WO2014180196A1 - 信息推荐处理方法及装置 - Google Patents

信息推荐处理方法及装置 Download PDF

Info

Publication number
WO2014180196A1
WO2014180196A1 PCT/CN2014/074403 CN2014074403W WO2014180196A1 WO 2014180196 A1 WO2014180196 A1 WO 2014180196A1 CN 2014074403 W CN2014074403 W CN 2014074403W WO 2014180196 A1 WO2014180196 A1 WO 2014180196A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
recommended
range
recommendation
time
Prior art date
Application number
PCT/CN2014/074403
Other languages
English (en)
French (fr)
Inventor
丘志宏
齐泉
Original Assignee
华为技术有限公司
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
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2014180196A1 publication Critical patent/WO2014180196A1/zh
Priority to US14/795,189 priority Critical patent/US20150324448A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28Databases characterised by their database models, e.g. relational or object models
    • G06F16/284Relational databases
    • G06F16/285Clustering or classification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/903Querying
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/22Indexing; Data structures therefor; Storage structures
    • G06F16/2228Indexing structures
    • G06F16/2237Vectors, bitmaps or matrices
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24Querying
    • G06F16/245Query processing
    • G06F16/2458Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2477Temporal data queries
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9535Search customisation based on user profiles and personalisation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/955Retrieval from the web using information identifiers, e.g. uniform resource locators [URL]

