CN113254789A - Method and device for pushing meteorological service content - Google Patents

Method and device for pushing meteorological service content Download PDF

Info

Publication number
CN113254789A
CN113254789A CN202110731097.3A CN202110731097A CN113254789A CN 113254789 A CN113254789 A CN 113254789A CN 202110731097 A CN202110731097 A CN 202110731097A CN 113254789 A CN113254789 A CN 113254789A
Authority
CN
China
Prior art keywords
content
vector
positive feedback
candidate
meteorological
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110731097.3A
Other languages
Chinese (zh)
Other versions
CN113254789B (en
Inventor
匡秋明
郑江平
刘进
彭敏
王一鹤
张耿
于廷照
胡骏楠
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.)
Public Meteorological Service Center Of China Meteorological Administration National Early Warning Information Release Center
Original Assignee
Public Meteorological Service Center Of China Meteorological Administration National Early Warning Information Release Center
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 Public Meteorological Service Center Of China Meteorological Administration National Early Warning Information Release Center filed Critical Public Meteorological Service Center Of China Meteorological Administration National Early Warning Information Release Center
Priority to CN202110731097.3A priority Critical patent/CN113254789B/en
Publication of CN113254789A publication Critical patent/CN113254789A/en
Application granted granted Critical
Publication of CN113254789B publication Critical patent/CN113254789B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Animal Behavior & Ethology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention provides a method and a device for pushing meteorological service content, wherein the method comprises the following steps: acquiring a positive feedback content set of a target user having positive feedback on browsed meteorological service contents; based on the positive feedback content set and the content similarity matrix, calculating the interest degree of the target user in the weather service content, and generating a candidate weather service content push set; respectively extracting words from a positive feedback content set of a target user and a candidate meteorological service content push set, constructing a word vector, an entity vector and an entity context vector based on the extracted words and a meteorological knowledge dictionary, and splicing to obtain a vector splicing matrix; inputting the vector splicing matrix corresponding to the content into a graph convolution neural network to obtain a content vector; based on each content vector, acquiring a weight score of the positive feedback meteorological service content and acquiring a push score of the candidate meteorological service content; and acquiring a plurality of candidate meteorological service contents ranked at the top for pushing. The pushing accuracy can be improved.

