al. Scalable mining of small visual objects
CN107679186B (zh) 基于实体库进行实体搜索的方法及装置
WO2018184427A1 (zh) 一种推荐司法知识的方法及装置
GB2493587A (en) Entity resolution system identifying non-distinct names in a set of names
WO2020135247A1 (zh) 法律文书解析方法及装置
CN110222074A (zh) 索引查找方法、查找装置、电子设备及存储介质
CN109271614A (zh) 一种数据查重方法
CN105653546B (zh) 一种目标主题的检索方法和系统
CN107832444A (zh) 基于搜索日志的事件发现方法及装置
Piryani et al. Detecting temporal ambiguity in questions
WO2016058520A1 (zh) 人脸图片人名识别方法和装置
Jhawar et al. Author name disambiguation in PubMed using ensemble-based classification algorithms
Zhang et al. Scalable visual instance mining with threads of features
CN107644033B (zh) 一种在非关系型数据库中进行数据查询的方法与设备
CN104572904A (zh) 一种标签关联程度的确定方法及装置
US10217025B2 (en) Method and apparatus for determining relevance between news and for calculating relevance among multiple pieces of news
WO2015176624A1 (zh) 突发时效性查询词识别方法和系统
CN109992708B (zh) 一种元数据查询的方法、装置、设备以及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16877462

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16877462

Country of ref document: EP

Kind code of ref document: A1