Definitions

  • the present invention relates to communication technologies, and in particular, to an information recommendation processing method and apparatus. Background technique
  • the embodiment of the invention provides an information recommendation processing method and device for solving the problem of recommending outdated information to a user.
  • a first aspect of the embodiments of the present invention provides a method for information recommendation processing, including: acquiring an information set, where the information set includes a plurality of pieces of information to be recommended, and the information to be recommended includes a time for identifying a time for generating the information to be recommended.
  • the plurality of pieces of to-be-recommended information in the information set are divided into the to-be-recommended information in the range and the to-be-recommended information in the range according to the information recommendation time range information and the time label corresponding to each to-be-recommended information; Determining the to-be-recommended information for the recommendation in the to-be-recommended information in the range; wherein, the time identified by the time tag of the to-be-recommended information in the range is included in the information recommendation time range.
  • the determining, to the recommended information, the recommended information to be recommended in the to-be-recommended information in the range includes:
  • the recommendation information includes the number of the keywords, respectively, and obtains an information gain corresponding to the keyword;
  • the determining, according to the information gain, the recommended information in the to-be-recommended information in the range The information to be recommended includes:
  • the digital vector matrix is composed according to the digital vector corresponding to the information to be recommended in each range, and the preset clustering or classification algorithm is applied to obtain the information to be recommended in the recommended range.
  • the method further includes:
  • the information to be recommended in the range is filtered according to the information gain corresponding to each keyword, and the digital vector corresponding to the information to be recommended in the filtered range is obtained; correspondingly, the information to be recommended according to each range is corresponding.
  • the digital vector consists of a number vector matrix including:
  • the digital vector matrix is composed according to a digital vector corresponding to the information to be recommended in the filtered range.
  • the search term includes: a search term input by the user; or, from the associated information of the user The search term taken.
  • a second aspect of the embodiments of the present invention provides a new type of recommended processing apparatus, including:
  • An obtaining module configured to obtain an information set, where the information set includes a plurality of pieces of information to be recommended, where the information to be recommended includes a time label for identifying a time when the information to be recommended is generated;
  • a dividing module configured to divide the plurality of pieces of to-be-recommended information in the information set into the to-be-recommended information in the range and the to-be-recommended information in the range according to the information recommendation time range information and the time label corresponding to each to-be-recommended information;
  • a recommendation module configured to determine, in the to-be-recommended information in the range, the information to be recommended for recommendation
  • the time indicated by the time stamp of the information to be recommended in the range is included in the information recommendation time range.
  • the recommendation module is specifically configured to acquire at least one keyword included in the to-be-recommended information in the range, and according to the range
  • the recommended information and the number of the information to be recommended outside the range, and the information to be recommended in the range and the information to be recommended in the range respectively include the number of the keywords, and obtain the information gain corresponding to the keyword;
  • the information gain within the range of the to-be-recommended information determines the to-be-recommended information for recommendation.
  • the recommended module includes:
  • An obtaining unit configured to obtain, according to an information gain corresponding to each keyword included in the to-be-recommended information in the range, a digital vector corresponding to the to-be-recommended information in each range;
  • the recommendation unit is configured to form a digital vector matrix according to the digital vector corresponding to the information to be recommended in each range, and apply a preset clustering or classification algorithm to obtain the information to be recommended in the recommended range.
  • the device further includes:
  • the acquiring module is specifically configured to acquire multiple to-be-acquired according to the search term
  • the recommendation information forms the information set; the search term includes: a search term input by the user; or a search term extracted from the associated information of the user.
  • a third aspect of the embodiments of the present invention provides an information recommendation processing apparatus, including:
  • the memory is configured to store an instruction
  • the processor coupled to the memory, is configured to execute an instruction stored in the memory, for acquiring an information set, where the information set includes a plurality of pieces of information to be recommended, and the information to be recommended includes The time label of the to-be-recommended information generation time; the plurality of pieces of to-be-recommended information in the information set are divided into the to-be-recommended information and the range to be recommended according to the information recommendation time range information and the time label corresponding to each to-be-recommended information.
  • the information to be recommended for the recommendation is determined in the to-be-recommended information in the range; wherein the time identified by the time tag of the information to be recommended in the range is included in the information recommendation time range.
  • the processor is specifically configured to acquire at least one keyword included in the to-be-recommended information in the range, and according to the range
  • the recommended information and the number of the information to be recommended outside the range, and the information to be recommended in the range and the information to be recommended in the range respectively include the number of the keywords, and obtain the information gain corresponding to the keyword;
  • the information gain determines the to-be-recommended information for recommendation in the to-be-recommended information in the range.
  • the processor is specifically configured to use, according to the keywords included in the to-be-recommended information in the range Corresponding information gain, obtaining a digital vector corresponding to the information to be recommended in each range; composing a digital vector matrix according to the digital vector corresponding to the information to be recommended in each range, applying a preset clustering or classification algorithm to obtain a range for recommendation Information to be recommended.
  • the processor is further configured to: The recommended information is filtered, and the digital vector corresponding to the information to be recommended in the filtered range is obtained; and the digital vector matrix is formed according to the digital vector corresponding to the information to be recommended in the filtered range.
  • the recommendation information forms the information set; wherein, the search term includes: a search term input by the user; or a search term extracted from the associated information of the user.
  • the obtained information to be recommended is divided into the to-be-recommended information in the range and the to-be-recommended information in the range according to the information recommendation time range information and the time label corresponding to each information to be recommended, and the information to be recommended in the range is
  • the information to be recommended for recommendation is selected to the user, so that the information recommended by the user considers the time stamp of the information, and the information recommended to the user is time-sensitive.
  • FIG. 1 is a schematic flowchart of Embodiment 1 of an information recommendation processing method provided by the present invention
  • FIG. 2 is a schematic flowchart of Embodiment 2 of an information recommendation processing method provided by the present invention
  • FIG. 3 is a schematic diagram of Embodiment 1 of an information recommendation processing apparatus provided by the present invention
  • FIG. 4 is a schematic structural diagram of Embodiment 2 of the information recommendation processing apparatus provided by the present invention
  • FIG. 5 is a schematic structural diagram of Embodiment 3 of the information recommendation processing apparatus provided by the present invention
  • the symbol "*" represents a multiplication sign in the formula, and the symbol “/” is expressed in the formula.
  • the division sign, the symbol "/” is indicated or related in the text part.
  • Embodiment 1 is a schematic flowchart of Embodiment 1 of an information recommendation processing method provided by the present invention.
  • the execution body of the method may be an information recommendation processing device, and the device may be integrated into servers of different websites. As shown in FIG. 1 , the process includes :
  • the search engine may obtain multiple pieces of information on each website, or directly obtain multiple pieces of information or all information of a website, and may also de-duplicate the obtained information to form an information set. Just exclude the exact same information.
  • the plurality of pieces of to-be-recommended information in the information set are divided into the to-be-recommended information in the range and the to-be-recommended information in the range according to the information recommendation time range information and the time label corresponding to each information to be recommended.
  • time indicated by the time label of the information to be recommended in the above range is included in the above information recommendation time range.
  • the information recommendation time range may be determined according to the attribute of the information to be recommended, for example, for "news", then the information recommendation time range is the day.
  • the information recommendation time range may also be determined according to the record of recommending information to the user. For example, the user logs in to the microblog at 8:00 in the morning, and the microblog recommends some information to the user. The user logs in to the microblog again at 12 noon, and then recommends to the user. Recommended information for updates between 8 and 12 o'clock.
  • the information recommendation time range may also be determined according to the received time range of the user input, for example, the user logs in to the microblog, and the time option is set in the search engine of the microblog. The user can customize or select a time range, and the microblog recommends to the user. Information within the time range entered by the user.
  • the information to be recommended may be sorted according to the time label corresponding to each information to be recommended in the above information set, and the information to be recommended is divided into the to-be-recommended information and the to-be-recommended information in the range according to the information recommendation time range.
  • S103 Determine to-be-recommended information for recommendation within the to-be-recommended information in the foregoing range. After the information to be recommended and the information to be recommended outside the scope are determined, not all the information in the scope is recommended to the user, but the screening is performed again, for example, some hot information or information of interest to the user is recommended to the user.
  • the information recommendation time range information and the time corresponding to each to-be-recommended information The inter-label is divided into the to-be-recommended information in the range and the to-be-recommended information in the range, and the information to be recommended for recommendation is selected in the to-be-recommended information in the range, so that the information recommended by the user is considered.
  • the time stamp of the information realizes that the information recommended to the user is time-sensitive.
  • Embodiment 2 is a schematic flowchart of Embodiment 2 of an information recommendation processing method provided by the present invention, where
  • determining to-be-recommended information for recommendation in the to-be-recommended information in the foregoing range specifically, acquiring at least one keyword included in the to-be-recommended information in the foregoing range, and according to the to-be-recommended information in the above range and the foregoing range
  • the number of information to be recommended, and the information to be recommended and the information to be recommended in the above range respectively include the number of the keywords, and obtain the information gain corresponding to the keyword, and determine the information to be recommended in the above range according to the information gain. Recommended information for recommendation.
  • an algorithm for word frequency, relative word frequency or anti-document word frequency may be used, and the to-be-recommended information for recommendation is determined according to the to-be-recommended information in the above range and the occurrence frequency of the to-be-recommended information words in the above range.
  • the number of to-be-recommended information and the number of to-be-recommended information in the above range, and the to-be-recommended information and the out-of-scope recommendation information in the above range respectively include the number of the keywords, and obtain the information gain corresponding to the keyword.
  • the method includes:
  • S20K divides all the information in the information set into words, specifically, after the information to be recommended and the information to be recommended outside the range, the scores are divided into their respective subsets. For example, in the information to be recommended in the scope, there is a message "# Favorite mobile phone brand # of course, Huawei is being used! Support domestic products!, using word segmentation technology to cut into words and then convert to "the favorite mobile phone brand is of course We are using Huawei to support Chinese goods. Ten words, in which word segmentation technology will remove the stop word "".
  • P- is the ratio of the information to be recommended outside the range to the above information set.