Description

Method and device for pushing meteorological service content
Technical Field
The invention relates to the technical field of weather service, in particular to a method and a device for pushing weather service content.
Background
The appearance and popularization of the internet brings a great amount of information to users, and the requirements of the users on information sharing, information acquisition and information inquiry are met. However, with the rapid development of the internet, the amount of information on the internet is greatly increased, so that when a user faces massive information, the user is difficult to obtain information really useful for the user, and the use efficiency of the information is reduced. For example, when people travel outside, people need to know related geological weather service contents, but the existing method for acquiring the weather service contents needs to search on the internet by users, and for mass contents obtained by searching, the users need to select useful contents from the mass contents for browsing, so that the time spent on browsing is long, the browsing efficiency of the users is low, and the pushing accuracy is low; however, most weather service contents pushed to the user on holidays are weather forecast contents related to famous scenic spots, so that personalized requirements of the user are difficult to meet, and the pushed weather service contents are more and neglected, so that the pushing efficiency is not high.
Disclosure of Invention
In view of the above, the present invention provides a method and an apparatus for pushing weather service content to improve the accuracy of pushing and meet the personalized requirements of users.
In a first aspect, an embodiment of the present invention provides a method for pushing weather service content, including:
acquiring a positive feedback content set of a target user having positive feedback on browsed meteorological service contents;
calculating the interest degree of the target user to the weather service content in the content similarity matrix based on the positive feedback content set and the content similarity matrix constructed in advance, and generating a candidate weather service content push set based on the calculated interest degree;
respectively extracting words from a positive feedback content set and a candidate meteorological service content push set of a target user, constructing word vectors based on the extracted words and a preset meteorological knowledge dictionary, and constructing entity vectors and entity context vectors corresponding to the extracted words based on the extracted words and a preset geological disaster meteorological knowledge map of torrential flood;
mapping the entity vector and the entity context vector to a space of the word vector, and splicing the word vector, the mapped entity vector and the entity context vector to obtain a vector splicing matrix;
aiming at each content, inputting a vector splicing matrix corresponding to the content into a graph convolution neural network to obtain a content vector;
acquiring a weight score of each positive feedback meteorological service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate meteorological service content push set, and acquiring a push score of each candidate meteorological service content in the candidate meteorological service content push set based on the content vector of the positive feedback content set, the content vector of the candidate meteorological service content push set and the weight score;
and sequencing according to the pushing scores, and acquiring a plurality of candidate meteorological service contents which are sequenced at the front for pushing.
With reference to the first aspect, an embodiment of the present invention provides a first possible implementation manner of the first aspect, where the acquiring a positive feedback content set that a target user has positive feedback on browsed weather service content includes:
acquiring meteorological service content browsed by a target user;
if the weather service content is matched with the interested weather service classification preset by the target user, placing the weather service content in a positive feedback content set;
if the weather service content is not matched with the preset weather service classification of the target user, if the score of the weather service content exceeds a preset score threshold value, placing the weather service content in a positive feedback content set;
if the weather service content is not scored, if the browsing time of the weather service content exceeds a preset time threshold, the weather service content is placed in a positive feedback content set.
With reference to the first aspect, an embodiment of the present invention provides a second possible implementation manner of the first aspect, where constructing the content similarity matrix includes:
acquiring a positive feedback meteorological service content set with positive feedback on browsed meteorological service contents by a user;
and generating a content similarity matrix by utilizing a collaborative filtering recommendation algorithm based on each positive feedback meteorological service content set.
With reference to the second possible implementation manner of the first aspect, an embodiment of the present invention provides a third possible implementation manner of the first aspect, where the generating a content similarity matrix by using a collaborative filtering recommendation algorithm based on each positive feedback weather service content set includes:
merging the positive feedback meteorological service content sets of each user and performing content deduplication processing to obtain a content deduplication set, constructing a positive feedback content matrix based on the content deduplication set, wherein rows and columns of the positive feedback content matrix are respectively contents contained in the content deduplication set, and the number of rows and columns is equal to the number of content strips contained in the content deduplication set;
editing row and column values of a positive feedback content matrix according to the positive feedback meteorological service content set of each user;
and normalizing the row and column values of the edited positive feedback content matrix to obtain a content similarity matrix.
With reference to the first aspect and any one of the first to third possible implementation manners of the first aspect, an embodiment of the present invention provides a fourth possible implementation manner of the first aspect, wherein the calculating, based on a positive feedback content set and a pre-constructed content similarity matrix, a degree of interest of a target user in weather service content in the content similarity matrix includes:
aiming at target weather service contents in the content similarity matrix, acquiring a similar content set similar to the target weather service contents;
acquiring a union of a similar content set and a positive feedback content set to obtain a content union;
aiming at each union content in the content union, acquiring union similarity between the union content and target weather service content from a content similarity matrix, and acquiring union score of the union content by a target user;
calculating the product of the union set similarity and the union set score to obtain the union set interest degree of the union set content;
and summing the interest degree of the union of the contents of each union to obtain the interest degree of the target user on the target weather service contents.
With reference to the first aspect and any one of the first to third possible implementation manners of the first aspect, an embodiment of the present invention provides a fifth possible implementation manner of the first aspect, where the constructing a word vector based on the extracted words and a preset meteorological knowledge dictionary includes:
and aiming at each extracted word, setting the word vector value of the extracted word to be 1 and setting the word vector values of other words to be 0 according to a preset meteorological knowledge dictionary to obtain the word vector of the extracted word.
With reference to the first aspect and any one of the first to third possible implementation manners of the first aspect, an embodiment of the present invention provides a sixth possible implementation manner of the first aspect, where the constructing an entity vector and an entity context vector corresponding to the extracted word based on the extracted word and a preset geological flood disaster weather knowledge graph includes:
judging whether the extracted words are entities in the mountain torrent geological disaster meteorological knowledgegraph or not aiming at each extracted word, and if so, constructing a sub-graph of a triple including a first entity, a relation and a second entity according to an initial entity corresponding to the extracted word in the mountain torrent geological disaster meteorological knowledgegraph and other entities in the mountain torrent geological disaster meteorological knowledgegraph and related to the initial entity, wherein the first entity is the initial entity;
setting a first entity vector, a relation vector and a second entity vector for each triple in the sub-spectrum, and assigning the first entity vector, the relation vector and the second entity vector by using a translation embedding algorithm until a difference value between a sum of the first entity vector value and the relation vector value and a second entity vector value is within a preset vector value error threshold value to obtain a sub-spectrum vector set;
extracting a vector corresponding to the initial entity from the sub-graph spectrum vector set to obtain an entity vector;
and calculating the average value of all entity vectors connected with the initial entity in the sub-map vector set to obtain the entity context vector.
With reference to the first aspect and any one of the first to third possible implementation manners of the first aspect, an embodiment of the present invention provides a seventh possible implementation manner of the first aspect, where the obtaining a weight score of each positive feedback weather service content in the positive feedback content set based on a content vector of the positive feedback content set and a content vector of a candidate weather service content push set includes:
extracting a content vector of a positive feedback meteorological service content in the positive feedback content set, and combining the content vector with the content vector of each candidate meteorological service content in the candidate meteorological service content push set respectively;
inputting the combination vector into a preset attention network aiming at the combination vector of each combination, and normalizing the output of the attention network by using a normalization function to obtain a weight value of the combination vector;
and carrying out mean value calculation on the weight values of the combined vectors to obtain the weight value corresponding to the extracted positive feedback meteorological service content.
With reference to the seventh possible implementation manner of the first aspect, an embodiment of the present invention provides an eighth possible implementation manner of the first aspect, where the obtaining a push score of each candidate weather service content in the candidate weather service content push set based on a content vector of the positive feedback content set, a content vector of the candidate weather service content push set, and a weight score includes:
aiming at each candidate meteorological service content of the candidate meteorological service content push set, respectively combining the content vector of the candidate meteorological service content with the content vector of each positive feedback meteorological service content of the positive feedback content set to obtain a candidate combined vector;
for each candidate combination vector, inputting the candidate combination vector into a preset attention network, and normalizing the output of the attention network by using a normalization function to obtain the score of the candidate combination vector;
acquiring a weight value of a combined vector corresponding to the positive feedback meteorological service content in the candidate combined vector, and calculating the product of the value of the candidate combined vector and the acquired weight value to obtain the weight value of the candidate combined vector;
and carrying out weighted average on the weighted value of each candidate combination vector to obtain the pushing value of the candidate meteorological service content.
In a second aspect, an embodiment of the present invention further provides an apparatus for pushing weather service content, including:
the content acquisition module is used for acquiring a positive feedback content set of the target user with positive feedback on the browsed meteorological service content;
the candidate content generation module is used for calculating the interest degree of the target user in the meteorological service content in the content similarity matrix based on the positive feedback content set and the content similarity matrix which is constructed in advance, and generating a candidate meteorological service content push set based on the calculated interest degree;
the vector construction module is used for respectively extracting words from a positive feedback content set and a candidate meteorological service content push set of a target user, constructing word vectors based on the extracted words and a preset meteorological knowledge dictionary, and constructing entity vectors and entity context vectors corresponding to the extracted words based on the extracted words and a preset geological flood disaster meteorological knowledge map;
the vector splicing module is used for mapping the entity vector and the entity context vector into a space of the word vector, and splicing the word vector, the mapped entity vector and the entity context vector to obtain a vector splicing matrix;
the content vector acquisition module is used for inputting a vector splicing matrix corresponding to each content into the graph convolution neural network to obtain a content vector;
the push score calculation module is used for acquiring the weight score of each positive feedback meteorological service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate meteorological service content push set, and acquiring the push score of each candidate meteorological service content in the candidate meteorological service content push set based on the content vector of the positive feedback content set, the content vector of the candidate meteorological service content push set and the weight score;
and the content pushing module is used for sequencing according to the pushing scores, and acquiring a plurality of candidate meteorological service contents which are sequenced at the front for pushing.
In a third aspect, an embodiment of the present application provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, performs the steps of the method described above.
According to the method and the device for pushing the meteorological service content, provided by the embodiment of the invention, a positive feedback content set with positive feedback on the browsed meteorological service content by a target user is obtained; calculating the interest degree of the target user to the weather service content in the content similarity matrix based on the positive feedback content set and the content similarity matrix constructed in advance, and generating a candidate weather service content push set based on the calculated interest degree; respectively extracting words from a positive feedback content set and a candidate meteorological service content push set of a target user, constructing word vectors based on the extracted words and a preset meteorological knowledge dictionary, and constructing entity vectors and entity context vectors corresponding to the extracted words based on the extracted words and a preset geological disaster meteorological knowledge map of torrential flood; mapping the entity vector and the entity context vector to a space of the word vector, and splicing the word vector, the mapped entity vector and the entity context vector to obtain a vector splicing matrix; aiming at each content, inputting a vector splicing matrix corresponding to the content into a graph convolution neural network to obtain a content vector; acquiring a weight score of each positive feedback meteorological service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate meteorological service content push set, and acquiring a push score of each candidate meteorological service content in the candidate meteorological service content push set based on the content vector of the positive feedback content set, the content vector of the candidate meteorological service content push set and the weight score; and sequencing according to the pushing scores, and acquiring a plurality of candidate meteorological service contents which are sequenced at the front for pushing. Therefore, by acquiring the weather service content browsed by the user, combining the collaborative filtering recommendation algorithm and feature learning based on the mountain torrent geological disaster weather knowledge map, the pushed weather service content is obtained, the focusing performance of the pushed weather service content can be improved, and the pushing accuracy and the pushing efficiency are improved.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flow chart of a method for pushing weather service content according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating a content similarity matrix generated by a collaborative filtering recommendation algorithm based on each positive feedback weather service content set according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating an apparatus for pushing weather service content according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a computer device 400 according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
In the existing method for obtaining the weather service content through online searching, as a user has less knowledge about the weather service content, under the condition of less input weather service keywords, massive weather service content can be obtained through searching, the time spent by the user for selecting useful weather service content is long, and the browsing efficiency is low; for the weather service content pushed to the user in holidays at present, the personalized requirements of the user are difficult to meet, and the pushed weather service content is often ignored, so that the pushing efficiency is not high. In the embodiment of the invention, the meteorological service content pushing method based on the mountain torrent geological disaster meteorological knowledgegraph is provided by acquiring the meteorological service content browsed by the user and combining the collaborative filtering recommendation algorithm and the characteristic learning based on the mountain torrent geological disaster meteorological knowledgegraph, so that the meteorological service content is pushed, the focusing performance of the pushed meteorological service content is improved, and the pushing accuracy and the pushing efficiency are improved.
The embodiment of the invention provides a method and a device for pushing meteorological service content, which are described by embodiments below.
Fig. 1 is a flow chart illustrating a method for pushing weather service content according to an embodiment of the present invention. As shown in fig. 1, the method includes:
step 101, acquiring a positive feedback content set of a target user having positive feedback on browsed meteorological service contents;
in the embodiment of the present invention, as an optional embodiment, acquiring a positive feedback content set in which a target user has positive feedback on browsed weather service content includes:
acquiring meteorological service content browsed by a target user;
if the weather service content is matched with the interested weather service classification preset by the target user, placing the weather service content in a positive feedback content set;
if the weather service content is not matched with the preset weather service classification of the target user, if the score of the weather service content exceeds a preset score threshold value, placing the weather service content in a positive feedback content set;
if the weather service content is not scored, if the browsing time of the weather service content exceeds a preset time threshold, the weather service content is placed in a positive feedback content set.
In the embodiment of the invention, the score, the browsing duration and the like of the weather service contents browsed by the user are obtained by collecting and recording the historical behaviors of the weather service contents browsed by the user, and the positive feedback content selection is carried out on the weather service contents browsed by the user by combining the interested weather service classification set by the user. As an alternative embodiment, the weather service classification of interest includes, but is not limited to: landslide, debris flow, collapse, ground collapse and the like, and the scoring information is the scoring of the browsed weather service contents by the user.
In the embodiment of the invention, for the convenience of distinguishing, the weather service content with concentrated positive feedback content is called as positive feedback weather service content.
In the embodiment of the present invention, if a positive feedback content set that the target user has positive feedback on the browsed weather service content is not acquired, as another optional embodiment, the method further includes:
the method comprises the steps of collecting attribute information and/or user explicit preference information of the preference information of a target user, and obtaining push contents based on the user explicit preference information.
In this embodiment of the present invention, as an optional embodiment, the attribute information includes but is not limited to: user geographical location information, family identity information, and the like; the preference information is the interested weather service content classification information set by the target user. The family identity information may be a child, a parent, and the like. For example, for a target user whose family identity information is a child, the weather service content which is simple and easy to understand and popular and vivid is selected for pushing in consideration of the reading ability of the child.
102, calculating the interest degree of a target user in the meteorological service content in a content similarity matrix based on a positive feedback content set and a pre-constructed content similarity matrix, and generating a candidate meteorological service content push set based on the calculated interest degree;
in the embodiment of the present invention, constructing a content similarity matrix includes:
acquiring a positive feedback meteorological service content set with positive feedback on browsed meteorological service contents by a user;
and generating a content similarity matrix by utilizing a collaborative filtering recommendation algorithm based on each positive feedback meteorological service content set.
In the embodiment of the invention, each user corresponds to a positive feedback meteorological service content set, and a content similarity matrix is constructed by collecting the positive feedback meteorological service content sets of massive users.
In the embodiment of the invention, a content similarity matrix is generated by adopting a collaborative filtering recommendation algorithm based on contents (articles), and the collaborative filtering recommendation algorithm considers that if most users like the content B, the content A and the content B have high content similarity. Therefore, as an alternative embodiment, the generating the content similarity matrix by using the collaborative filtering recommendation algorithm based on each positive feedback weather service content set includes:
merging the positive feedback meteorological service content sets of each user and performing content deduplication processing to obtain a content deduplication set, constructing a positive feedback content matrix based on the content deduplication set, wherein rows and columns of the positive feedback content matrix are respectively contents contained in the content deduplication set, and the number of rows and columns is equal to the number of content strips contained in the content deduplication set;
editing row and column values of a positive feedback content matrix according to the positive feedback meteorological service content set of each user;
and normalizing the row and column values of the edited positive feedback content matrix to obtain a content similarity matrix.