  • the information in the above information sets has information. 126569.
  • H ( C ) -20640 / 126569 * (log(20640 / 126569) ) - 105929 / 126569 * ((log(105929 I 126569) )).
  • H(CIT) P(t+)*H(Clt+)+P(t-)*H(Clt-), and H(CIT) is used to know whether or not words are included in each piece of information.
  • T the above information set is classified according to the uncertainty of the information to be recommended and the information to be recommended outside the scope.
  • the word T appears, marked as t+, the word T does not appear, and is marked as t-
  • P(t+) represents the ratio of the number of pieces of information containing the word T to the total amount of information in the above information set
  • H(Clt+) indicates that the above information set contains
  • P(t-) represents the ratio of the number of information not including the word T to the total number of information in the above information set
  • H(Clt-) indicates the information subset of the above information set not including the word T Information entropy.
  • H(CIT) P(t+)*(-(p+lt+)*log(p+lt+)-(p-lt+)*log(p-lt+))+P(t-)*(-(p+ Lt-)*log(p+lt -) -(p-
  • the information set contains the proportion of the total number of pieces of the word ⁇ , taking the above "national goods" as an example.
  • (p+lt+) 20491/125531.
  • (p-lt+) is the ratio of the number of pieces of information containing the word T in the information to be recommended outside the range to the total number of pieces of the word T in the above information set
  • (p+ Lt-) is the ratio of the number of pieces of information in the to-be-recommended information that does not contain the word T to the total number of pieces of information in the above information set that does not contain the word T
  • (p-lt-) is the out-of-range information to be recommended.
  • the number of pieces of information of T accounts for the proportion of the total number of pieces of information in the above information set that does not contain the word ⁇ .
  • the calculation method is used to calculate the information gain values of the respective words after the segmentation, and the information to be recommended for recommendation is selected according to the calculated information gain value.
  • determining the to-be-recommended information for recommendation in the to-be-recommended information in the above range according to the information gain specifically, obtaining information in each range according to information gain corresponding to each keyword included in the information in the above range A digital vector corresponding to the information to be recommended; then, a digital vector matrix is formed according to the digital vector corresponding to the information to be recommended in each range, and a preset clustering or classification algorithm is applied to obtain the information to be recommended in the recommended range.
  • the information to be recommended in the range is represented as the above-mentioned digital vector, and then these digital vectors are formed into a vector matrix.
  • Table 2 is the partial result of the microblog website outputted by the clustering algorithm on the basis of the processing of multiple microblogs by the above embodiment:
  • semantic analysis tools can be used to classify and classify the central phrases of each class into a useful piece of information that is recommended to the user.
  • the information to be recommended in the above range may be filtered according to the information gain corresponding to each keyword, and the digital vector corresponding to the information to be recommended in the filtered range may be obtained;
  • the digital vector matrix corresponding to the to-be-recommended information in each range constitutes a digital vector matrix, and specifically, the digital vector matrix is formed according to the digital vector corresponding to the to-be-recommended information in the filtered range. That is, after calculating the information gain of each word, the words may be sorted according to the level of the information gain value, and the information of the words whose information gain is less than the preset threshold is deleted from the information to be recommended in the range, thereby avoiding Users recommend some repetitive spam, advertisements, etc.
  • the information appearing in the negative example is generally outdated information, and some recurring information will appear in both the information to be recommended in the range and the information to be recommended in the range, such as an advertisement.
  • the information recommendation time range is the same day, then the number of times the advertisement will appear in the recommended information outside the range will be much larger than the number of times the advertisement will appear in the recommended information in the range, according to the above formula (5)
  • Calculate the information gain of the words contained in this advertisement will be very low, then the advertisement will be deleted when recommending information to the user on the same day, and will not be recommended to the user, thus avoiding the user seeing some recurring information. And outdated information.
  • the obtaining information set may be configured to obtain a plurality of pieces of information to be recommended according to the search term to form the information set; wherein the search term may be: (1) a search term input by the user; or, (2) A search term extracted from the user's associated information. This allows the user's interests to be taken into account before recommending information to the user, so that the information recommended to the user is information of interest to the user.
  • the user can directly input some search words in the search engine, and the related information is obtained by the search engine.
  • the search term may be extracted from some information customized by the user, for example, the user-defined tag information in the microblog may be directly extracted as a search term; or may be extracted according to the browsing record of the user. Search words, such as recent users on the e-commerce website are browsing history books many times, then Use "history books" as a search term.
  • Weibo servers do not allow other search engines to perform large-scale information search on their websites. Then, Weibo's own search tool can periodically use the above search words to its information. The search is performed, and after being deduplicated, it is saved locally, and is acquired by the information recommendation processing device through a dedicated search interface.
  • the information to be recommended by the user is obtained according to the search term associated with the user, and the information to be recommended is divided into the information to be recommended in the range according to the information recommendation time range information and the time label corresponding to each information to be recommended. And the information to be recommended outside the scope, and the information to be recommended for recommendation is selected to the user in the information to be recommended in the range, so that the information recommended by the user considers the time label of the information, and the information recommended to the user is time-sensitive. Moreover, the information to be recommended in the range can be filtered according to the gain information of each keyword, and some repetitive information and advertisement information and the like can be removed.
  • FIG. 3 is a schematic structural diagram of Embodiment 1 of the information recommendation processing apparatus provided by the present invention.
  • the apparatus may be integrated into servers of different websites.
  • the apparatus includes: an obtaining module 301, a dividing module 302, and a recommending module 303. , among them:
  • the obtaining module 301 is configured to obtain an information set, where the information set includes a plurality of pieces of information to be recommended, the information to be recommended includes a time label for identifying a time when the information to be recommended is generated, and a dividing module 302, configured to recommend according to the information.
  • the time range information and the time label corresponding to each of the to-be-recommended information, the plurality of to-be-recommended information in the information set are divided into the to-be-recommended information in the range and the to-be-recommended information in the range; the recommendation module 303 is configured to be in the range
  • the information to be recommended for the recommendation is determined in the recommendation information.
  • the time indicated by the time label of the information to be recommended in the range is included in the information recommendation time range.
  • Embodiment 4 is a schematic structural diagram of Embodiment 2 of an information recommendation processing apparatus according to the present invention.
  • the recommendation module 303 is specifically configured to obtain at least one keyword included in the to-be-recommended information in the foregoing range, and according to the to-be-recommended information in the range and the number of to-be-recommended information in the range, and the The information to be recommended in the range and the information to be recommended in the range respectively include the number of the keywords, and obtain the information gain corresponding to the keyword; and determine the Recommended information. Further, as shown in FIG. 4, the recommendation module 303 includes an obtaining unit 401 and a recommending unit 402, where:
  • the obtaining unit 401 is configured to obtain a digital vector corresponding to the to-be-recommended information in each range according to the information gain corresponding to each keyword included in the to-be-recommended information in the range, and the recommendation unit 402 is configured to use the information to be recommended according to each range.
  • the corresponding digital vector constitutes a digital vector matrix, and a preset clustering or classification algorithm is applied to obtain information to be recommended within the recommended range.
  • FIG. 5 is a schematic structural diagram of Embodiment 3 of the information recommendation processing apparatus provided by the present invention.
  • the apparatus further includes: a screening module 501, where the screening module 501 is configured according to FIG.
  • the information gain corresponding to each keyword, the information to be recommended in the range is filtered, and the digital vector corresponding to the information to be recommended in the filtered range is obtained;
  • the recommendation unit 402 is configured to be used according to the selected range.
  • the digital vector corresponding to the recommendation information constitutes the digital vector matrix.
  • the obtaining module 301 is specifically configured to obtain a plurality of pieces of information to be recommended according to the search term, and form the information set.
  • the search term includes: a search term input by the user; or, from the associated information of the user. Extracted search terms.
  • FIG. 6 is a schematic structural diagram of Embodiment 4 of the information recommendation processing apparatus provided by the present invention.
  • the apparatus includes: a memory 601 and a processor 602, wherein the memory 601 is configured to store an instruction, the processor 602 and the memory. Coupled, the processor 602 is configured to execute instructions stored in the memory, specifically:
  • the processor 602 is configured to acquire a set of information, where the information set includes a plurality of pieces of information to be recommended, and the information to be recommended includes a time label for identifying a time when the information to be recommended is generated; a time tag corresponding to the information, the plurality of pieces of information to be recommended in the information set are divided into the to-be-recommended information in the range and the to-be-recommended information in the range; and the to-be-recommended information for recommendation is determined in the to-be-recommended information in the range; The time indicated by the time stamp of the information to be recommended in the range is included in the information recommendation time range.
  • the processor 602 is specifically configured to acquire at least one keyword included in the to-be-recommended information in the range, and according to the to-be-recommended information in the range and the number of to-be-recommended information in the range, and the The to-be-recommended information in the range and the to-be-recommended information in the range include And the number of the keywords, the information gain corresponding to the keyword is obtained; and the to-be-recommended information for recommendation is determined in the to-be-recommended information in the range according to the information gain.
  • the processor 602 is configured to obtain, according to the information gain corresponding to each keyword included in the to-be-recommended information in the range, a digital vector corresponding to the to-be-recommended information in each range;
  • the digital vector corresponding to the recommendation information constitutes a digital vector matrix, and a preset clustering or classification algorithm is applied to obtain information to be recommended in the recommended range.
  • the processor 602 is further configured to: filter, according to information gains corresponding to the keywords, the information to be recommended in the range, and obtain a digital vector corresponding to the information to be recommended in the filtered range;
  • the digital vector corresponding to the information to be recommended in the range constitutes the digital vector matrix.
  • the processor 602 is specifically configured to acquire a plurality of pieces of to-be-recommended information according to the search term, and form the information set.
  • the search term includes: a search term input by the user; or, extracting from the associated information of the user. Search term.
  • the foregoing apparatus may be used to implement the foregoing method embodiments, and the implementation manners are similar, and are not described herein again.
  • the disclosed apparatus and method may be implemented in other manners.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.
  • the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e., may be located in one place, or may be distributed over multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units.
  • the above-described integrated unit implemented in the form of a software functional unit can be stored in a computer readable storage medium.
  • the software functional unit is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to perform the method of various embodiments of the present invention.
  • the foregoing storage medium includes: a USB flash drive, a removable hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like, which can store program code. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种信息推荐处理方法及装置,该方法包括:获取信息集,所述信息集中包括多条待推荐信息,所述待推荐信息包括用于标识所述待推荐信息产生时间的时间标签;根据信息推荐时间范围信息以及各待推荐信息对应的时间标签,将所述信息集中的多条待推荐信息划分为范围内待推荐信息和范围外待推荐信息;在所述范围内待推荐信息内确定用于推荐的待推荐信息。该方法由于给用户推荐的信息考虑了信息的时间标签,因此给用户推荐的信息时效性高。