Fig. 2 is a schematic diagram illustrating a content similarity matrix generated by using a collaborative filtering recommendation algorithm based on each positive feedback weather service content set according to an embodiment of the present invention. As shown in fig. 2, in the embodiment of the present invention, assuming that the number of users is 5, the first positive feedback weather service content set of the first user includes: content a, content b, content d; the second positive feedback weather service content set for the second user comprises: content b, content c, content e; the third positive feedback weather service content set for the third user comprises: content c, content d; the fourth positive feedback weather service content set for the fourth user comprises: content b, content c, content d; the fifth positive feedback weather service content set for the fifth user includes: content a, content d; then the content deduplication set comprises: and content a, content b, content c, content d and content e, pairwise matching the content to construct a positive feedback content matrix, wherein the constructed positive feedback content matrix is a 5x5 matrix, and the positive feedback content matrix is assigned according to the positive feedback meteorological service content sets of the users respectively, namely each positive feedback content set corresponds to a content sub-matrix, and each content sub-matrix is a 5x5 matrix. For example, for a first user, in the first content sub-matrix, the row and column values of the first row and the second column, the first row and the fourth column, the second row and the first column, the second row and the fourth column, the fourth row and the first column, and the fourth row and the second column are all 1, and the values of the other rows and columns are 0; for a second user, in the second content sub-matrix, row and column values of a second row, a third column, a second row, a fifth column, a third row, a second column, a third row, a fifth column, a fifth row, a second column and a fifth column are all 1, and values of other rows and columns are 0; for a third user, in the third content sub-matrix, the row values of the third row, the fourth column, the fourth row and the third column are all 1, and the values of other rows and columns are 0; for a fourth user, in the fourth content sub-matrix, row and column values of a second row, a third column, a second row, a fourth column, a third row, a second column, a third row, a fourth column, a fourth row, a second column, a fourth row, and a third column are all 1, and values of other rows and columns are 0; for the fifth user, in the fifth content sub-matrix, the row and column values of the fourth row and column, and the first column are all 1, and the other row and column values are 0. And summing the first content sub-matrix, the second content sub-matrix, the third content sub-matrix, the fourth content sub-matrix and the fifth content sub-matrix to obtain a positive feedback content matrix, wherein in the positive feedback content matrix, a row and a column C [ i ] [ j ] represent the number of users who like the content i and the content j at the same time. And carrying out normalization processing on the positive feedback content matrix to obtain a content similarity matrix.
In the embodiment of the present invention, as an optional embodiment, the normalization processing is performed on the positive feedback content matrix by using the following formula:
Figure F_210519171612060_060232001
in the formula (I), the compound is shown in the specification,
W ij as contentiAnd contentjThe similarity between them, i.e. in the content similarity matrix, the firstiGo to the firstjNormalized values of the columns;
N(i)for positive feedback of contents in the content matrixiThe corresponding number of users;
N(j)for positive feedback of contents in the content matrixjThe corresponding number of users.
In the embodiment of the present invention, the first and second substrates,
Figure F_210519171612122_122732002
as contentiAnd contentjThe corresponding number of users, i.e. in the positive feedback content matrix, the secondiGo to the firstjThe number of users of the column.
In the embodiment of the present invention, as another optional embodiment, the positive feedback content matrix is normalized by using the following formula:
Figure F_210519171612200_200857003
in the embodiment of the invention, the possibility that hot contents are similar to a plurality of contents can be reduced by using the formula.
In the embodiment of the present invention, as another optional embodiment, in the following, the content similarity matrix may be updated, for example, when a new user is acquired, or a content set of the user in the positive feedback weather service changes, the content similarity matrix is updated in real time according to the above manner.
In the embodiment of the present invention, as an optional embodiment, calculating the interest degree of the target user in the weather service content in the content similarity matrix based on the positive feedback content set and the content similarity matrix constructed in advance includes:
aiming at target weather service contents in the content similarity matrix, acquiring a similar content set similar to the target weather service contents;
acquiring a union of a similar content set and a positive feedback content set to obtain a content union;
aiming at each union content in the content union, acquiring union similarity between the union content and target weather service content from a content similarity matrix, and acquiring union score of the union content by a target user;
calculating the product of the union set similarity and the union set score to obtain the union set interest degree of the union set content;
and summing the interest degree of the union of the contents of each union to obtain the interest degree of the target user on the target weather service contents.
In the embodiment of the present invention, the target weather service content may be a content randomly selected from the content similarity matrix, and as an optional embodiment, the following formula is used to calculate the interest degree of the target user in the content (weather service content):
Figure F_210519171612278_278982004
in the formula (I), the compound is shown in the specification,
p uj is a target useruFor contentjThe degree of interest of;
R(u)is a target useruA positive feedback content set;
S(j,K)is and contentjMost similarKA set of similar content for the individual content;
r ui is a target useruFor contentiThe score of (1).
In the embodiment of the invention, the similar content set can be obtained through the content similarity matrix, and a plurality of contents with higher similarity with the target weather service content can be selected from the similar content set.
In the embodiment of the invention, as an optional embodiment, a content is subjected tojAssumptions and contentsjMost similarKThe content set of each content isS(j,K)Set of contents liked by userN(u)If there is an intersection, then for each content in the intersectioniTo provide contentiAnd contentjDegree of similarity ofW ij To the content with the useriIs scoredr ui Multiplying and then accumulating to obtain the useruFor contentjDegree of interest ofP uj 。
In the embodiment of the invention, the scores of the contents of the users can be normalized so as to make the scores consistent in scale. As an alternative embodiment, the more similar the content is to the content that the user has historically been interested in, the more likely it is to get a higher ranking in the user's push list.
In the embodiment of the invention, the preset number of weather service contents before sequencing are selected according to the sequence of the calculated interest degrees from large to small, and the candidate weather service content push set is generated. For example, the candidate parameter M is set, and M weather service contents with a higher interest level are selected from the calculated interest levels corresponding to the weather service contents, thereby obtaining a candidate weather service content push set. As an alternative, M is taken to be 40.
In the embodiment of the invention, for the convenience of distinguishing, the weather service content in the candidate weather service content push set is called as candidate weather service content.
103, respectively extracting words from the positive feedback content set of the target user and the candidate meteorological service content push set, constructing word vectors based on the extracted words and a preset meteorological knowledge dictionary, and constructing entity vectors and entity context vectors corresponding to the extracted words based on the extracted words and a preset geological flood disaster meteorological knowledge map;
in the embodiment of the invention, relevant meteorological service contents are collected in advance, words of the collected meteorological service contents are extracted, repeated words are removed, and a meteorological knowledge dictionary containing V words is constructed.
In the embodiment of the invention, the words of the positive feedback content set and the candidate weather service content push set of the target user are sequentially extracted, so that the keywords corresponding to the contents (the positive feedback weather service contents and the candidate weather service contents) are extracted. And if the extracted word phases are repeated, repeated word removing processing is not carried out.
In the embodiment of the present invention, as an optional embodiment, constructing a word vector based on the extracted words and a preset meteorological knowledge dictionary includes:
and aiming at each extracted word, setting the word vector value of the extracted word to be 1 and setting the word vector values of other words to be 0 according to a preset meteorological knowledge dictionary to obtain the word vector of the extracted word.
In the embodiment of the invention, according to a weather knowledge dictionary, word extraction is carried out on each content in a positive feedback content set and a candidate weather service content push set, in a V-dimensional weather knowledge dictionary, a code value (word vector value) corresponding to the word is set to be 1, the code values of other words are all 0, a word vector of the extracted word is obtained, and thus a V-dimensional one-hot code (one-hot) vector is generated for each extracted word.
In the embodiment of the present invention, as an optional embodiment, constructing an entity vector and an entity context vector corresponding to an extracted word based on the extracted word and a preset mountain torrent geological disaster meteorological knowledgebase, includes:
judging whether the extracted words are entities in the mountain torrent geological disaster meteorological knowledgegraph or not aiming at each extracted word, and if so, constructing a sub-graph of a triple including a first entity, a relation and a second entity according to an initial entity corresponding to the extracted word in the mountain torrent geological disaster meteorological knowledgegraph and other entities in the mountain torrent geological disaster meteorological knowledgegraph and related to the initial entity, wherein the first entity is the initial entity;
setting a first entity vector, a relation vector and a second entity vector for each triple in the sub-spectrum, and assigning the first entity vector, the relation vector and the second entity vector by using a translation embedding algorithm until a difference value between a sum of the first entity vector value and the relation vector value and a second entity vector value is within a preset vector value error threshold value to obtain a sub-spectrum vector set;
extracting a vector corresponding to the initial entity from the sub-graph spectrum vector set to obtain an entity vector;
and calculating the average value of all entity vectors connected with the initial entity in the sub-map vector set to obtain the entity context vector.
In the embodiment of the invention, the sub-map is constructed based on the words extracted from each content (the content in the positive feedback content set and the candidate meteorological service content push set), the meteorological knowledge dictionary and the relationship and the entity defined in the mountain torrent geological disaster meteorological knowledge map. And if the extracted words are entities in the mountain torrent geological disaster meteorological knowledge graph, each extracted word corresponds to a sub-graph, and the corresponding sub-graphs are the same under the condition that the extracted words are the same. The sub-map includes a plurality of triplets (entities, relationships, entities). Then, a translation Embedding (Translating Embedding) algorithm is adopted for feature learning, specifically, initial vectors are respectively set for each entity and relationship in the sub-map, for each (entity h, relationship r, entity t) triplet, the entity h is regarded as a head node, the entity t is regarded as a tail node, the value of the vector is continuously adjusted so that the vector can satisfy h + r = t, h + r and t of all triplets are close until the average difference of all triplets meets the requirement, and at this time, the vector corresponding to the initial entity is the entity vector corresponding to the extracted word.
In the embodiment of the invention, if the extracted word is not an entity in the mountain torrent geological disaster meteorological knowledge map, the vector of the extracted word is constructed by using the 0 vector with the same dimensionality.
In the embodiment of the invention, the entity can be further described by utilizing the context of the entity, so that the accuracy of subsequent pushing is improved. Specifically, the entity vectors of the entities in the sub-map vector set are used, and the average value of the entity vectors of the entities connected with the initial entity is used as the entity context vector of the initial entity.