Description

信息推荐处理方法及装置
技术领域
本发明涉及通信技术, 尤其涉及一种信息推荐处理方法及装置。 背景技术
随着互联网的不断发展, 网络上的信息量呈现出爆炸性地增长, 信息 的更新频率也越来越快, 当用户浏览网页的时候会有各式各样的信息呈现 给用户, 使得用户应接不暇。 特别地, 在电子商务领域, 随着电子商务规 模的不断扩大, 商品个数和种类快速增长, 顾客需要花费大量的时间才能 找到自己想买的商品。这种浏览大量无关信息和产品的过程无疑会使淹没 在信息过载问题中的消费者不断流失。 在互联网浏览领域, 随着博客、 维 基、 微博的发展, 大量的网络信息由用户个人产生, 信息的组织散乱, 质 量和可信度参差不齐,使得用户需要花费大量时间才能找到自己感兴趣的
I Ft自、。
现有技术中, 为了解决上述问题, 采用个性化推荐的方式向用户推荐 感兴趣的信息和商品。
但是, 随着信息更新越来越快, 现有技术中, 向用户推荐的信息很多 时候是已经过时的信息, 给用户带来信息浏览的负担。 发明内容
本发明实施例提供一种信息推荐处理方法及装置, 用于解决向用户推 荐过时信息的问题。
本发明实施例第一方面提供一种信息推荐处理方法, 包括: 获取信息集, 所述信息集中包括多条待推荐信息, 所述待推荐信息包 括用于标识所述待推荐信息产生时间的时间标签;
根据信息推荐时间范围信息以及各待推荐信息对应的时间标签, 将所 述信息集中的多条待推荐信息划分为范围内待推荐信息和范围外待推荐 信息; 在所述范围内待推荐信息内确定用于推荐的待推荐信息; 其中, 所述范围内待推荐信息的时间标签所标识的时间包含在所述信 息推荐时间范围内。
结合第一方面, 在第一方面的第一种可能的实施方式中, 所述在所述 范围内待推荐信息内确定用于推荐的待推荐信息包括:
获取所述范围内待推荐信息所包括的至少一个关键词, 并根据所述范 围内待推荐信息和所述范围外待推荐信息的数量, 以及所述范围内待推荐 信息和所述范围外待推荐信息分别包括所述关键词的数量, 获取所述关键 词对应的信息增益;
根据所述信息增益在所述范围内待推荐信息确定所述用于推荐的待 推荐信息。
结合第一方面的第一种可能的实施方式, 在第一方面的第二种可能的 实施方式中, 所述根据所述信息增益在所述范围内待推荐信息中确定所述 用于推荐的待推荐信息包括:
根据所述范围内待推荐信息所包括的各关键词所对应的信息增益, 获 取各范围内待推荐信息对应的数字向量;
根据各范围内待推荐信息对应的数字向量组成数字向量矩阵, 应用预 设的聚类或分类算法, 获取用于推荐的范围内待推荐信息。
结合第一方面的第二种可能的实施方式, 在第一方面的第三种可能的 实施方式中, 所述方法还包括:
根据各关键词所对应的信息增益, 对所述范围内待推荐信息进行筛 选, 并获取经过筛选的范围内待推荐信息所对应的数字向量; 相应地, 所述根据各范围内待推荐信息对应的数字向量组成数字向量矩阵包 括:
根据经过筛选的范围内待推荐信息所对应的数字向量, 组成所述数字 向量矩阵。
结合第一方面至第一方面的第三种可能的实施方式中任一项, 在第一 方面的第四种可能的实施方式中, 所述获取信息集包括:
根据搜索词获取多条待推荐信息, 形成所述信息集;
所述搜索词包括: 用户输入的搜索词; 或者, 从用户的关联信息中提 取的搜索词。
本发明实施例第二方面提供一种新型推荐处理装置, 包括:
获取模块, 用于获取信息集, 所述信息集中包括多条待推荐信息, 所 述待推荐信息包括用于标识所述待推荐信息产生时间的时间标签;
划分模块, 用于根据信息推荐时间范围信息以及各待推荐信息对应的 时间标签, 将所述信息集中的多条待推荐信息划分为范围内待推荐信息和 范围外待推荐信息;
推荐模块, 用于在所述范围内待推荐信息内确定用于推荐的待推荐信 息;
其中, 所述范围内待推荐信息的时间标签所标识的时间包含在所述信 息推荐时间范围内。
结合第二方面, 在第二方面的第一种可能的实施方式中, 所述推荐模 块, 具体用于获取所述范围内待推荐信息所包括的至少一个关键词, 并根 据所述范围内待推荐信息和所述范围外待推荐信息的数量, 以及所述范围 内待推荐信息和所述范围外待推荐信息分别包括所述关键词的数量, 获取 所述关键词对应的信息增益; 根据所述信息增益在所述范围内待推荐信息 确定所述用于推荐的待推荐信息。
结合第二方面的第一种可能的实施方式, 在第二方面的第二种可能的 实施方式中, 所述推荐模块, 包括:
获取单元, 用于根据所述范围内待推荐信息所包括的各关键词所对应 的信息增益, 获取各范围内待推荐信息对应的数字向量;
推荐单元, 用于根据各范围内待推荐信息对应的数字向量组成数字向 量矩阵,应用预设的聚类或分类算法,获取用于推荐的范围内待推荐信息。
结合第二方面的第二种可能的实施方式, 在第二方面的第三种可能的 实施方式种, 所述装置还包括:
筛选模块, 用于根据各关键词所对应的信息增益, 对所述范围内待推 荐信息进行筛选, 并获取经过筛选的范围内待推荐信息所对应的数字向 所述推荐单元, 用于根据经过筛选的范围内待推荐信息所对应的数字 向量, 组成所述数字向量矩阵。 结合第二方面至第二方面的第三种可能的实施方式中任一项, 在第二 方面的第四种可能的实施方式中, 所述获取模块, 具体用于根据搜索词获 取多条待推荐信息, 形成所述信息集; 所述搜索词包括: 用户输入的搜索 词; 或者, 从用户的关联信息中提取的搜索词。
本发明实施例第三方面提供一种信息推荐处理装置, 包括:
存储器和处理器, 其中:
所述存储器, 用于存储指令;
所述处理器, 与所述存储器耦合, 被配置为执行存储在所述存储器中 的指令, 用于获取信息集, 所述信息集中包括多条待推荐信息, 所述待推 荐信息包括用于标识所述待推荐信息产生时间的时间标签; 根据信息推荐 时间范围信息以及各待推荐信息对应的时间标签, 将所述信息集中的多条 待推荐信息划分为范围内待推荐信息和范围外待推荐信息; 在所述范围内 待推荐信息内确定用于推荐的待推荐信息; 其中, 所述范围内待推荐信息 的时间标签所标识的时间包含在所述信息推荐时间范围内。
结合第三方面,在第三方面的第一种可能的实施方式中,所述处理器, 具体用于获取所述范围内待推荐信息所包括的至少一个关键词, 并根据所 述范围内待推荐信息和所述范围外待推荐信息的数量, 以及所述范围内待 推荐信息和所述范围外待推荐信息分别包括所述关键词的数量, 获取所述 关键词对应的信息增益; 根据所述信息增益在所述范围内待推荐信息中确 定所述用于推荐的待推荐信息。
结合第三方面的第一种可能的实施方式, 在第三方面的第二种可能的 实施方式中, 所述处理器, 具体用于根据所述范围内待推荐信息所包括的 各关键词所对应的信息增益, 获取各范围内待推荐信息对应的数字向量; 根据各范围内待推荐信息对应的数字向量组成数字向量矩阵, 应用预设的 聚类或分类算法, 获取用于推荐的范围内待推荐信息。
结合第三方面的第二种可能的实施方式, 在第三方面的第三种可能的 实施方式中, 所述处理器, 还用于根据各关键词所对应的信息增益, 对所 述范围内待推荐信息进行筛选, 并获取经过筛选的范围内待推荐信息所对 应的数字向量; 根据经过筛选的范围内待推荐信息所对应的数字向量, 组 成所述数字向量矩阵。 结合第三方面至第三方面的第三种可能的实施方式中任一项, 在第三 方面的第四种可能的实施方式中, 所述处理器, 具体用于根据搜索词获取 多条待推荐信息, 形成所述信息集; 其中, 所述搜索词包括: 用户输入的 搜索词; 或者, 从用户的关联信息中提取的搜索词。
本发明实施例中, 根据信息推荐时间范围信息以及各待推荐信息对应 的时间标签, 将获取到的待推荐信息划分为范围内待推荐信息和范围外待 推荐信息, 并在范围内待推荐信息中选择用于推荐的待推荐信息给用户, 这样给用户推荐的信息考虑了信息的时间标签, 实现了给用户推荐的信息 时效性高。 附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案, 下面将对 实施例或现有技术描述中所需要使用的附图作一简单地介绍, 显而易见 地, 下面描述中的附图是本发明的一些实施例, 对于本领域普通技术人员 来讲, 在不付出创造性劳动性的前提下, 还可以根据这些附图获得其他的 附图。
图 1为本发明提供的信息推荐处理方法实施例一的流程示意图; 图 2为本发明提供的信息推荐处理方法实施例二的流程示意图; 图 3为本发明提供的信息推荐处理装置实施例一的结构示意图; 图 4为本发明提供的信息推荐处理装置实施例二的结构示意图; 图 5为本发明提供的信息推荐处理装置实施例三的结构示意图; 图 6为本发明提供的信息推荐处理装置实施例四的结构示意图。 具体实施方式
为使本发明实施例的目的、 技术方案和优点更加清楚, 下面将结合本 发明实施例中的附图, 对本发明实施例中的技术方案进行清楚、 完整地描 述, 显然,所描述的实施例是本发明一部分实施例, 而不是全部的实施例。 基于本发明中的实施例, 本领域普通技术人员在没有作出创造性劳动前提 下所获得的所有其他实施例, 都属于本发明保护的范围。
本发明实施例中符号 " * "在公式中表示乘号, 符号 "/"在公式中表 示除号, 符号 "/"在文字部分表示或者关系。
图 1为本发明提供的信息推荐处理方法实施例一的流程示意图, 该方 法的执行主体可以是信息推荐处理装置, 该装置可以集成于不同网站的服 务器中, 如图 1所示, 该流程包括:
S101、 获取信息集, 该信息集中包括多条待推荐信息, 该待推荐信息 包括用于标识上述带推荐信息产生时间的时间标签。
具体地可以是通过搜索引擎在各个网站上获取到多条信息, 或者直接 随机获取某网站的多条信息或所有信息, 还可以对获取到的信息进行去 重, 组成信息集, 这里去重一般只是排除掉完全一样的信息。
S102、 根据信息推荐时间范围信息以及各待推荐信息对应的时间标 签, 将上述信息集中的多条待推荐信息划分为范围内待推荐信息和范围外 待推荐信息。
需要说明的是, 上述范围内待推荐信息的时间标签所标识的时间包含 在上述信息推荐时间范围内。