In the embodiment of the invention, for the extracted words which are not the entities in the mountain torrent geological disaster meteorological knowledgebase, 0 vector with the same dimensionality is used for representing the entity context vector.
Step 104, mapping the entity vectors and the entity context vectors into a space of word vectors, and splicing the word vectors, the mapped entity vectors and the entity context vectors to obtain a vector splicing matrix;
in the embodiment of the invention, the entity vector and the entity context vector are mapped into a space of the word vector to perform dimension reduction processing on the entity vector and the entity context vector, so that the entity vector, the entity context vector and the word vector have the same dimension, and then the word vector, the mapped entity vector and the entity context vector are spliced to obtain the vector splicing matrix.
In the embodiment of the invention, for each extracted word, the word vector corresponding to the extracted word, the entity vector subjected to mapping processing and the entity context vector are spliced to obtain the vector splicing matrix corresponding to the extracted word.
Step 105, aiming at each content, inputting a vector splicing matrix corresponding to the content into a graph convolution neural network to obtain a content vector;
in the embodiment of the invention, the content comprises positive feedback meteorological service content and candidate meteorological service content. And fusing the vector splicing matrix under the framework of the graph convolution neural network to obtain the content vector.
Step 106, acquiring the weight score of each positive feedback weather service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate weather service content push set, and acquiring the push score of each candidate weather service content in the candidate weather service content push set based on the content vector of the positive feedback content set, the content vector of the candidate weather service content push set and the weight score;
in this embodiment of the present invention, as an optional embodiment, the obtaining a weight score of each positive feedback weather service content in the positive feedback content set based on a content vector of the positive feedback content set and a content vector of the candidate weather service content push set includes:
extracting a content vector of a positive feedback meteorological service content in the positive feedback content set, and combining the content vector with the content vector of each candidate meteorological service content in the candidate meteorological service content push set respectively;
inputting the combination vector into a preset attention network aiming at the combination vector of each combination, and normalizing the output of the attention network by using a normalization function to obtain a weight value of the combination vector;
and carrying out mean value calculation on the weight values of the combined vectors to obtain the weight value corresponding to the extracted positive feedback meteorological service content.
In the embodiment of the invention, the positive feedback weather service contents in the positive feedback content set are traversed, for a certain positive feedback weather service content extracted from the positive feedback content set, each candidate weather service content in the set is pushed according to the candidate weather service content, the content vector corresponding to the extracted positive feedback weather service content and the content vector corresponding to the candidate weather service content in the candidate weather service content push set are input into the attention network and are normalized through the softmax function to obtain a weight value (the weight value of the combined vector), each candidate weather service content in the candidate weather service content push set is traversed according to the same method, the weight values of the content vector of the extracted weather service content and the content vector of the candidate weather service content in the candidate weather service content push set are respectively calculated, and the average value calculation is carried out on all the traversed weight values, the weight score of the extracted positive feedback weather service content can be obtained. And repeating the steps to obtain the weight score corresponding to each positive feedback meteorological service content in the positive feedback content set. For example, if the positive feedback content set includes 40 pieces of positive feedback weather service content, which are respectively the first weather service content (corresponding to the first content vector) to the fortieth weather service content (corresponding to the fortieth content vector), the candidate weather service content push set includes 50 pieces of candidate weather service content, which are respectively the fortieth weather service content (corresponding to the fortieth content vector) to the ninety weather service content (corresponding to the ninety content vector), taking the first weather service content as an example, the first content vector is respectively spliced and combined with the fortieth content vector, the forty-second content vector, … and the ninety content vector to obtain a combined vector, a second combined vector, … and a fifty combined vector, the first combined vector is input into a pre-trained attention network, and a weighted value of the first combined vector is obtained through normalization, inputting the second combination vector into a pre-trained attention network, and obtaining a weight value of the second combination vector through normalization, …, until obtaining a weight value of a fiftieth combination vector; and carrying out weighted average on the weight values of the first combination vector to the fiftieth combination vector to obtain a weight score corresponding to the first weather service content (first content vector).
In this embodiment of the present invention, as an optional embodiment, the obtaining a push score of each candidate weather service content in the candidate weather service content push set based on a content vector of the positive feedback content set, a content vector of the candidate weather service content push set, and a weight score includes:
aiming at each candidate meteorological service content of the candidate meteorological service content push set, respectively combining the content vector of the candidate meteorological service content with the content vector of each positive feedback meteorological service content of the positive feedback content set to obtain a candidate combined vector;
for each candidate combination vector, inputting the candidate combination vector into a preset attention network, and normalizing the output of the attention network by using a normalization function to obtain the score of the candidate combination vector;
acquiring a weight value of a combined vector corresponding to the positive feedback meteorological service content in the candidate combined vector, and calculating the product of the value of the candidate combined vector and the acquired weight value to obtain the weight value of the candidate combined vector;
and carrying out weighted average on the weighted value of each candidate combination vector to obtain the pushing value of the candidate meteorological service content.
In the embodiment of the present invention, as described above, for a positive feedback content set including 40 pieces of positive feedback weather service content and a candidate weather service content push set including 50 pieces of candidate weather service content, taking the fortieth weather service content as an example, a forty-first content vector is respectively spliced and combined with a first content vector, a second content vector, …, and a fortieth content vector to obtain a first candidate combined vector, a second candidate combined vector, …, and a forty-candidate combined vector, the first candidate combined vector is input to a pre-trained attention network, and a score of the first candidate combined vector is obtained through normalization, the second candidate combined vector is input to a pre-trained attention network, and a score of the second candidate combined vector is obtained through normalization, …, until a score of the first candidate combined vector is obtained; and multiplying the score of the first candidate combination vector and the weight score of the first content vector to obtain a weighted value of the first candidate combination vector, and carrying out weighted average on the weighted value of the first candidate combination vector to the weighted value of the fortieth candidate combination vector to obtain a pushed score of the fortieth weather service content.
And 107, sequencing according to the pushing scores, and acquiring a plurality of candidate weather service contents which are sequenced at the front for pushing.
In the embodiment of the invention, a pushing parameter N is set, and N items with the highest scores are pushed for a user. As an alternative embodiment, N is taken to be 20. For example, according to the push scores of the candidate weather service contents, the candidate weather service contents are sorted from high to low, and finally, 20 candidate weather service contents with higher push scores are pushed.
In this embodiment of the present invention, as an optional embodiment, the method further includes:
the attribute information of the target user is obtained, a plurality of candidate weather service contents which are ranked in the front are screened on the basis of the attribute information, and the screened candidate weather service contents are pushed to the target user.
In the embodiment of the invention, when the candidate weather service content is pushed, the family identity information and/or the geographic position information of the target user at the moment are/is acquired, and the plurality of candidate weather service contents ranked at the top are filtered again based on the family identity information and/or the geographic position information of the target user at the moment, so that the accuracy of the pushed candidate weather service content is further improved.
In the embodiment of the invention, a test is carried out on a mountain torrent geological disaster meteorological service content map with 1.5 ten thousand entities and 6 ten thousand pieces of relational data, the content of the map mainly comprises meteorological knowledge, disaster records, countermeasures and the like related to mountain torrent geological disasters, wherein the data are from Baidu encyclopedias and Chinese Wikipedia. 100 recruited experimenters browsed 10 pieces of content according to their own preferences, and 20 pushes were generated for each experimenter in 0.2 seconds from the 10 user behavior records described above. The experiment result shows that the comprehensive satisfaction degree of experimenters to the push content is about 92%.
In the embodiment of the invention, the pushed weather service content is obtained through acquisition of the historical browsing behavior of the target user, construction of a candidate weather service content pushing set and characteristic learning of a weather knowledge map based on mountain torrent geological disasters. Wherein, a candidate meteorological service content push set is constructed by adopting a collaborative filtering recommendation algorithm, the collaborative filtering recommendation algorithm can process unstructured complex objects and has higher processing speed, and a Deep Knowledge sensing Network (DKN) algorithm is used for learning based on the characteristics of the mountain torrent geological disaster meteorological wisdom, the DKN algorithm can utilize the constructed mountain torrent geological disaster meteorological wisdom to mine the interest of users, can effectively solve the problems of sparsity and cold start, enhance the push effect, and can reduce the calculation burden brought by introducing the mountain torrent geological disaster meteorological disaster wisdom, so that the push accuracy can be effectively improved by combining the collaborative filtering recommendation algorithm with the mountain torrent geological disaster meteorological Knowledge wisdom, more personalized push is provided for the users, the relevance of the push content is stronger, and the users are attracted, and a good pushing effect is realized.
Fig. 3 is a schematic structural diagram of a device for pushing weather service content according to an embodiment of the present invention. As shown in fig. 3, the apparatus includes:
the content acquisition module 301 is configured to acquire a positive feedback content set that a target user has positive feedback on browsed weather service content;
in this embodiment of the present invention, as an optional embodiment, the content obtaining module 301 includes:
a content acquiring unit (not shown in the figure) for acquiring the weather service content browsed by the target user;