可以根据待推荐信息的属性确定信息推荐时间范围, 例如对于 "新 闻" , 那么信息推荐时间范围就为当天。 也可以根据向用户推荐信息的记 录确定信息推荐时间范围, 例如, 用户早上 8点登录了微博, 微博向用户 推荐了一些信息, 该用户在中午 12点再次登录微博, 则向用户推荐 8点 -12点之间的更新的推荐信息。 还可以根据接收到的用户输入的时间范围 确定信息推荐时间范围, 例如用户登录微博, 在微博的搜索引擎中设置时 间选项, 用户可自定义或选择一个时间范围, 则微博向用户推荐用户输入 的时间范围内的信息。
可以在上述信息集中根据各待推荐信息对应的时间标签, 对这些待推 荐信息进行排序, 再根据信息推荐时间范围将这些待推荐信息划分为范围 内待推荐信息和范围外待推荐信息。
S103、 在上述范围内待推荐信息内确定用于推荐的待推荐信息。 确定了范围内待推荐信息和范围外待推荐信息之后, 并不是将所有范 围内的信息都推荐给用户, 而是要进行再次筛选, 例如将一些热点信息或 者用户感兴趣的信息推荐给用户。
本实施例中, 根据信息推荐时间范围信息以及各待推荐信息对应的时 间标签, 将获取到的待推荐信息划分为范围内待推荐信息和范围外待推荐 信息, 并在范围内待推荐信息中选择用于推荐的待推荐信息给用户, 这样 给用户推荐的信息考虑了信息的时间标签, 实现了给用户推荐的信息时效 性高。
图 2为本发明提供的信息推荐处理方法实施例二的流程示意图, 上述
S103中,在上述范围内待推荐信息内确定用于推荐的待推荐信息,具体为, 获取上述范围内待推荐信息所包括的至少一个关键词, 并根据上述范围内 待推荐信息和上述范围外待推荐信息的数量, 以及上述范围内待推荐信息 和范围外待推荐信息分别包括该关键词的数量, 获取该关键词对应的信息 增益, 根据该信息增益在上述范围内待推荐信息中确定用于推荐的待推荐 信息。 另外, 除了采用信息增益, 也可以采用词频、 相对词频或反文档词 频的算法, 根据上述范围内待推荐信息和上述范围外待推荐信息词语的出 现频率, 确定用于推荐的待推荐信息。
举例说明根据上述范围内待推荐信息和上述范围外待推荐信息的数 量, 以及上述范围内待推荐信息和范围外待推荐信息分别包括该关键词的 数量, 获取该关键词对应的信息增益。 假设将截止计算当天 "一周内" 的 信息划为范围内待推荐信息, 范围内待推荐信息有 10640条, 范围外待推 荐信息有 105929条。 具体地, 该方法包括:
S20K 将信息集中的所有信息切分成词语, 具体地可以是在划分完范 围内待推荐信息和范围外待推荐信息之后, 分别在各自的子集中进行划 分。 例如在范围内待推荐信息中, 有一条信息为 " #最喜欢的手机品牌 #当 然是正在使用的华为呀! 支持国货! " , 采用分词技术切分成词语后转换 为 "最喜欢 手机 品牌 当然是 正在 使用 华为 呀 支持 国货" 十个词 语, 其中分词技术会去掉停用词 "的" 。
S202、根据范围内待推荐信息和范围外待推荐信息的数量计算信息熵
H ( C) , 具体地, 采用公式 (1 ) : H ( C) =- (p+) *log (p+) - (P- ) *log (P- ) 计算信息熵, 其中 p+为范围内待推荐信息占上述信息集的比例, P- 为范围外待推荐信息占上述信息集的比例, 本发明实施例中只划分了范围 内和范围外两种情况, 因此 p+与 P-的和为 1。 假设范围内待推荐信息有 10640条, 范围外待推荐信息有 105929条, 则上述信息集中共有信息 126569条。 H ( C ) =-20640 / 126569 * (log(20640 / 126569) ) - 105929 / 126569 * ((log(105929 I 126569) ))。
S203、 计算上述切分后各词语的条件熵 H(CIT)。 以 "国货"作为关键 词为例, 以表 1表示包含该关键词的信息的条数的一个统计结果,
表 1
Figure imgf000009_0001
采用公式 (2 ) : H(CIT)=P(t+)*H(Clt+)+P(t-)*H(Clt-)计算上述条件熵, H(CIT)表示知道各条信息中是否包含词 T的条件下, 上述信息集按照范围 内待推荐信息和范围外待推荐信息分类的不确定程度。 其中, 词 T出现, 标记为 t+, 词 T不出现, 标记为 t-, P(t+)表示包含词 T的信息数量占上述 信息集总信息数量的比例, H(Clt+)表示上述信息集中包含词 T的信息子集 的信息熵, P(t-)表示不包含词 T的信息数量占上述信息集总信息数量的比 例, H(Clt-)表示上述信息集中不包含词 T的信息子集的信息熵。
根据上述公式 (1 ) 将公式 (2 ) 展开为公式 (3 ) :
H(CIT)=P(t+)*(-(p+lt+)*log(p+lt+)-(p-lt+)*log(p-lt+))+P(t-)*(-(p+lt-)*log(p+lt -) -(p-|t-)*log(p-lt-)), (p+lt+)为范围内待推荐信息中包含词 T的信息条数占 上述信息集中包含词 τ的总信息条数的比例, 以上述 "国货" 为例
(p+lt+)=20491/125531, 同理, (p-lt+)为范围外待推荐信息中包含词 T的信 息条数占上述信息集中包含词 T的总信息条数的比例, (p+lt-)为范围内待 推荐信息中不包含词 T的信息条数占上述信息集中不包含词 T的总信息条 数的比例, (p-lt-)为范围外待推荐信息中不包含词 T的信息条数占上述信 息集中不包含词 τ的总信息条数的比例。
S204、 计算上述切分后各词语的信息增益 IG ( T) , 具体的根据公式 ( 4 ) : IG(T)=H(C) - H(CIT)计算信息增益, 根据前述公式将公式 (4 ) 展 开为公式 (5 ) :
IG(T)=P(t+)*H(Clt+)+P(t-)*H(Clt-)-(P(t+)*(-(p+lt+)*log(p+lt+)-(p-lt+)*log(p-l t+))+P(t-)*(-(p+lt-)*log(p+lt-)-(p-lt-)*log(p-lt-))) , 那么以上述 "国货"为例:
IG (国货) = -20640 I 126569 * (log(20640 I 126569) ) - 105929 I 126569 * ((log(105929 I 126569) )) - 1038 I 126569 * (-149 I 1038 * (log(149 I 1038) ) - 889 I 1038 * (log(889 I 1038) )) - 125531 I 126569 * (-20491 I 125531 * (log(20491 I 125531) ) - 105040 I 125531 * (log(105040 I 125531) ))) =
0.000017。采用该计算方式分别计算出上述切分后的各词语的信息增益值, 并根据计算出的信息增益值来选择用于推荐的待推荐信息。
进一歩地, 根据上述信息增益在上述范围内待推荐信息中确定上述用 于推荐的待推荐信息, 具体为, 根据上述范围内信息所包括的各关键词所 对应的信息增益, 获取各范围内待推荐信息对应的数字向量; 然后, 根据 各范围内待推荐信息对应的数字向量组成数字向量矩阵, 应用预设的聚类 或分类算法, 获取用于推荐的范围内待推荐信息。
举例说明,上述信息" #最喜欢的手机品牌 #当然是正在使用的华为呀! 支持国货! ",转换为"最喜欢 手机 品牌 当然是 正在 使用 华为 呀 支 持 国货" 之后, 假设切分后的 10个词的信息增益依次为
0.000001 ,0.03,0.004,0.00006,0.000008,0.000001,0.003,0.0004,0.000006,0.00 0017, 于是该条信息对应的数字向量为
{0.000001 ,0.03,0.004,0.00006,0.000008,0.000001,0.003,0.0004,0.000006,0.0 00017 }, 将范围内待推荐信息都表示为上述数字向量, 然后将这些数字向 量组成向量矩阵。 将获取到的向量矩阵输入到预设的聚类或分类算法中, 可以采用现有的聚类算法: kmeans算法、层次聚类算法等, 也可以采用现 有的分类算法: 朴素贝叶斯分类算法、 贝叶斯网络分类算法等, 以 kmeans 算法为例, 通过这种算法将每条信息放到对应的类中, 并且计算出每条信 息与类中心的距离, 最后从每一类中挑选出与类中心距离最小的信息推荐 给用户。 这样就可以挑选出包含信息量最大的这一类信息推荐给用户。
以表 2为例, 表 2为微博网站针对多条微博经过上述实施例处理的基 础上, 经过聚类算法后输出的部分结果:
表 2
Figure imgf000010_0001
1 0.216215357 /@张三 :2G的过瘾, 1G的实惠 @李四: 大家赶紧来抢 吧, 绝对不会后悔。 //@华为商城: #商城新鲜事儿 #【华 为 Mediapad 10 FHD—— 首发预售享优惠套餐! 】童 鞋们, 别说华仔不够义气 //@华为商城: 童鞋们, 更有 配备 2G RAM+16G机身内存的更高配置版本一同上 市! 详情: http:〃 t.cn/zWEz9sw
1 0.220000961 //@穆然欢喜: 这个好,把京东劫光光 [赞] //@全球 IT数 码排行榜: #和华为一起打劫京东 # MediaPad至清 至快 至真 至轻薄, 超越 NEW PAD不要犹豫抄底低价抓到 手, 与华为一起来打劫京东吧!
1 0.230278106 @哎呀呀好粉红 咱下午说的这货偷偷上市了…价位嘛
2999没.9…然后规格说明里不停的提底座键盘可能是 想暗示咱买 然后我就极度厌烦了…真送 e5就不错 但 只是有机会 让人三思 //@华为 MediaPad:所有在华为商 城、 京东商城参加预购的朋友都有机会获取华为 E5, 搭配 WiFi MediaPad 10 FHD 体验更佳!
2 0.084241 #华为 P1让智慧更美丽 #[bofu啃西瓜]转一转, 相信会 有好运被我转出来的! ! ! @也而之蓝 @Miss八月 未央 @fox芬 地址: http://t.cn/zW8kEDm
2 0.084242 #华为 PI让智慧更美丽 #[bofu啃西瓜]转一转, 相信会 有好运被我转出来的! ! ! @张三 @李四 地址: http://t.cn/zW8kEDm
2 0.084251 #华为 PI让智慧更美丽 #[bofu啃西瓜]转一转, 相信会 有好运被我转出来的! ! ! @成成 @向往天空的白 ©gunananan 地址: http:〃 t.cn/zW8kEDm