the first matching unit is used for placing the meteorological service content into a positive feedback content set if the meteorological service content is matched with the interesting meteorological service classification preset by the target user;
the second matching unit is used for placing the meteorological service content into a positive feedback content set if the meteorological service content is not matched with the meteorological service classification preset by the target user and if the score of the meteorological service content exceeds a preset score threshold value;
and the time length matching unit is used for placing the meteorological service content into a positive feedback content set if the meteorological service content is not scored and the browsing time length of the meteorological service content exceeds a preset time length threshold value.
The candidate content generation module 302 is configured to calculate the interest degree of the target user in the weather service content in the content similarity matrix based on the positive feedback content set and a pre-constructed content similarity matrix, and generate a candidate weather service content push set based on the calculated interest degree;
in this embodiment of the present invention, as an optional embodiment, the candidate content generating module 302 includes:
the similarity acquisition unit is used for acquiring a similar content set similar to the target meteorological service content aiming at the target meteorological service content in the content similarity matrix;
the union processing unit is used for acquiring a union of the similar content set and the positive feedback content set to obtain a content union;
the score obtaining unit is used for obtaining the union set similarity between the union set content and the target weather service content from the content similarity matrix aiming at each union set content in the content union set, and obtaining the union set score of the union set content by the target user;
the product unit is used for calculating the product of the union set similarity and the union set score to obtain the union set interest degree of the union set content;
and the summing unit is used for summing the interest degree of the union of the contents of the unions to obtain the interest degree of the target user on the target weather service contents.
In the embodiment of the present invention, as an optional embodiment, the normalization processing is performed on the positive feedback content matrix by using the following formula:
Figure F_210519171612482_482107005
in the embodiment of the present invention, as another optional embodiment, the positive feedback content matrix is normalized by using the following formula:
Figure F_210519171612544_544607006
the interest degree of the target user in the content (weather service content) is calculated by the following formula:
Figure F_210519171612622_622732007
the vector construction module 303 is configured to perform word extraction on the positive feedback content set and the candidate weather service content push set of the target user, construct a word vector based on the extracted words and a preset weather knowledge dictionary, and construct an entity vector and an entity context vector corresponding to the extracted words based on the extracted words and a preset mountain torrent geological disaster weather knowledge map;
in the embodiment of the present invention, as an optional embodiment, constructing a word vector based on the extracted words and a preset meteorological knowledge dictionary includes:
and aiming at each extracted word, setting the word vector value of the extracted word to be 1 and setting the word vector values of other words to be 0 according to a preset meteorological knowledge dictionary to obtain the word vector of the extracted word.
In this embodiment of the present invention, as an optional embodiment, the vector constructing module 303 is specifically configured to:
judging whether the extracted words are entities in the mountain torrent geological disaster meteorological knowledgegraph or not aiming at each extracted word, and if so, constructing a sub-graph of a triple including a first entity, a relation and a second entity according to an initial entity corresponding to the extracted word in the mountain torrent geological disaster meteorological knowledgegraph and other entities in the mountain torrent geological disaster meteorological knowledgegraph and related to the initial entity, wherein the first entity is the initial entity;
setting a first entity vector, a relation vector and a second entity vector for each triple in the sub-spectrum, and assigning the first entity vector, the relation vector and the second entity vector by using a translation embedding algorithm until a difference value between a sum of the first entity vector value and the relation vector value and a second entity vector value is within a preset vector value error threshold value to obtain a sub-spectrum vector set;
extracting a vector corresponding to the initial entity from the sub-graph spectrum vector set to obtain an entity vector;
and calculating the average value of all entity vectors connected with the initial entity in the sub-map vector set to obtain the entity context vector.
The vector splicing module 304 is configured to map the entity vector and the entity context vector into a space of a word vector, and splice the word vector, the mapped entity vector, and the entity context vector to obtain a vector splicing matrix;
in the embodiment of the invention, the dimension reduction processing is carried out on the entity vector and the entity context vector.
A content vector obtaining module 305, configured to, for each content, input a vector stitching matrix corresponding to the content into a graph convolution neural network to obtain a content vector;
in the embodiment of the invention, the content comprises positive feedback meteorological service content and candidate meteorological service content. And fusing the vector splicing matrix under the framework of the graph convolution neural network to obtain the content vector.
The push score calculation module 306 is configured to obtain a weight score of each positive feedback weather service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate weather service content push set, and obtain a push score of each candidate weather service content in the candidate weather service content push set based on the content vector of the positive feedback content set, the content vector of the candidate weather service content push set, and the weight score;
in this embodiment of the present invention, as an optional embodiment, the pushed score calculating module 306 is specifically configured to:
extracting a content vector of a positive feedback meteorological service content in the positive feedback content set, and combining the content vector with the content vector of each candidate meteorological service content in the candidate meteorological service content push set respectively;
inputting the combination vector into a preset attention network aiming at the combination vector of each combination, and normalizing the output of the attention network by using a normalization function to obtain a weight value of the combination vector;
and carrying out mean value calculation on the weight values of the combined vectors to obtain the weight value corresponding to the extracted positive feedback meteorological service content.
In this embodiment of the present invention, as another optional embodiment, the pushed score calculating module 306 is further specifically configured to:
aiming at each candidate meteorological service content of the candidate meteorological service content push set, respectively combining the content vector of the candidate meteorological service content with the content vector of each positive feedback meteorological service content of the positive feedback content set to obtain a candidate combined vector;
for each candidate combination vector, inputting the candidate combination vector into a preset attention network, and normalizing the output of the attention network by using a normalization function to obtain the score of the candidate combination vector;
acquiring a weight value of a combined vector corresponding to the positive feedback meteorological service content in the candidate combined vector, and calculating the product of the value of the candidate combined vector and the acquired weight value to obtain the weight value of the candidate combined vector;
and carrying out weighted average on the weighted value of each candidate combination vector to obtain the pushing value of the candidate meteorological service content.
And the content pushing module 307 is configured to sort according to the pushing scores, and obtain a plurality of candidate weather service contents sorted in the front for pushing.
In this embodiment of the present invention, as an optional embodiment, the apparatus further includes:
a content similarity matrix construction module (not shown in the figure) for acquiring a positive feedback weather service content set of the user having positive feedback on the browsed weather service content; and generating a content similarity matrix by utilizing a collaborative filtering recommendation algorithm based on each positive feedback meteorological service content set.
In this embodiment, as an optional embodiment, the content similarity matrix constructing module includes:
the content collection unit is used for acquiring a positive feedback meteorological service content set of which the user has positive feedback on browsed meteorological service contents;
the matrix construction unit is used for merging the positive feedback meteorological service content sets of all users and performing content duplication elimination processing to obtain a content duplication elimination set, and constructing a positive feedback content matrix based on the content duplication elimination set, wherein the rows and columns of the positive feedback content matrix are respectively the content contained in the content duplication elimination set, and the number of rows and columns is equal to the number of content strips contained in the content duplication elimination set;
the editing unit is used for editing row and column values of the positive feedback content matrix according to the positive feedback meteorological service content set of each user;
and the normalizing unit is used for normalizing the row and column values of the edited positive feedback content matrix to obtain a content similarity matrix.
In this embodiment, as another optional embodiment, the apparatus further includes:
and the attribute information matching module is used for acquiring the attribute information of the target user, screening a plurality of candidate weather service contents which are ranked in the front on the basis of the attribute information, and pushing the screened candidate weather service contents to the target user.
In the embodiment of the invention, when the candidate weather service content is pushed, the family identity information and/or the geographic position information of the target user at the moment are/is acquired, and the plurality of candidate weather service contents ranked at the top are filtered again based on the family identity information and/or the geographic position information of the target user at the moment, so that the accuracy of the pushed candidate weather service content is further improved.
As shown in fig. 4, an embodiment of the present application provides a computer device 400 for executing the method for pushing weather service content in fig. 1, the device includes a memory 401, a processor 402 and a computer program stored on the memory 401 and operable on the processor 402, wherein the processor 402 implements the steps of the method for pushing weather service content when executing the computer program.
Specifically, the memory 401 and the processor 402 can be general-purpose memory and processor, and are not limited to specific examples, and the processor 402 can execute the method for pushing weather service content when executing the computer program stored in the memory 401.
Corresponding to the method for pushing weather service content in fig. 1, the present application also provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the steps of the method for pushing weather service content.
In particular, the storage medium can be a general-purpose storage medium, such as a removable disk, a hard disk, etc., and when the computer program on the storage medium is executed, the method for pushing the weather service content can be executed.