根据上述结果, 向用户推荐下述两条微博: 1 ) /@张三 :2G的过瘾, 1G的实惠 @李四: 大家赶紧来抢吧, 绝对不会后悔。 //@华为商城: #商城 新鲜事儿#【华为 Mediapad 10 FHD—— 首发预售享优惠套餐!】童鞋们, 别说华仔不够义气 //@华为商城: 童鞋们, 更有配备 2G RAM+16G机身内 存的更高配置版本一同上市! 详情: http://t.cn/zWEz9sw。 2 ) #华为 PI让 智慧更美丽 #[bofu啃西瓜]转一转, 相信会有好运被我转出来的! ! ! @ 也而之蓝 @Miss八月未央 @fox芬地址: http:〃 t.cn/zW8kEDm。
另外, 还可以利用语义分析工具, 将聚类或分类后, 每类的中心词组 织成一条有用的信息, 推荐给用户。
进一歩地, 在上述实施例的基础上, 可以根据各关键词所对应的信息 增益, 对上述范围内待推荐信息进行筛选, 并获取经过筛选的范围内待推 荐信息所对应的数字向量; 相应地, 上述根据各范围内待推荐信息对应的 数字向量组成数字向量矩阵, 具体为, 根据经过筛选的范围内待推荐信息 所对应的数字向量, 组成上述数字向量矩阵。 也就是, 在计算出各词语的 信息增益之后, 可以对词语按照信息增益值的高低进行排序, 将信息增益 小于预设阈值的词语所在的信息从范围内待推荐信息中删除, 这样可以避 免向用户推荐一些重复出现的垃圾信息、 广告等。 从上述实施例可以看出 负例中出现的信息一般是过时的信息, 对于一些重复出现的信息会既出现 在范围内待推荐信息中, 也出现在范围外待推荐信息中, 例如一则广告, 连续播放一个月, 信息推荐时间范围为当天, 那么这则广告在范围外待推 荐信息中会出现的次数会远大于这则广告在范围内待推荐信息中会出现 的次数, 根据上述公式 (5 ) 计算出这则广告中包含的词语的信息增益一 定会很低, 那么当天向用户推荐信息时就会将这则广告删除, 而不会推荐 给用户, 这样避免用户看到一些重复出现的信息以及过时的信息。
更进一歩地, 上述获取信息集, 可以为根据搜索词获取多条待推荐信 息, 形成该信息集; 其中, 该搜索词可以为: (1 ) 用户自己输入的搜索 词; 或者, (2 ) 从用户的关联信息中提取的搜索词。 这样可以实现在向 用户推荐信息之前将用户的兴趣考虑进去, 以便向用户推荐的信息是用户 感兴趣的信息。
具体实现过程中, 上述 (1 ) 方式中, 用户可以在搜索引擎中直接输 入一些搜索词, 由搜索引擎获取相关信息。 上述 (2 ) 方式中, 可以是从 用户自定义的一些信息中提取搜索词, 例如微博中用户自定义的标签信 息, 就可以直接提取出来作为搜索词; 也可以根据用户的浏览记录来提取 搜索词, 例如最近用户在电子商务网站是多次浏览了历史类书籍, 那么可 以将 "历史类书籍" 作为搜索词。
需要说明的是, 一些网站服务器, 例如微博的服务器, 不允许其它搜 索引擎对其网站进行大规模的信息搜索, 那么, 可以使微博自己的搜索工 具周期性的采用上述搜索词对其信息进行搜索, 去重以后保存在本地, 由 信息推荐处理装置通过专用的搜索接口进行获取。
本实施例中, 通过根据与用户关联的搜索词获取用户感兴趣的信息, 根据信息推荐时间范围信息以及各待推荐信息对应的时间标签, 将获取到 的待推荐信息划分为范围内待推荐信息和范围外待推荐信息, 并在范围内 待推荐信息中选择用于推荐的待推荐信息给用户, 这样给用户推荐的信息 考虑了信息的时间标签, 实现了给用户推荐的信息时效性高。 并且可以根 据各关键词的增益信息对范围内待推荐信息进行筛选, 可以去除一些重复 出现的信息以及广告信息等垃圾信息。
图 3为本发明提供的信息推荐处理装置实施例一的结构示意图, 该装 置可以集成于不同网站的服务器中, 如图 3所示, 该装置包括: 获取模块 301, 划分模块 302和推荐模块 303, 其中:
获取模块 301, 用于获取信息集, 所述信息集中包括多条待推荐信息, 所述待推荐信息包括用于标识所述待推荐信息产生时间的时间标签; 划分 模块 302, 用于根据信息推荐时间范围信息以及各待推荐信息对应的时间 标签, 将所述信息集中的多条待推荐信息划分为范围内待推荐信息和范围 外待推荐信息; 推荐模块 303, 用于在所述范围内待推荐信息内确定用于 推荐的待推荐信息; 其中, 所述范围内待推荐信息的时间标签所标识的时 间包含在所述信息推荐时间范围内。
上述各模块用于执行图 1所示方法实施例, 其实现原理和技术效果类 似, 在此不再赘述。
图 4为本发明提供的信息推荐处理装置实施例二的结构示意图, 在图
3的基础上, 推荐模块 303, 具体用于获取上述范围内待推荐信息所包括 的至少一个关键词, 并根据所述范围内待推荐信息和所述范围外待推荐信 息的数量, 以及所述范围内待推荐信息和所述范围外待推荐信息分别包括 所述关键词的数量, 获取所述关键词对应的信息增益; 根据上述信息增益 在上述范围内待推荐信息中确定用于推荐的待推荐信息。 进一歩地, 如图 4所示, 推荐模块 303, 包括获取单元 401和推荐单 元 402, 其中:
获取单元 401, 用于根据上述范围内待推荐信息所包括的各关键词所 对应的信息增益, 获取各范围内待推荐信息对应的数字向量; 推荐单元 402, 用于根据各范围内待推荐信息对应的数字向量组成数字向量矩阵, 应用预设的聚类或分类算法, 获取用于推荐的范围内待推荐信息。
图 5为本发明提供的信息推荐处理装置实施例三的结构示意图, 如图 5所示, 在图 4的基础上, 该装置还包括: 筛选模块 501, 其中, 该筛选 模块 501, 用于根据各关键词所对应的信息增益, 对所述范围内待推荐信 息进行筛选, 并获取经过筛选的范围内待推荐信息所对应的数字向量; 上 述推荐单元 402, 用于根据经过筛选的范围内待推荐信息所对应的数字向 量, 组成所述数字向量矩阵。
进一歩地, 上述获取模块 301, 具体用于根据搜索词获取多条待推荐 信息, 形成所述信息集; 其中, 所述搜索词包括: 用户输入的搜索词; 或 者, 从用户的关联信息中提取的搜索词。
上述各模块用于执行前述方法实施例, 其实现原理和技术效果类似, 在此不再赘述。
图 6为本发明提供的信息推荐处理装置实施例四的结构示意图, 如图 6所示, 该装置包括: 存储器 601和处理器 602, 其中存储器 601用于存 储指令, 处理器 602与所述存储器耦合, 所述处理器 602被配置为执行存 储在所述存储器中的指令, 具体地:
处理器 602用于获取信息集, 所述信息集中包括多条待推荐信息, 所 述待推荐信息包括用于标识所述待推荐信息产生时间的时间标签; 根据信 息推荐时间范围信息以及各待推荐信息对应的时间标签, 将所述信息集中 的多条待推荐信息划分为范围内待推荐信息和范围外待推荐信息; 在所述 范围内待推荐信息内确定用于推荐的待推荐信息; 其中, 所述范围内待推 荐信息的时间标签所标识的时间包含在所述信息推荐时间范围内。
进一歩地, 该处理器 602具体用于获取所述范围内待推荐信息所包括 的至少一个关键词, 并根据所述范围内待推荐信息和所述范围外待推荐信 息的数量, 以及所述范围内待推荐信息和所述范围外待推荐信息分别包括 所述关键词的数量, 获取所述关键词对应的信息增益; 根据所述信息增益 在所述范围内待推荐信息中确定所述用于推荐的待推荐信息。
更进一歩地, 所述处理器 602, 用于根据所述范围内待推荐信息所包 括的各关键词所对应的信息增益, 获取各范围内待推荐信息对应的数字向 量; 根据各范围内待推荐信息对应的数字向量组成数字向量矩阵, 应用预 设的聚类或分类算法, 获取用于推荐的范围内待推荐信息。
所述处理器 602, 还用于根据各关键词所对应的信息增益, 对所述范 围内待推荐信息进行筛选, 并获取经过筛选的范围内待推荐信息所对应的 数字向量; 根据经过筛选的范围内待推荐信息所对应的数字向量, 组成所 述数字向量矩阵。
另外, 所述处理器 602, 具体用于根据搜索词获取多条待推荐信息, 形成所述信息集; 其中, 所述搜索词包括: 用户输入的搜索词; 或者, 从 用户的关联信息中提取的搜索词。
上述装置可用于执行前述方法实施例, 其实现方式类似, 在此不再赘 述。
在本发明所提供的几个实施例中,应该理解到,所揭露的装置和方法, 可以通过其它的方式实现。 例如, 以上所描述的装置实施例仅仅是示意性 的, 例如, 所述单元的划分, 仅仅为一种逻辑功能划分, 实际实现时可以 有另外的划分方式, 例如多个单元或组件可以结合或者可以集成到另一个 系统, 或一些特征可以忽略, 或不执行。 另一点, 所显示或讨论的相互之 间的耦合或直接耦合或通信连接可以是通过一些接口, 装置或单元的间接 耦合或通信连接, 可以是电性, 机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的, 作为单元显示的部件可以是或者也可以不是物理单元, 即可以位于一个地 方, 或者也可以分布到多个网络单元上。 可以根据实际的需要选择其中的 部分或者全部单元来实现本实施例方案的目的。
另外, 在本发明各个实施例中的各功能单元可以集成在一个处理单元 中, 也可以是各个单元单独物理存在, 也可以两个或两个以上单元集成在 一个单元中。 上述集成的单元既可以采用硬件的形式实现, 也可以采用硬 件加软件功能单元的形式实现。 上述以软件功能单元的形式实现的集成的单元, 可以存储在一个计算 机可读取存储介质中。 上述软件功能单元存储在一个存储介质中, 包括若 干指令用以使得一台计算机设备 (可以是个人计算机, 服务器, 或者网络 设备等) 或处理器 (processor) 执行本发明各个实施例所述方法的部分歩 骤。 而前述的存储介质包括: U盘、 移动硬盘、 只读存储器 (Read-Only Memory, ROM ) 、 随机存取存储器 (Random Access Memory, RAM ) 、 磁碟或者光盘等各种可以存储程序代码的介质。
最后应说明的是: 以上各实施例仅用以说明本发明的技术方案, 而非 对其限制; 尽管参照前述各实施例对本发明进行了详细的说明, 本领域的 普通技术人员应当理解: 其依然可以对前述各实施例所记载的技术方案进 行修改, 或者对其中部分或者全部技术特征进行等同替换; 而这些修改或 者替换, 并不使相应技术方案的本质脱离本发明各实施例技术方案的范 围。

Claims

权 利 要 求 书
1、 一种信息推荐处理方法, 其特征在于, 包括:
获取信息集, 所述信息集中包括多条待推荐信息, 所述待推荐信息包 括用于标识所述待推荐信息产生时间的时间标签;
根据信息推荐时间范围信息以及各待推荐信息对应的时间标签, 将所 述信息集中的多条待推荐信息划分为范围内待推荐信息和范围外待推荐
I Ή自、 .,
在所述范围内待推荐信息内确定用于推荐的待推荐信息;
其中, 所述范围内待推荐信息的时间标签所标识的时间包含在所述信 息推荐时间范围内。
2、 根据权利要求 1所述的方法, 其特征在于, 所述在所述范围内待 推荐信息内确定用于推荐的待推荐信息包括:
获取所述范围内待推荐信息所包括的至少一个关键词, 并根据所述范 围内待推荐信息和所述范围外待推荐信息的数量, 以及所述范围内待推荐 信息和所述范围外待推荐信息分别包括所述关键词的数量, 获取所述关键 词对应的信息增益;
根据所述信息增益在所述范围内待推荐信息确定所述用于推荐的待 推荐信息。
3、 根据权利要求 2所述的方法, 其特征在于, 所述根据所述信息增 益在所述范围内待推荐信息中确定所述用于推荐的待推荐信息包括: 根据所述范围内待推荐信息所包括的各关键词所对应的信息增益, 获 取各范围内待推荐信息对应的数字向量;
根据各范围内待推荐信息对应的数字向量组成数字向量矩阵, 应用预 设的聚类或分类算法, 获取用于推荐的范围内待推荐信息。
4、 根据权利要求 3所述的方法, 其特征在于, 所述方法还包括: 根据各关键词所对应的信息增益, 对所述范围内待推荐信息进行筛 选, 并获取经过筛选的范围内待推荐信息所对应的数字向量; 相应地, 所述根据各范围内待推荐信息对应的数字向量组成数字向量矩阵包 括:
根据经过筛选的范围内待推荐信息所对应的数字向量, 组成所述数字 向量矩阵。
5、 根据权利要求 1-4任一项所述的方法, 其特征在于, 所述获取信息 集包括:
根据搜索词获取多条待推荐信息, 形成所述信息集;
其中, 所述搜索词包括: 用户输入的搜索词; 或者, 从用户的关联信 息中提取的搜索词。
6、 一种信息推荐处理装置, 其特征在于, 包括:
获取模块, 用于获取信息集, 所述信息集中包括多条待推荐信息, 所 述待推荐信息包括用于标识所述待推荐信息产生时间的时间标签;
划分模块, 用于根据信息推荐时间范围信息以及各待推荐信息对应的 时间标签, 将所述信息集中的多条待推荐信息划分为范围内待推荐信息和 范围外待推荐信息;
推荐模块, 用于在所述范围内待推荐信息内确定用于推荐的待推荐信 息;
其中, 所述范围内待推荐信息的时间标签所标识的时间包含在所述信 息推荐时间范围内。
7、 根据权利要求 6所述的装置, 其特征在于, 所述推荐模块, 具体 用于获取所述范围内待推荐信息所包括的至少一个关键词, 并根据所述范 围内待推荐信息和所述范围外待推荐信息的数量, 以及所述范围内待推荐 信息和所述范围外待推荐信息分别包括所述关键词的数量, 获取所述关键 词对应的信息增益; 根据所述信息增益在所述范围内待推荐信息中确定所 述用于推荐的待推荐信息。
8、 根据权利要求 7所述的装置, 其特征在于, 所述推荐模块, 包括: 获取单元, 用于根据所述范围内待推荐信息所包括的各关键词所对应 的信息增益, 获取各范围内待推荐信息对应的数字向量;
推荐单元, 用于根据各范围内待推荐信息对应的数字向量组成数字向 量矩阵,应用预设的聚类或分类算法,获取用于推荐的范围内待推荐信息。
9、 根据权利要求 8所述的装置, 其特征在于, 还包括:
筛选模块, 用于根据各关键词所对应的信息增益, 对所述范围内待推 荐信息进行筛选, 并获取经过筛选的范围内待推荐信息所对应的数字向 所述推荐单元, 用于根据经过筛选的范围内待推荐信息所对应的数字 向量, 组成所述数字向量矩阵。
10、 根据权利要求 6-9任一项所述的装置, 其特征在于, 所述获取模 块, 具体用于根据搜索词获取多条待推荐信息, 形成所述信息集; 其中, 所述搜索词包括: 用户输入的搜索词; 或者, 从用户的关联信息中提取的 搜索词。
11、 一种信息推荐处理装置, 其特征在于, 包括:
存储器和处理器, 其中:
所述存储器, 用于存储指令;
所述处理器, 与所述存储器耦合, 被配置为执行存储在所述存储器中 的指令, 用于获取信息集, 所述信息集中包括多条待推荐信息, 所述待推 荐信息包括用于标识所述待推荐信息产生时间的时间标签; 根据信息推荐 时间范围信息以及各待推荐信息对应的时间标签, 将所述信息集中的多条 待推荐信息划分为范围内待推荐信息和范围外待推荐信息; 在所述范围内 待推荐信息内确定用于推荐的待推荐信息; 其中, 所述范围内待推荐信息 的时间标签所标识的时间包含在所述信息推荐时间范围内。
12、 根据权利要求 11 所述的装置, 其特征在于, 所述处理器, 具体 用于获取所述范围内待推荐信息所包括的至少一个关键词, 并根据所述范 围内待推荐信息和所述范围外待推荐信息的数量, 以及所述范围内待推荐 信息和所述范围外待推荐信息分别包括所述关键词的数量, 获取所述关键 词对应的信息增益; 根据所述信息增益在所述范围内待推荐信息中确定所 述用于推荐的待推荐信息。
13、 根据权利要求 12所述的装置, 其特征在于, 所述处理器, 具体 用于根据所述范围内待推荐信息所包括的各关键词所对应的信息增益, 获 取各范围内待推荐信息对应的数字向量; 根据各范围内待推荐信息对应的 数字向量组成数字向量矩阵, 应用预设的聚类或分类算法, 获取用于推荐 的范围内待推荐信息。
14、 根据权利要求 13所述的装置, 其特征在于, 所述处理器, 还用 于根据各关键词所对应的信息增益, 对所述范围内待推荐信息进行筛选, 并获取经过筛选的范围内待推荐信息所对应的数字向量; 根据经过筛选的 范围内待推荐信息所对应的数字向量, 组成所述数字向量矩阵。
15、 根据权利要求 11-14任一项所述的装置, 其特征在于, 所述处理 器, 具体用于根据搜索词获取多条待推荐信息, 形成所述信息集; 其中, 所述搜索词包括: 用户输入的搜索词; 或者, 从用户的关联信息中提取的 搜索词。
PCT/CN2014/074403 2013-05-08 2014-03-31 信息推荐处理方法及装置 WO2014180196A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US14/795,189 US20150324448A1 (en) 2013-05-08 2015-07-09 Information Recommendation Processing Method and Apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201310165715.8A CN104142940B (zh) 2013-05-08 2013-05-08 信息推荐处理方法及装置
CN201310165715.8 2013-05-08