In the embodiments provided in the present application, it should be understood that the disclosed system and method may be implemented in other ways. The above-described system embodiments are merely illustrative, and for example, the division of the units is only one logical functional division, and there may be other divisions in actual implementation, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of systems or units through some communication interfaces, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus once an item is defined in one figure, it need not be further defined and explained in subsequent figures, and moreover, the terms "first", "second", "third", etc. are used merely to distinguish one description from another and are not to be construed as indicating or implying relative importance.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present application, and are used for illustrating the technical solutions of the present application, but not limiting the same, and the scope of the present application is not limited thereto, and although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope disclosed in the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the present disclosure, which should be construed in light of the above teachings. Are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. A method for pushing weather service content, comprising:
acquiring a positive feedback content set of a target user having positive feedback on browsed meteorological service contents;
calculating the interest degree of the target user to the weather service content in the content similarity matrix based on the positive feedback content set and the content similarity matrix constructed in advance, and generating a candidate weather service content push set based on the calculated interest degree;
respectively extracting words from a positive feedback content set and a candidate meteorological service content push set of a target user, constructing word vectors based on the extracted words and a preset meteorological knowledge dictionary, and constructing entity vectors and entity context vectors corresponding to the extracted words based on the extracted words and a preset geological disaster meteorological knowledge map of torrential flood;
mapping the entity vector and the entity context vector to a space of the word vector, and splicing the word vector, the mapped entity vector and the entity context vector to obtain a vector splicing matrix;
aiming at each content, inputting a vector splicing matrix corresponding to the content into a graph convolution neural network to obtain a content vector;
acquiring a weight score of each positive feedback meteorological service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate meteorological service content push set, and acquiring a push score of each candidate meteorological service content in the candidate meteorological service content push set based on the content vector of the positive feedback content set, the content vector of the candidate meteorological service content push set and the weight score;
and sequencing according to the pushing scores, and acquiring a plurality of candidate meteorological service contents which are sequenced at the front for pushing.
2. The method of claim 1, wherein obtaining a positive feedback content set that the target user has positive feedback on the browsed weather service content comprises:
acquiring meteorological service content browsed by a target user;
if the weather service content is matched with the interested weather service classification preset by the target user, placing the weather service content in a positive feedback content set;
if the weather service content is not matched with the preset weather service classification of the target user, if the score of the weather service content exceeds a preset score threshold value, placing the weather service content in a positive feedback content set;
if the weather service content is not scored, if the browsing time of the weather service content exceeds a preset time threshold, the weather service content is placed in a positive feedback content set.
3. The method of claim 1, wherein constructing the content similarity matrix comprises:
acquiring a positive feedback meteorological service content set with positive feedback on browsed meteorological service contents by a user;
and generating a content similarity matrix by utilizing a collaborative filtering recommendation algorithm based on each positive feedback meteorological service content set.
4. The method of claim 3, wherein generating a content similarity matrix using a collaborative filtering recommendation algorithm based on each set of positive feedback weather service content comprises:
merging the positive feedback meteorological service content sets of each user and performing content deduplication processing to obtain a content deduplication set, constructing a positive feedback content matrix based on the content deduplication set, wherein rows and columns of the positive feedback content matrix are respectively contents contained in the content deduplication set, and the number of rows and columns is equal to the number of content strips contained in the content deduplication set;
editing row and column values of a positive feedback content matrix according to the positive feedback meteorological service content set of each user;
and normalizing the row and column values of the edited positive feedback content matrix to obtain a content similarity matrix.
5. The method according to any one of claims 1 to 4, wherein the calculating the interest degree of the target user in the weather service content in the content similarity matrix based on the positive feedback content set and the pre-constructed content similarity matrix comprises:
aiming at target weather service contents in the content similarity matrix, acquiring a similar content set similar to the target weather service contents;
acquiring a union of a similar content set and a positive feedback content set to obtain a content union;
aiming at each union content in the content union, acquiring union similarity between the union content and target weather service content from a content similarity matrix, and acquiring union score of the union content by a target user;
calculating the product of the union set similarity and the union set score to obtain the union set interest degree of the union set content;
and summing the interest degree of the union of the contents of each union to obtain the interest degree of the target user on the target weather service contents.
6. The method according to any one of claims 1 to 4, wherein the constructing a word vector based on the extracted words and a preset meteorological knowledge dictionary comprises:
and aiming at each extracted word, setting the word vector value of the extracted word to be 1 and setting the word vector values of other words to be 0 according to a preset meteorological knowledge dictionary to obtain the word vector of the extracted word.
7. The method according to any one of claims 1 to 4, wherein the constructing of the entity vector and the entity context vector corresponding to the extracted word based on the extracted word and a pre-set mountain torrent geological disaster meteorological knowledgebase comprises:
judging whether the extracted words are entities in the mountain torrent geological disaster meteorological knowledgegraph or not aiming at each extracted word, and if so, constructing a sub-graph of a triple including a first entity, a relation and a second entity according to an initial entity corresponding to the extracted word in the mountain torrent geological disaster meteorological knowledgegraph and other entities in the mountain torrent geological disaster meteorological knowledgegraph and related to the initial entity, wherein the first entity is the initial entity;
setting a first entity vector, a relation vector and a second entity vector for each triple in the sub-spectrum, and assigning the first entity vector, the relation vector and the second entity vector by using a translation embedding algorithm until a difference value between a sum of the first entity vector value and the relation vector value and a second entity vector value is within a preset vector value error threshold value to obtain a sub-spectrum vector set;
extracting a vector corresponding to the initial entity from the sub-graph spectrum vector set to obtain an entity vector;
and calculating the average value of all entity vectors connected with the initial entity in the sub-map vector set to obtain the entity context vector.
8. The method of any of claims 1 to 4, wherein obtaining a weight score for each positive feedback weather service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate weather service content push set comprises:
extracting a content vector of a positive feedback meteorological service content in the positive feedback content set, and combining the content vector with the content vector of each candidate meteorological service content in the candidate meteorological service content push set respectively;
inputting the combination vector into a preset attention network aiming at the combination vector of each combination, and normalizing the output of the attention network by using a normalization function to obtain a weight value of the combination vector;
and carrying out mean value calculation on the weight values of the combined vectors to obtain the weight value corresponding to the extracted positive feedback meteorological service content.
9. The method of claim 8, wherein obtaining the push score for each candidate weather service content in the candidate weather service content push set based on the content vector of the positive feedback content set, the content vector of the candidate weather service content push set, and the weight score comprises:
aiming at each candidate meteorological service content of the candidate meteorological service content push set, respectively combining the content vector of the candidate meteorological service content with the content vector of each positive feedback meteorological service content of the positive feedback content set to obtain a candidate combined vector;
for each candidate combination vector, inputting the candidate combination vector into a preset attention network, and normalizing the output of the attention network by using a normalization function to obtain the score of the candidate combination vector;
acquiring a weight value of a combined vector corresponding to the positive feedback meteorological service content in the candidate combined vector, and calculating the product of the value of the candidate combined vector and the acquired weight value to obtain the weight value of the candidate combined vector;
and carrying out weighted average on the weighted value of each candidate combination vector to obtain the pushing value of the candidate meteorological service content.
10. An apparatus for pushing weather service content, comprising:
the content acquisition module is used for acquiring a positive feedback content set of the target user with positive feedback on the browsed meteorological service content;
the candidate content generation module is used for calculating the interest degree of the target user in the meteorological service content in the content similarity matrix based on the positive feedback content set and the content similarity matrix which is constructed in advance, and generating a candidate meteorological service content push set based on the calculated interest degree;
the vector construction module is used for respectively extracting words from a positive feedback content set and a candidate meteorological service content push set of a target user, constructing word vectors based on the extracted words and a preset meteorological knowledge dictionary, and constructing entity vectors and entity context vectors corresponding to the extracted words based on the extracted words and a preset geological flood disaster meteorological knowledge map;
the vector splicing module is used for mapping the entity vector and the entity context vector into a space of the word vector, and splicing the word vector, the mapped entity vector and the entity context vector to obtain a vector splicing matrix;
the content vector acquisition module is used for inputting a vector splicing matrix corresponding to each content into the graph convolution neural network to obtain a content vector;
the push score calculation module is used for acquiring the weight score of each positive feedback meteorological service content in the positive feedback content set based on the content vector of the positive feedback content set and the content vector of the candidate meteorological service content push set, and acquiring the push score of each candidate meteorological service content in the candidate meteorological service content push set based on the content vector of the positive feedback content set, the content vector of the candidate meteorological service content push set and the weight score;
and the content pushing module is used for sequencing according to the pushing scores, and acquiring a plurality of candidate meteorological service contents which are sequenced at the front for pushing.
CN202110731097.3A 2021-06-30 2021-06-30 Method and device for pushing meteorological service content Active CN113254789B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110731097.3A CN113254789B (en) 2021-06-30 2021-06-30 Method and device for pushing meteorological service content