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US14/795,189 Continuation US20150324448A1 (en) 2013-05-08 2015-07-09 Information Recommendation Processing Method and Apparatus

Publications (1)

Publication Number Publication Date
WO2014180196A1 true WO2014180196A1 (zh) 2014-11-13

Family

ID=51852114

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/074403 WO2014180196A1 (zh) 2013-05-08 2014-03-31 信息推荐处理方法及装置

Country Status (3)

Country Link
US (1) US20150324448A1 (zh)
CN (1) CN104142940B (zh)
WO (1) WO2014180196A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111782980A (zh) * 2020-06-30 2020-10-16 北京百度网讯科技有限公司 地图兴趣点的挖掘方法、装置、设备及存储介质

Families Citing this family (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104657468B (zh) * 2015-02-12 2018-07-31 中国科学院自动化研究所 基于图像与文本的视频的快速分类方法
CN104915439A (zh) * 2015-06-25 2015-09-16 百度在线网络技术(北京)有限公司 搜索结果推送方法和装置
WO2017044958A1 (en) * 2015-09-11 2017-03-16 Ayasdi, Inc. Network representation for evolution of clusters and groups
CN106598987B (zh) * 2015-10-16 2020-08-07 阿里巴巴集团控股有限公司 信息推荐方法及装置
CN105608117B (zh) * 2015-12-14 2019-12-10 微梦创科网络科技(中国)有限公司 一种信息推荐方法及装置
CN105630868B (zh) * 2015-12-15 2019-05-31 北京奇虎科技有限公司 一种向用户推荐内容的方法及系统
CN105608627A (zh) * 2016-02-01 2016-05-25 广东欧珀移动通信有限公司 基于社交网络平台的信息更新方法及装置
CN105740436B (zh) * 2016-02-01 2019-12-27 北京京东尚科信息技术有限公司 基于互联网搜索技术的文字作品推送方法和装置
CN105608154B (zh) * 2016-02-14 2019-10-15 广州网律互联网科技有限公司 一种基于隐马尔科夫链模型的智能推荐算法
CN106454536B (zh) * 2016-09-19 2019-07-26 广州视源电子科技股份有限公司 信息推荐度的确定方法及装置
US11205103B2 (en) 2016-12-09 2021-12-21 The Research Foundation for the State University Semisupervised autoencoder for sentiment analysis
CN106934002B (zh) * 2017-03-06 2020-07-07 冠生园(集团)有限公司 一种搜索关键字数字化解析方法和引擎
CN107463679A (zh) * 2017-08-07 2017-12-12 石林星 一种信息推荐方法及装置
CN107463698B (zh) 2017-08-15 2020-11-20 北京百度网讯科技有限公司 基于人工智能推送信息的方法和装置
CN107657004A (zh) * 2017-09-21 2018-02-02 广州华多网络科技有限公司 视频推荐方法、系统及设备
CN107943907A (zh) * 2017-11-17 2018-04-20 南京感度信息技术有限责任公司 一种基于内容标签的知识库推荐系统
CN108268619B (zh) * 2018-01-08 2020-06-30 阿里巴巴集团控股有限公司 内容推荐方法及装置
CN109543111B (zh) * 2018-11-28 2021-09-21 广州虎牙信息科技有限公司 推荐信息筛选方法、装置、存储介质及服务器
CN110727840B (zh) * 2019-09-27 2022-07-05 浙江大搜车软件技术有限公司 车辆问询标签推送方法、装置、计算机设备和存储介质
CN113886708A (zh) * 2021-10-26 2022-01-04 平安银行股份有限公司 基于用户信息的产品推荐方法、装置、设备及存储介质
CN114169976A (zh) * 2021-12-27 2022-03-11 中国建设银行股份有限公司 金融数据的推荐方法、装置和设备
CN114661705A (zh) * 2022-04-15 2022-06-24 广州双知网络科技有限公司 一种基于云计算的大数据分析系统
JP7448595B2 (ja) 2022-07-20 2024-03-12 楽天グループ株式会社 情報処理システム、情報処理方法及びプログラム

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102662986A (zh) * 2012-01-13 2012-09-12 中国科学院计算技术研究所 微博消息检索系统与方法
US20130031093A1 (en) * 2011-07-25 2013-01-31 Sony Computer Entertainment Inc. Information processing system, information processing method, program, and non-transitory information storage medium

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008083504A1 (en) * 2007-01-10 2008-07-17 Nick Koudas Method and system for information discovery and text analysis
JP2008178037A (ja) * 2007-01-22 2008-07-31 Sony Corp 情報処理装置、情報処理方法及び情報処理プログラム
CN101609472B (zh) * 2009-08-13 2011-08-17 腾讯科技(深圳)有限公司 一种基于问答平台的关键词评价方法和装置
CN102346894B (zh) * 2010-08-03 2017-03-01 阿里巴巴集团控股有限公司 推荐信息的输出方法、系统及服务器

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130031093A1 (en) * 2011-07-25 2013-01-31 Sony Computer Entertainment Inc. Information processing system, information processing method, program, and non-transitory information storage medium
CN102662986A (zh) * 2012-01-13 2012-09-12 中国科学院计算技术研究所 微博消息检索系统与方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111782980A (zh) * 2020-06-30 2020-10-16 北京百度网讯科技有限公司 地图兴趣点的挖掘方法、装置、设备及存储介质
CN111782980B (zh) * 2020-06-30 2023-08-04 北京百度网讯科技有限公司 地图兴趣点的挖掘方法、装置、设备及存储介质

Also Published As

Publication number Publication date
US20150324448A1 (en) 2015-11-12
CN104142940A (zh) 2014-11-12
CN104142940B (zh) 2017-11-17

Similar Documents

Publication Publication Date Title
WO2014180196A1 (zh) 信息推荐处理方法及装置
US20210144233A1 (en) Method and apparatus for real-time personalization
US10685065B2 (en) Method and system for recommending content to a user
JP6621914B2 (ja) オンライン・ソーシャル・ネットワーク上でのキー・トピックの検出
JP6483092B2 (ja) 更新レイヤを伴うデータベース・シャーディング
US9442903B2 (en) Generating preview data for online content
KR102373796B1 (ko) 소셜 네트워킹 사용자의 확장된 추적 및 광고 타겟팅
US11042590B2 (en) Methods, systems and techniques for personalized search query suggestions
US20160371589A1 (en) Systems and methods for online content recommendation
EP3117339A1 (en) Systems and methods for keyword suggestion
US11899728B2 (en) Methods, systems and techniques for ranking personalized and generic search query suggestions
TW201905736A (zh) 資訊推送方法及系統
JP2014137757A (ja) 情報処理システム、情報処理装置、情報処理方法及びプログラム
CN107491465B (zh) 用于搜索内容的方法和装置以及数据处理系统
US20150287080A1 (en) Contextual promotions
US11256453B1 (en) Retargeting events service for online advertising
US20160012130A1 (en) Aiding composition of themed articles about popular and novel topics and offering users a navigable experience of associated content
Nguyen et al. Learning to rank for personalised fashion recommender systems via implicit feedback
US20170097986A1 (en) Methods, systems and techniques for providing search query suggestions based on non-personal data and user personal data according to availability of user personal data
Pradhan et al. Recommendation system using lexicon based sentimental analysis with collaborative filtering
US9367624B2 (en) Identity workflow that utilizes multiple storage engines to support various lifecycles
WO2017062245A1 (en) Method and system for dynamically generating a card
CN111787042B (zh) 用于推送信息的方法和装置
US20140278983A1 (en) Using entity repository to enhance advertisement display
CN113704630B (zh) 一种信息推送方法、装置、可读存储介质和电子设备

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: 14794050

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: 14794050

Country of ref document: EP

Kind code of ref document: A1