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110731097.3A CN113254789B (en) 2021-06-30 2021-06-30 Method and device for pushing meteorological service content

Publications (2)

Publication Number Publication Date
CN113254789A true CN113254789A (en) 2021-08-13
CN113254789B CN113254789B (en) 2021-09-17

Family

ID=77190210

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110731097.3A Active CN113254789B (en) 2021-06-30 2021-06-30 Method and device for pushing meteorological service content

Country Status (1)

Country Link
CN (1) CN113254789B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113887613A (en) * 2021-09-29 2022-01-04 平安银行股份有限公司 Deep learning method, device and equipment based on attention mechanism and storage medium
CN116578781A (en) * 2023-04-28 2023-08-11 北京天译科技有限公司 Meteorological service pushing method and server applying neural network algorithm

Citations (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102611785A (en) * 2011-01-20 2012-07-25 北京邮电大学 Personalized active news recommending service system and method for mobile phone user
EP2518679A1 (en) * 2011-04-26 2012-10-31 YooChoose GmbH Method and system fo recommending geo-tagged items
CN103064924A (en) * 2012-12-17 2013-04-24 浙江鸿程计算机系统有限公司 Travel destination situation recommendation method based on geotagged photo excavation
CN108664564A (en) * 2018-04-13 2018-10-16 东华大学 A kind of improvement collaborative filtering recommending method based on item contents feature
CN108874914A (en) * 2018-05-29 2018-11-23 吉林大学 A kind of information recommendation method based on the long-pending and neural collaborative filtering of picture scroll
CN109635200A (en) * 2018-12-18 2019-04-16 南京邮电大学 A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user
CN110191171A (en) * 2019-05-27 2019-08-30 北京气象在线科技有限公司 Weather information transmission method
CN110490717A (en) * 2019-09-05 2019-11-22 齐鲁工业大学 Method of Commodity Recommendation and system based on user conversation and figure convolutional neural networks
CN110580314A (en) * 2019-07-16 2019-12-17 华南师范大学 Course recommendation method and system based on graph convolution neural network and dynamic weight
CN112417294A (en) * 2020-12-03 2021-02-26 广州瀚信通信科技股份有限公司 Intelligent business recommendation method based on neural network mining model
CN112559764A (en) * 2020-12-10 2021-03-26 北京中视广信科技有限公司 Content recommendation method based on domain knowledge graph
CN112612973A (en) * 2020-12-31 2021-04-06 重庆邮电大学 Personalized intelligent clothing matching recommendation method combining knowledge graph
CN112667346A (en) * 2021-03-16 2021-04-16 深圳市火乐科技发展有限公司 Weather data display method and device, electronic equipment and storage medium
CN112883175A (en) * 2021-02-10 2021-06-01 武汉大学 Meteorological service interaction method and system combining pre-training model and template generation
CN112927801A (en) * 2021-01-24 2021-06-08 武汉东湖大数据交易中心股份有限公司 Health prediction model construction method and device based on dressing meteorological index
CN112966091A (en) * 2019-12-12 2021-06-15 复旦大学 Knowledge graph recommendation system fusing entity information and heat

Patent Citations (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102611785A (en) * 2011-01-20 2012-07-25 北京邮电大学 Personalized active news recommending service system and method for mobile phone user
EP2518679A1 (en) * 2011-04-26 2012-10-31 YooChoose GmbH Method and system fo recommending geo-tagged items
CN103064924A (en) * 2012-12-17 2013-04-24 浙江鸿程计算机系统有限公司 Travel destination situation recommendation method based on geotagged photo excavation
CN108664564A (en) * 2018-04-13 2018-10-16 东华大学 A kind of improvement collaborative filtering recommending method based on item contents feature
CN108874914A (en) * 2018-05-29 2018-11-23 吉林大学 A kind of information recommendation method based on the long-pending and neural collaborative filtering of picture scroll
CN109635200A (en) * 2018-12-18 2019-04-16 南京邮电大学 A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user
CN110191171A (en) * 2019-05-27 2019-08-30 北京气象在线科技有限公司 Weather information transmission method
CN110580314A (en) * 2019-07-16 2019-12-17 华南师范大学 Course recommendation method and system based on graph convolution neural network and dynamic weight
CN110490717A (en) * 2019-09-05 2019-11-22 齐鲁工业大学 Method of Commodity Recommendation and system based on user conversation and figure convolutional neural networks
CN112966091A (en) * 2019-12-12 2021-06-15 复旦大学 Knowledge graph recommendation system fusing entity information and heat
CN112417294A (en) * 2020-12-03 2021-02-26 广州瀚信通信科技股份有限公司 Intelligent business recommendation method based on neural network mining model
CN112559764A (en) * 2020-12-10 2021-03-26 北京中视广信科技有限公司 Content recommendation method based on domain knowledge graph
CN112612973A (en) * 2020-12-31 2021-04-06 重庆邮电大学 Personalized intelligent clothing matching recommendation method combining knowledge graph
CN112927801A (en) * 2021-01-24 2021-06-08 武汉东湖大数据交易中心股份有限公司 Health prediction model construction method and device based on dressing meteorological index
CN112883175A (en) * 2021-02-10 2021-06-01 武汉大学 Meteorological service interaction method and system combining pre-training model and template generation
CN112667346A (en) * 2021-03-16 2021-04-16 深圳市火乐科技发展有限公司 Weather data display method and device, electronic equipment and storage medium

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
WANG TONGXUAN等: ""Integrating knowledge graph into deep neural network-based recommender system"", 《2019 4TH IEEE INTERNATIONAL CONFERENCE ON CYBERNETICS (CYBCONF)》 *
常亮 等: ""知识图谱的推荐系统综述"", 《智能系统学报》 *
梅钰 等: ""气象文本推荐研究"", 《计算机应用与软件》 *
陈曦 等: ""融合知识图谱和协同过滤的学生成绩预测方法"", 《计算机应用》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113887613A (en) * 2021-09-29 2022-01-04 平安银行股份有限公司 Deep learning method, device and equipment based on attention mechanism and storage medium
CN116578781A (en) * 2023-04-28 2023-08-11 北京天译科技有限公司 Meteorological service pushing method and server applying neural network algorithm
CN116578781B (en) * 2023-04-28 2023-10-24 北京天译科技有限公司 Meteorological service pushing method and server applying neural network algorithm

Also Published As

Publication number Publication date
CN113254789B (en) 2021-09-17

Similar Documents

Publication Publication Date Title
CN107944986B (en) Method, system and equipment for recommending O2O commodities
CN108280114B (en) Deep learning-based user literature reading interest analysis method
CN113254789B (en) Method and device for pushing meteorological service content
CN112966091B (en) Knowledge map recommendation system fusing entity information and heat
CN111259173B (en) Search information recommendation method and device
CN105224699A (en) A kind of news recommend method and device
CN106096609B (en) A kind of merchandise query keyword automatic generation method based on OCR
CN110750640A (en) Text data classification method and device based on neural network model and storage medium
KR101224312B1 (en) Friend recommendation method for SNS user, recording medium for the same, and SNS and server using the same
CN105719191A (en) System and method of discovering social group having unspecified behavior senses in multi-dimensional space
CN111353044B (en) Comment-based emotion analysis method and system
CN109508385A (en) A kind of character relation analysis method in web page news data based on Bayesian network
CN111429161B (en) Feature extraction method, feature extraction device, storage medium and electronic equipment
Islam et al. Review analysis of ride-sharing applications using machine learning approaches: Bangladesh perspective
CN110910235A (en) Method for detecting abnormal behavior in credit based on user relationship network
US10521727B2 (en) System, method, and storage medium for generating hypotheses in data sets
CN114153965A (en) Content and map combined public opinion event recommendation method, system and terminal
Singgalen Comparative analysis of decision tree and support vector machine algorithm in sentiment classification for birds of paradise content
CN115641179A (en) Information pushing method and device and electronic equipment
Wang et al. Understanding the spatial dimension of natural language by measuring the spatial semantic similarity of words through a scalable geospatial context window
CN115409630A (en) Insurance product accurate recommendation method based on mixed recommendation algorithm
CN112765465B (en) User map-based recommendation method
CN109711651B (en) Hidden factor model wecker task recommendation method with correction vector
CN111611455A (en) User group division method based on user emotional behavior characteristics under microblog hot topics
Pratama et al. Tourism recommendation system using weighted hybrid method in Bali Island

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant