WO2020207074A1 - 一种信息推送的方法及设备 - Google Patents

一种信息推送的方法及设备 Download PDF

Info

Publication number
WO2020207074A1
WO2020207074A1 PCT/CN2019/130897 CN2019130897W WO2020207074A1 WO 2020207074 A1 WO2020207074 A1 WO 2020207074A1 CN 2019130897 W CN2019130897 W CN 2019130897W WO 2020207074 A1 WO2020207074 A1 WO 2020207074A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
candidate
similarity
target
target object
Prior art date
Application number
PCT/CN2019/130897
Other languages
English (en)
French (fr)
Inventor
龙信文
胡希平
郑建波
高英
程俊
Original Assignee
中国科学院深圳先进技术研究院
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 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2020207074A1 publication Critical patent/WO2020207074A1/zh

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
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/957—Browsing optimisation, e.g. caching or content distillation

Definitions

  • the invention belongs to the technical field of data processing, and in particular relates to a method and equipment for pushing information.
  • the existing information push technology mainly recommends the related information associated with the query record to the user based on the user's query record for the information.
  • the information pushed in this way has a strong relationship with the query record, which is difficult Actively expand the field of information pushed, thereby reducing the diversity and richness of the pushed information, and reducing the efficiency of users to obtain information.
  • the embodiments of the present invention provide an information push method and device to solve the existing information push technology, mainly based on the user's query record for information, recommend associated information associated with the query record to the user Because the information pushed in this way has a strong relationship with the query record, it is difficult to actively expand the field of the pushed information, thereby reducing the diversity and richness of the pushed information, and reducing the efficiency of users to obtain information. .
  • the first aspect of the embodiments of the present invention provides a method for pushing information, including:
  • the associated information database contains historical query information of the associated object
  • a second aspect of the embodiments of the present invention provides a device for pushing information, including:
  • the information query record obtaining unit is used to obtain the information query record of the target object
  • a similarity calculation unit configured to calculate the similarity between the candidate object and the target object based on the information query record and the candidate query record of the candidate object;
  • An associated object selecting unit configured to select an associated object of the target object from the candidate objects based on the similarity corresponding to each candidate object
  • a recommended information list generating unit configured to generate a recommended information list matching the target object based on the associated information database of the associated object; the associated information database contains historical query information of the associated object;
  • the to-be-pushed information list generating unit is configured to merge the recommended information lists of all the associated objects to generate the to-be-pushed information list of the target object.
  • a third aspect of the embodiments of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and running on the processor.
  • the processor executes the computer program, Realize the steps of the first aspect.
  • a fourth aspect of the embodiments of the present invention provides a computer-readable storage medium that stores a computer program that implements the steps of the first aspect when the computer program is executed by a processor.
  • the embodiment of the present invention obtains the query record of the target object that needs to push information, calculates the similarity between the candidate object and the target object according to the query record, and determines the associated object that has a strong association relationship with the target object according to the similarity, which means The browsing habits and information levels of the associated objects are similar to those of the target object. Therefore, a recommended information list can be generated according to the associated information database of the associated object, and a list of information to be pushed for the target object can be generated through the recommended information lists of multiple associated objects. The target object pushes the information contained in the list to be pushed.
  • the present invention does not only rely on the query record of the target object to determine the push information, but also determines the associated object with higher object attribute similarity to the target object based on the query record of the target object.
  • the recommended information list is generated according to the associated information database of the associated object. Since different associated objects browse more information fields and the similarity between the associated object and the target object is relatively high, it can be inferred that the information field of the associated object is also very interesting. It is likely to be the information field that the target object is interested in. It achieves the purpose of automatically expanding the information field of the push information. While ensuring accurate push, it can also increase the scope of the information field and improve the information push efficiency of the target object.
  • FIG. 1 is an implementation flowchart of an information push method provided by the first embodiment of the present invention
  • FIG. 2 is a specific implementation flowchart of an information push method S102 according to the second embodiment of the present invention.
  • FIG. 3 is a specific implementation flowchart of an information push method S1023 provided by the third embodiment of the present invention.
  • FIG. 4 is a specific implementation flowchart of an information pushing method S1025 provided by the fourth embodiment of the present invention.
  • FIG. 5 is a specific implementation flowchart of an information pushing method S104 provided by the fifth embodiment of the present invention.
  • FIG. 6 is a specific implementation flowchart of a method S105 for pushing information provided by the sixth embodiment of the present invention.
  • FIG. 7 is a structural block diagram of a device for pushing information according to an embodiment of the present invention.
  • Fig. 8 is a schematic diagram of a terminal device according to another embodiment of the present invention.
  • the embodiment of the present invention obtains the query record of the target object that needs to push information, calculates the similarity between the candidate object and the target object according to the query record, and determines the associated object that has a strong association relationship with the target object according to the similarity, which means The browsing habits and information levels of the associated objects are similar to those of the target object. Therefore, a recommended information list can be generated according to the associated information database of the associated object, and a list of information to be pushed for the target object can be generated through the recommended information lists of multiple associated objects.
  • the target object pushes the information contained in the list to be pushed, which solves the existing information push technology. It is mainly based on the user's query record for the information, recommending the associated information associated with the query record to the user.
  • the information pushed in this way is due to There is a strong relationship with the query record, and it is difficult to proactively expand the field of the pushed information, thereby reducing the diversity and richness of the pushed information, and reducing the efficiency of users to obtain
  • the execution subject of the process is the terminal device.
  • the terminal equipment includes, but is not limited to: servers, computers, smart phones, and tablet computers that can perform information push operations.
  • Fig. 1 shows an implementation flow chart of the information pushing method provided by the first embodiment of the present invention, which is detailed as follows:
  • the terminal device can identify the user who needs to push as the target object. If the terminal device is a user terminal of a certain user, the user terminal can identify the user logged in to the user terminal as the target user, and generate a corresponding list of information to be pushed for the user to which the user terminal belongs. If the terminal device is a server, the server can receive the user's push request, and according to the user identification carried in the push request, identify the requesting user corresponding to the user identification as the target user, and push information to the target user; of course, the server It is also possible to automatically push information for user objects on the entire platform. In this case, the terminal device can create multiple parallel threads, and simultaneously produce multiple target object lists of information to be pushed through multiple parallel threads.
  • the terminal device will Add a creation identifier for each user account. After the list of information to be pushed for a certain user account is created, the creation identifier of the user account will be set as the first value. The next time the target object is selected, you can Create the bit value of the identifier, identify the user account that has not been created, and identify it as the target object.
  • the list to be pushed has a valid duration.
  • the terminal device After the terminal device creates a list of information to be pushed for the target object, it will create a valid timer. When the count value of the valid timer is equal to When the valid duration is reached, it will be recognized that the list to be pushed is not an invalid push list, and the list of information to be pushed is regenerated for the target object.
  • the terminal device can also set multiple list invalidation conditions.
  • the list to be pushed is identified as an invalid push list; if It is detected that the number of query records newly created by the user is greater than the preset update number threshold, and the list to be pushed is identified as an invalid push list.
  • the terminal device can create an object database for the target object.
  • the object database can be used to store all operation records of the target object.
  • the operation records include information query records, which are specifically used to record the target object. Information queried.
  • the terminal device can be applied to the field of language learning.
  • the information query record can be used to record the words queried by the target object, and the information query records corresponding to all the queried words are stored in the object of the target object In the database; the terminal device can also be applied to the news recommendation field.
  • the information query record can be used to record the news articles browsed by the target object, and the information query record recording the queried article identifier is stored in the In the object database of the target object.
  • the similarity between the candidate object and the target object is calculated according to the information query record and the candidate query record of the candidate object.
  • the terminal device may also obtain the query records of all candidate objects except the target object, and calculate the degree of matching between the information query record and the candidate query record Find out the similarity between the candidate object and the target object. If the query content between the information query record of the target object and the candidate query record of the candidate object is more similar, it means that the interest and/or learning stage between the two users are closer, and the similarity between the two is higher.
  • the target object and the candidate object may be on the same application platform.
  • Each candidate object is configured with a corresponding object database, and records of information that the candidate object has queried are stored in the object database associated with the candidate object, so the object database stores candidate query records of the candidate object.
  • the candidate query record is the same as the information query record. Depending on the type of information, it can record information that the candidate has queried, such as queried words or browsed articles.
  • the manner of calculating the similarity according to the information query record and the candidate query record may be: the terminal device can respectively detect the number of occurrences of the character contained in each candidate query record in the information query record, and based on the number of occurrences As the similarity factor of the candidate query record, the similarity of the candidate object is calculated based on the similarity factors of all candidate query records.
  • an associated object of the target object is selected from the candidate objects.
  • the terminal device can calculate the similarity of each candidate object according to the order of the similarity values, The candidate objects are sorted, and the related objects of the target object are selected from the candidate objects.
  • the terminal device may be set with a threshold for the number of associations.
  • the terminal device may sort from large to small according to the similarity value, and select the first N candidate objects as associated objects, where N is The above-mentioned associated number threshold; the terminal device may also be set with a similarity threshold, and the terminal device may select a similarity value greater than the similarity threshold as the associated object.
  • the number of associated objects is not fixed.
  • a recommended information list matching the target object is generated based on the associated information database of the associated object; the associated information database contains historical query information of the associated object.
  • each object has a corresponding information library, and the information library is used to record the information that the object focuses on, stored information, and/or browsed information. Therefore, after determining the associated object of the target object, the associated information base of the object identifier can be determined from the database according to the object identifier of the associated object.
  • the associated information base is used to store all historical query information of the associated object, so that it can be based on Historical query information, extract the concerned information and stored information of the associated object.
  • the terminal device may calculate the recommendation priority of the historical query information of the associated object in the associated information database, determine the recommended information from the historical query information based on the recommended priority, and generate recommended information from all the recommended information List.
  • the method of calculating the recommendation priority may be: the terminal device can identify the information type of each historical query information, and select the information type that contains the largest number of historical query information as the main information type, and Set the recommended priority of all primary information types to the highest level, and set the recommended priority of all secondary information types as secondary information types, and set the recommended priority of all secondary information types to secondary information types, and so on, according to each history The level of the information type to which the query information belongs, and the recommended priority of the historical query information is determined.
  • the recommended information lists of all the associated objects are merged to generate the to-be-pushed information list of the target object.
  • the terminal device can extract the recommended information contained in the recommended information list of each associated object, and rearrange all the recommended information to obtain a list of information to be pushed about the target object.
  • the terminal device may perform a push operation to the target object according to the recommendation information contained in the to-be-push list.
  • the terminal device can determine the recommendation order according to the recommendation priority of each recommendation information, and sequentially push information to the target object based on the recommendation order.
  • the recommendation priority can refer to the calculation method of S104, which will not be repeated here; or, the recommendation priority can be determined by the similarity of its corresponding associated object, if the corresponding associated object is more similar to the target object , The higher the push priority of the recommendation information of the associated object.
  • the personal center network of the target user can be constructed based on the social network, and the similarity with the reference user can be calculated through the word database; when the similarity is greater than the similarity threshold, the closeness between the target user and the reference user can be calculated by the number of contacts;
  • the word score is determined according to the similarity and intimacy, and then words with a score greater than the score threshold are recommended to the target user.
  • the classification conditions of words are not clear, and the boundaries between word categories are blurred. According to the classification of words and the scores of word classifications, the similarity of points of interest between users is judged, and the personal center network is constructed.
  • the embodiment of the present invention provides an information matching and recommendation method based on group intelligence perception, introduces the concept of group intelligence perception, collects the information review history of all users, and finds out the similarity to the target user's interest and the degree of overlap in the learning stage according to the information review history A large number of associated users, so as to match and recommend information that the target user is really interested in.
  • the embodiment of the present invention can automatically match the learning stage of the target user, and automatically match and recommend words belonging to the current learning stage when the target user's learning stage is detected to change, which is more suitable for users to acquire long-term information.
  • the information matching and recommendation method based on group intelligence perception proposed in the embodiment of the present invention accurately matches the user’s interest in information, and at the same time allows the information recommendation to intelligently match the learning stage the user is in to achieve personalized customization of information push and optimize users The learning experience in turn improves learning efficiency.
  • the information pushing method obtains the query record of the target object that needs to perform information push, calculates the similarity between the candidate object and the target object according to the query record, and determines the similarity according to the similarity.
  • Associated objects that have a strong association with the target object means that the browsing habits and information levels of the associated object are similar to the target object. Therefore, a recommended information list can be generated according to the associated information database of the associated object, and recommended by multiple associated objects The information list generates a list of information to be pushed for the target object, and pushes the information contained in the list to be pushed to the target object.
  • the present invention does not only rely on the query record of the target object to determine the push information, but also determines the associated object with higher object attribute similarity to the target object based on the query record of the target object.
  • the recommended information list is generated according to the associated information database of the associated object. Since different associated objects browse more information fields and the similarity between the associated object and the target object is relatively high, it can be inferred that the information field of the associated object is also very interesting. It is likely to be the information field that the target object is interested in. It achieves the purpose of automatically expanding the information field of the push information. While ensuring accurate push, it can also increase the scope of the information field and improve the information push efficiency of the target object.
  • FIG. 2 shows a specific implementation flowchart of an information push method S102 provided by the second embodiment of the present invention.
  • an information push method S102 provided in this embodiment includes: S1021 to S1026, which are detailed as follows:
  • the calculating the similarity between the candidate object and the target object according to the information query record and the candidate query record of the candidate object includes:
  • S1021 extract multiple pieces of first information of the target object from the information query record, and extract multiple pieces of second information of the candidate object from the candidate query record.
  • the terminal device when the terminal device has obtained the information query record of the target object, it can extract the information that the target object has queried from each information query record, that is, the above-mentioned first information. By performing the above operations, multiple pieces of first information can be obtained.
  • an information query record can include one piece of first information, or two or more pieces of first information.
  • the target user when the target user writes query keywords, he can use the default query Or query the information in a batch query, so that one information query record can output multiple query results, so that one query record can correspond to multiple first information.
  • the terminal device may obtain the article browsed by the target object, perform semantic analysis on the article, and extract the article Corresponding core keywords, for example, count the occurrence frequency of each candidate keyword in the article, select the core keyword based on the occurrence frequency, and identify each core keyword as the first information corresponding to the information query record.
  • the specific operation process of extracting the second information from the candidate query record is the same as the step of extracting the first information. You can refer to the process of extracting the first information, which will not be repeated here.
  • each of the first information word vector conversion models are respectively generated to generate a first word vector corresponding to the first information, and each of the second information is imported into the word vector conversion model to generate the The second word vector corresponding to the second information.
  • the terminal device may store a word vector conversion model, which can convert the extracted information into a word vector expressed in the same dimension, so that the distance between the two information can be determined by the distance between the word vectors. The degree of relevance. Based on this, the terminal device can import the first information into the word vector conversion model, and calculate the first word vector corresponding to the first information; and import the second information into the word vector conversion model, and calculate the second information corresponding to the second information. Word vector.
  • the terminal device may determine the dimension value of the first information in each part of speech dimension according to multiple preset parts of speech dimensions, so as to construct the first information of the first information according to the dimension value of each part of speech dimension.
  • a word vector where the number of elements contained in the first word vector is consistent with the number of parts of speech dimensions.
  • the word vector conversion operation can also be performed in the above manner.
  • the terminal device can import the first information and the second information into the trained Word2Vec semantic model through the Word2Vec semantic model to obtain the first word vector and the second word vector.
  • the first word vector and the second word vector are imported into a preset first similarity factor calculation model, and a first similarity factor between the candidate object and the target object is calculated.
  • the terminal device can use the first word vector and the second word vector.
  • the vector distance between vectors determines the degree of association between different information. If the distance between two vectors is smaller, the degree of association between the two information is greater; on the contrary, if the distance between two vectors is greater, it means two The smaller the correlation between the information.
  • the terminal device may calculate the vector distance between each first word vector and the second word vector, and perform a weighted summation according to the vector distances, and use the weighted summation value as a candidate object The first similarity factor with the target object.
  • the target attribute of the target object is obtained, and the candidate attribute of the candidate object is obtained.
  • the terminal device can extract the target attribute of the target object from the object database according to the object identification of the target object.
  • the target attribute can include the age, gender, address, education background, education background, etc. and the target of the target object.
  • Object-related information can be extracted from the object database through the object identification of the candidate object.
  • a second similarity factor between the target object and the candidate object is calculated according to the target attribute and the candidate attribute.
  • the target attribute can determine the field of interest and/or the learning stage of the target object. Therefore, if the similarity between the target attribute and the candidate attribute is greater, it indicates the association between the candidate object and the target object. The greater the degree, based on this, the second similarity factor between the two objects can be calculated according to the target attribute and the candidate attribute.
  • the terminal device may count the number of attribute items with the same parameter value in the target attribute and candidate attributes, and import the number of identical attribute items into a preset hash conversion function to calculate the The hash value corresponding to the number of the same attribute items is identified as the second similarity factor between the target object and the candidate object. If the number of the attribute item is larger, the corresponding second similarity factor is larger; on the contrary, if the number of the attribute item is smaller, the corresponding second similarity factor is smaller.
  • the first similarity factor and the second similarity factor are imported into a preset similarity calculation model to determine the similarity between the target object and the candidate object; the similarity calculation
  • the specific model is:
  • Sim (A, X) is the similarity between the candidate object and the target control
  • Sim IC (A, X) is the first similarity factor
  • Sim SC (A, X) is the result The second similarity factor
  • W IC and W SC are preset coefficients.
  • the terminal device imports the calculated first similarity factor and the second similarity factor into the similarity calculation model, and performs a weighting calculation on the two similarity factors through the weighting weights corresponding to the first similarity factor and the second similarity factor. Sum, the summed value is regarded as the similarity between the target object and the candidate object.
  • the candidate objects include not only the objects recognized by the target object but also unfamiliar objects, a larger data volume can be increased and the accuracy of matching and recommendation can be improved. Furthermore, additional user information is introduced. , That is, the above-mentioned target attributes and candidate attributes, which represent the characteristics of the learning stage. Based on the original browsing records, for different learning stages, matching and recommending information that meets both the interest and the learning level improves the accuracy of the push information .
  • FIG. 3 shows a specific implementation flowchart of an information push method S1023 provided by the third embodiment of the present invention.
  • an information push method S1023 provided in this embodiment includes: S301 to S302, which are detailed as follows:
  • the importing the first word vector and the second word vector into a preset first similarity factor calculation model, and calculating the first similarity factor between the candidate object and the target object includes :
  • each of the second word vectors and the first word vectors are respectively imported into a preset similarity distance calculation model to determine the similarity distance of the second word vector;
  • the similarity distance calculation model is specifically:
  • v 2i is the similarity distance of the i-th second word vector; Is the i-th second word vector; Is the j-th first word vector; L 1 is the total number of the first word vector; L 2 is the total number of the second word vector; Max ⁇ x ⁇ is the maximum value selection function .
  • the terminal device can calculate the similarity distance between the two vectors through the cosine similarity, and select the first word vector with the shortest distance from the i-th second word vector from the first word vector, The cosine similarity between the closest first word vector and the second word vector is taken as the similarity distance of the second word vector.
  • the terminal device can generate an information dictionary based on the first information and the second information, and the information dictionary is configured with a corresponding information number for each of the first information and the second information, and the terminal device can be based on The information number configures a distance vector of length L, that is, the aforementioned v 2 , where L is the total number of information contained in the information dictionary.
  • L is the preset word vector length.
  • the terminal device can calculate the first similarity factor between the candidate object and the target object according to the similarity distance. Since the distance between the target object and itself is 0, an array of all 1s can be used to identify the similar distance between the target object and itself.
  • the terminal device encapsulates the similar distances of all second word vectors to form an L-dimensional distance vector about the candidate object, and calculates the vector distance between the two vectors based on the distance vector of the candidate object and the native vector of the target object.
  • the vector distance may also be calculated by using the cosine similarity, so that the calculated result is used as the first similarity factor between the target object and the candidate object.
  • the query record between the target object and the candidate object is calculated.
  • the above similarity improves the accuracy of calculation of the first similarity factor.
  • FIG. 4 shows a specific implementation flowchart of an information push method S1025 provided by the fourth embodiment of the present invention.
  • an information push method S1042 provided in this embodiment includes: S401 to S403, which are detailed as follows:
  • the target attribute includes the target age and the target address of the target object
  • the candidate attribute includes the candidate age and the candidate address of the candidate object
  • the calculation of the target attribute and the candidate attribute includes:
  • a first distance value between any two candidate addresses is calculated, and a second distance value between any one of the candidate addresses and the target address is calculated.
  • the target attribute includes the target address where the target object is located, and the candidate attribute of each candidate object also includes the candidate address where the candidate object is located.
  • the terminal device can calculate the target address and all candidate addresses for pairwise combination, thereby calculating the first distance value between any two candidate addresses and the second distance value between any candidate address and the target address .
  • a maximum distance value is selected from all the first distance values and the second distance value, and the maximum distance value is recognized as a distance reference value.
  • the terminal device can select a distance value with the largest distance value according to the first distance value and the second distance value, and identify the distance value with the largest value as the distance reference value, so that the maximum geographic location distance can be determined.
  • age A is the target age
  • age X is the candidate age
  • Add A is the target address
  • Add X is the candidate address
  • Max_Range is the distance reference value
  • ⁇ and ⁇ are preset coefficients
  • Min (age A ,age X ) is the minimum value selection function
  • Range (Add A ,Add X ) is the distance calculation function
  • the learning stage can be indirectly expressed from the age and geographic location of the user.
  • the similarity in the learning stage between two objects is determined by age and geographic location, thereby calculating the second similarity factor, which can improve the accuracy of subsequent calculation of the similarity between two users.
  • FIG. 5 shows a specific implementation flowchart of an information push method S104 provided by the fifth embodiment of the present invention.
  • an information push method S104 provided in this embodiment includes: S1041 to S1044, which are detailed as follows:
  • the related information database based on the related object generates a recommended information list matching the target object;
  • the related information database contains historical query information of the related object, including:
  • the historical query information whose query time is within a preset effective time range is selected as the effective query information.
  • the terminal device after determining the associated object, the terminal device needs to determine the recommended information pushed to the target object based on the associated object. Based on this, the terminal device can detect the query time of each historical query information of the associated object, and select the pre- The historical query information within the effective time range is set as the effective query information.
  • the effective time range can be from the current time to the time T, so that the historical query information recently browsed by the associated object can be obtained, so that the most interesting information of the associated object can be determined, and the timeliness of the recommended information is improved. accuracy.
  • the terminal device extracts the corresponding second information from each valid query information, and counts the number of first occurrences of each second information in all valid query records. If the number of occurrences is greater, it means that the second information The higher the recommendation degree corresponding to the second information, the greater the probability that the target object is interested in the second information.
  • the number of occurrences of the second information and the second word vector are imported into a recommendation degree calculation model to calculate the recommendation degree of the second information;
  • the recommendation degree calculation model is specifically:
  • the terminal device can convert the second information into the second word vector.
  • the specific conversion method into the second word vector refer to the related description of S1022, which will not be repeated here.
  • the above two parameters can be imported into the recommendation degree calculation model to calculate the recommendation degree of each second information.
  • the greater the number of occurrences the greater the corresponding recommendation degree; if the similarity between the second word vector of the second information and the first word vector of the target object is greater, the recommendation degree of the second information is greater Big.
  • the recommended information list is generated.
  • the terminal device may sort each second information from large to small according to the value of the recommendation degree of each second information, and generate a recommended information list corresponding to the associated object.
  • the terminal device may be set with a maximum number of recommendations.
  • the terminal device sorts each second information from largest to smallest according to the value of the recommendation degree, and selects the first N second information as the recommended information , Where N is the maximum recommended number of positions above.
  • the terminal device may also be set with a recommendation threshold, identify all second information with a recommendation degree greater than the recommendation threshold as recommendation information, and generate a recommendation information list.
  • the terminal device may be set with a maximum number of recommendations and a recommendation threshold, select second information with a recommendation degree greater than the recommendation threshold as candidate information, and select the first N candidate information as recommended information, and generate a recommended information list.
  • the recommended information that needs to be pushed is selected, and the recommended information list is generated, which improves the accuracy of the recommended information and the relevance to the target user, and achieves accuracy The purpose of the information push.
  • FIG. 6 shows a specific implementation flowchart of a method S105 for pushing information provided by the sixth embodiment of the present invention.
  • an information push method S105 provided in this embodiment includes: S1051 to S1053, which are detailed as follows:
  • the merging the recommended information lists of all the associated objects to generate the to-be-pushed information list of the target object includes:
  • the terminal device can separately count the second appearance times of each recommended information in all recommended information lists according to the recommended information contained in each recommended information list. If the number of appearances is greater, it means that the recommended information corresponds to The higher the priority of the recommendation, the greater the probability that the target object is interested in the recommendation information.
  • the push priority of each of the recommended information is determined according to the second number of occurrences.
  • the terminal device may separately calculate the push priority of each recommended information according to the second number of appearances, for example, directly use the second number of appearances as the push priority of the recommended information.
  • each recommendation information may be converted into a recommendation word vector, and based on the recommendation word vector and the second number of occurrences, it may be imported into the recommendation degree calculation model to calculate the push priority of each recommendation information.
  • the list of information to be pushed is generated based on the push priority.
  • the terminal device sorts all recommended information from large to small according to the value of the push priority, thereby generating a list of information to be pushed about the target object, and according to the push order in the list of information to be pushed, Push information to the target object.
  • the push priority of each recommended information is determined, and a list of information to be pushed with the push priority order is generated, which can push the user in an orderly manner , Improve the accuracy of push operation.
  • FIG. 7 shows a structural block diagram of an information pushing device provided by an embodiment of the present invention, and each unit included in the information pushing device is used to execute each step in the embodiment corresponding to FIG. 1.
  • each unit included in the information pushing device is used to execute each step in the embodiment corresponding to FIG. 1.
  • only the parts related to this embodiment are shown.
  • the information pushing equipment includes:
  • the information query record obtaining unit 71 is configured to obtain the information query record of the target object
  • the similarity calculation unit 72 is configured to calculate the similarity between the candidate object and the target object according to the information query record and the candidate query record of the candidate object;
  • An associated object selection unit 73 configured to select an associated object of the target object from the candidate objects based on the similarity corresponding to each candidate object;
  • the recommended information list generating unit 74 is configured to generate a recommended information list that matches the target object based on the associated information database of the associated object; the associated information database contains historical query information of the associated object;
  • the information to be pushed list generating unit 75 is configured to merge the recommended information lists of all the associated objects to generate a list of information to be pushed of the target object.
  • the similarity calculation unit 72 includes:
  • An information extraction unit configured to extract multiple pieces of first information about the target object from the information query record, and extract multiple pieces of second information about the candidate object from the candidate query record;
  • the word vector conversion unit is configured to convert each of the first information word vector models to generate a first word vector corresponding to the first information, and to import each of the second information into the word vector conversion model, Generating a second word vector corresponding to the second information;
  • the first similarity factor calculation unit is configured to import the first word vector and the second word vector into a preset first similarity factor calculation model, and calculate the first similarity factor between the candidate object and the target object. Similarity factor
  • the object attribute obtaining unit is configured to obtain the target attribute of the target object and obtain the candidate attribute of the candidate object;
  • a second similarity factor calculation unit configured to calculate a second similarity factor between the target object and the candidate object according to the target attribute and the candidate attribute;
  • a similarity calculation unit configured to import the first similarity factor and the second similarity factor into a preset similarity calculation model to determine the similarity between the target object and the candidate object;
  • the similarity calculation model is specifically:
  • Sim (A, X) is the similarity between the candidate object and the target control
  • Sim IC (A, X) is the first similarity factor
  • Sim SC (A, X) is the result The second similarity factor
  • W IC and W SC are preset coefficients.
  • the first similarity factor calculation unit includes:
  • a similarity distance calculation unit configured to respectively import each of the second word vectors and the first word vector into a preset similarity distance calculation model to determine the similarity distance of the second word vector; the similarity distance calculation model Specifically:
  • v 2i is the similarity distance of the i-th second word vector; Is the i-th second word vector; Is the j-th first word vector; L 1 is the total number of the first word vector; L 2 is the total number of the second word vector; Max ⁇ x ⁇ is the maximum value selection function ;
  • the similarity distance introduction unit is configured to calculate the first similarity factor between the candidate object and the target object according to the similarity distance of each of the second word vectors; the model for calculating the similarity is specifically :
  • L is the preset word vector length.
  • the target attribute includes a target age and a target address of the target object
  • the candidate attribute includes a candidate age and a candidate address of the candidate object
  • the second similarity factor calculation unit includes:
  • a distance value calculation unit configured to calculate a first distance value between any two candidate addresses, and calculate a second distance value between any one of the candidate addresses and the target address;
  • a distance reference value calculation unit configured to select a maximum distance value from all the first distance values and the second distance value, and identify the maximum distance value as the distance reference value;
  • the user information importing unit is configured to import the candidate address of the candidate object, the candidate age, the target address of the target object, and the target age into a second similarity factor calculation model, and calculate the relationship between the candidate object and the The second similarity factor between target objects; the second similarity factor calculation model is specifically:
  • age A is the target age
  • age X is the candidate age
  • Add A is the target address
  • Add X is the candidate address
  • Max_Range is the distance reference value
  • ⁇ and ⁇ are preset coefficients
  • Min (age A , age X ) is the minimum value selection function
  • Range (Add A , Add X ) is the distance calculation function.
  • the recommended information list generating unit 74 includes:
  • An effective information selecting unit configured to select the historical query information whose query time is within a preset effective time range as effective query information based on the query time of each of the historical query information;
  • the first occurrence counting unit is configured to count the first occurrences of each second information based on the second information corresponding to the valid query information;
  • the recommendation degree calculation unit is configured to import the number of occurrences of the second information and the second word vector into a recommendation degree calculation model to calculate the recommendation degree of the second information;
  • the recommendation degree calculation model is specifically:
  • the recommendation information selection unit is configured to generate the recommendation information list based on the recommendation degree of each of the second information.
  • the information to be pushed list generating unit 75 includes:
  • a second appearance count counting unit configured to count the second appearance times of each of the recommended information in all the recommended information lists based on the recommended information contained in each of the recommended information lists;
  • a push priority calculation unit configured to determine the push priority of each of the recommended information according to the second number of occurrences
  • the push information selection unit is configured to generate the list of information to be pushed based on the push priority.
  • the information pushing device also does not only rely on the query record of the target object to determine the pushed information, but also determines the associated object with a higher degree of similarity to the object attribute of the target object based on the query record of the target object. , And generate a recommended information list based on the associated information database of the associated object. Since different associated objects browse more information fields, and the associated object has a high similarity to the target object, it can be inferred that the associated object’s interested information field is also It is most likely that it is the information field that the target object is interested in. It achieves the purpose of automatically expanding the information field of the push information. While ensuring accurate push, it can also increase the scope of the information field and improve the information push efficiency of the target object.
  • Fig. 8 is a schematic diagram of a terminal device according to another embodiment of the present invention.
  • the terminal device 8 of this embodiment includes a processor 80, a memory 81, and a computer program 82 stored in the memory 81 and running on the processor 80, such as an information push program.
  • the processor 80 executes the computer program 82, the steps in the foregoing method embodiments for pushing information are implemented, such as S101 to S105 shown in FIG. 1.
  • the processor 80 executes the computer program 82
  • the functions of the units in the foregoing device embodiments such as the functions of the modules 71 to 75 shown in FIG. 7, are realized.
  • the computer program 82 may be divided into one or more units, and the one or more units are stored in the memory 81 and executed by the processor 80 to complete the present invention.
  • the one or more units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 82 in the terminal device 8.
  • the computer program 82 may be divided into an information query record acquisition unit, a similarity calculation unit, an associated object selection unit, a recommendation information list generation unit, and a to-be-push information list generation unit, and the specific functions of each unit are as described above.
  • the terminal device 8 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • the terminal device may include, but is not limited to, a processor 80 and a memory 81.
  • FIG. 8 is only an example of the terminal device 8 and does not constitute a limitation on the terminal device 8. It may include more or less components than shown in the figure, or a combination of certain components, or different components.
  • the terminal device may also include input and output devices, network access devices, buses, etc.
  • the so-called processor 80 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (ASIC), Ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the memory 81 may be an internal storage unit of the terminal device 8, such as a hard disk or memory of the terminal device 8.
  • the memory 81 may also be an external storage device of the terminal device 8, such as a plug-in hard disk equipped on the terminal device 8, a smart memory card (Smart Media Card, SMC), or a Secure Digital (SD) Card, Flash Card, etc. Further, the memory 81 may also include both an internal storage unit of the terminal device 8 and an external storage device.
  • the memory 81 is used to store the computer program and other programs and data required by the terminal device.
  • the memory 81 can also be used to temporarily store data that has been output or will be output.
  • the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种信息推送的方法及设备，应用于数据处理技术领域，包括：获取目标对象的信息查询记录（S101）；根据查询记录以及候选对象的候选查询记录，计算候选对象与目标对象之间的相似度（S102）；基于各个候选对象对应的相似度，从候选对象中选取目标对象的关联对象（S103）；基于关联对象的关联信息库，生成与目标对象匹配的推荐信息列表；关联信息库包含有关联对象的历史查询信息（S104）；对所有关联对象的推荐信息列表进行合并，生成目标对象的待推送信息列表（S105）。该方案根据目标对象的查询记录确定与目标对象的对象属性相似度较高的关联对象，生成推荐信息列表，能够自动扩展推送信息的信息领域，在保证精准推送的同时，扩大信息领域的范围，提高目标对象的信息推送效率。

Description

一种信息推送的方法及设备 技术领域
本发明属于数据处理技术领域，尤其涉及一种信息推送的方法及设备。
背景技术
随着信息更新速度的不断增加，信息数量正以几何级数的速度递增，如何能够精准地为用户推送信息，直接影响用户获取信息的效率。例如，在用户学习新语言，需要对该语言的单词进行记忆时，如何能够根据用户的兴趣以及学习阶段推送相关联的单词，则会对用户的学习效果造成直接影响。由此可见，现亟需一种对信息进行精准推送的方法。
现有的信息推送技术，主要是根据用户的对于信息的查询记录，推荐与查询记录相关联的关联信息给用户，通过此方式推送的信息由于与查询记录之间存在强关联的关系，较难主动扩展所推送信息的领域，从而降低了推送信息的多样性以及丰富性，降低了用户获取信息的效率。
技术问题
有鉴于此，本发明实施例提供了一种信息推送的方法及设备，以解决现有的信息推送技术，主要是根据用户的对于信息的查询记录，推荐与查询记录相关联的关联信息给用户，通过此方式推送的信息由于与查询记录之间存在强关联的关系，较难主动扩展所推送信息的领域，从而降低了推送信息的多样性以及丰富性，降低了用户获取信息的效率的问题。
技术解决方案
本发明实施例的第一方面提供了一种信息推送的方法，包括：
获取目标对象的信息查询记录；
根据所述信息查询记录以及候选对象的候选查询记录，计算所述候选对象与所述目标对象之间的相似度；
基于各个所述候选对象对应的所述相似度，从所述候选对象中选取所述目标对象的关联对象；
基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息；
对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表。
本发明实施例的第二方面提供了一种信息推送的设备，包括：
信息查询记录获取单元，用于获取目标对象的信息查询记录；
相似度计算单元，用于根据所述信息查询记录以及候选对象的候选查询记录，计算所 述候选对象与所述目标对象之间的相似度；
关联对象选取单元，用于基于各个所述候选对象对应的所述相似度，从所述候选对象中选取所述目标对象的关联对象；
推荐信息列表生成单元，用于基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息；
待推送信息列表生成单元，用于对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表。
本发明实施例的第三方面提供了一种终端设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现第一方面的各个步骤。
本发明实施例的第四方面提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现第一方面的各个步骤。
有益效果
实施本发明实施例提供的一种信息推送的方法及设备具有以下有益效果：
本发明实施例通过获取需要进行信息推送的目标对象的查询记录，根据查询记录计算候选对象与目标对象之间的相似度，并根据相似度确定与目标对象存在强关联关系的关联对象，即表示关联对象的浏览习惯以及信息层级与目标对象较为相似，因此可以根据关联对象的关联信息库，生成推荐信息列表，并通过多个关联对象的推荐信息列表，生成目标对象的待推送信息列表，向目标对象推送待推送列表内包含的信息。与现有的信息推送技术相比，本发明不只是仅仅依靠目标对象的查询记录进行推送信息的确定，而是根据目标对象的查询记录确定与目标对象的对象属性相似度较高的关联对象，并根据关联对象的关联信息库生成推荐信息列表，由于不同的关联对象所浏览的信息领域较多，且关联对象与目标对象的相似度较高，从而可以推断关联对象感兴趣的信息领域也很大可能是目标对象所感兴趣的信息领域，实现了自动扩展推送信息的信息领域的目的，在保证精准推送的同时，也能够提高信息领域的范围，提高目标对象的信息推送效率。
附图说明
为了更清楚地说明本发明实施例中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本发明第一实施例提供的一种信息推送的方法的实现流程图；
图2是本发明第二实施例提供的一种信息推送的方法S102具体实现流程图；
图3是本发明第三实施例提供的一种信息推送的方法S1023具体实现流程图；
图4是本发明第四实施例提供的一种信息推送的方法S1025具体实现流程图；
图5是本发明第五实施例提供的一种信息推送的方法S104具体实现流程图；
图6是本发明第六实施例提供的一种信息推送的方法S105具体实现流程图；
图7是本发明一实施例提供的一种信息推送的设备的结构框图；
图8是本发明另一实施例提供的一种终端设备的示意图。
本发明的实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
本发明实施例通过获取需要进行信息推送的目标对象的查询记录，根据查询记录计算候选对象与目标对象之间的相似度，并根据相似度确定与目标对象存在强关联关系的关联对象，即表示关联对象的浏览习惯以及信息层级与目标对象较为相似，因此可以根据关联对象的关联信息库，生成推荐信息列表，并通过多个关联对象的推荐信息列表，生成目标对象的待推送信息列表，向目标对象推送待推送列表内包含的信息，解决了现有的信息推送技术，主要是根据用户的对于信息的查询记录，推荐与查询记录相关联的关联信息给用户，通过此方式推送的信息由于与查询记录之间存在强关联的关系，较难主动扩展所推送信息的领域，从而降低了推送信息的多样性以及丰富性，降低了用户获取信息的效率的问题。
在本发明实施例中，流程的执行主体为终端设备。该终端设备包括但不限于：服务器、计算机、智能手机以及平板电脑等能够执行信息推送的操作的设备。图1示出了本发明第一实施例提供的信息推送的方法的实现流程图，详述如下：
在S101中，获取目标对象的信息查询记录。
在本实施例中，终端设备可以将需要进行推送的用户识别为目标对象。若终端设备为某一用户的用户终端，则用户终端可以将登陆于该用户终端的用户识别为目标用户，并为用户终端所属的用户生成与之对应的待推送信息列表。若终端设备为一服务器，服务器可以接收用户的推送请求，并根据推送请求中携带有用户标识，将该用户标识对应的请求用户识别为目标用户，并向该目标用户进行信息推送；当然，服务器还可以为全平台的用户对象自动进行信息推送，在该情况下，终端设备可以创建多条并行线程，通过多条并行线程同时生产多个目标对象的待推送信息列表，优选地，终端设备会为每个用户账户添加一 个创建标识符，在创建了某一用户账户的待推送信息列表后，会将该用户账户的创建标识符设置为第一位值，在下一次选取目标对象时，可以根据创建标识符的位值，识别出未创建的用户账户，并将其识别为目标对象。
可选地，在本实施例中，该待推送列表具有有效时长，终端设备在创建了一个目标对象的待推送信息列表后，会创建有个有效计时器，当该有效计时器的计数值等于该有效时长时，则会识别该待推送列表未无效推送列表，并重新为该目标对象生成待推送信息列表。当然，终端设备还可以设置多个列表无效条件，例如，当检测到待推送信息列表中每个推荐信息的推送次数大于预设的推送次数阈值，则识别该待推送列表为无效推送列表；若检测到用户新建的查询记录的个数大于预设的更新个数阈值，则识别该待推送列表为无效推送列表。
在本实施例中，终端设备可以为目标对象创建一个对象数据库，该对象数据库可以用于存储该目标对象的所有操作记录，该操作记录包括有信息查询记录，信息查询记录具体用于记录目标对象查询过的信息。例如，终端设备可以应用于语言学习领域，在该情况下，该信息查询记录可以用于记录目标对象查询过的单词，并将所有查询过的单词对应的信息查询记录存储于该目标对象的对象数据库内；终端设备还可以应用于新闻推荐领域，在该情况下，该信息查询记录可以用于记录目标对象浏览过的新闻文章，并将记录有查询过的文章标识的信息查询记录存储于该目标对象的对象数据库内。
在S102中，根据所述信息查询记录以及候选对象的候选查询记录，计算所述候选对象与所述目标对象之间的相似度。
在本实施例中，终端设备在获取了目标对象的信息查询记录后，还可以获取除目标对象外的所有候选对象的查询记录，并根据信息查询记录与候选查询记录之间的匹配度，计算出候选对象与目标对象之间的相似度。若目标对象的信息查询记录与候选对象的候选查询记录之间的查询内容越相似，则表示两个用户之间的兴趣和/或学习阶段越接近，从而两者之间的相似度越高。
在本实施例中，目标对象与候选对象可以处于同一个应用平台。各个候选对象配置有对应的对象数据库，将候选对象查询过信息的记录存储于该候选对象关联的对象数据库内，因此对象数据库存储有候选对象的候选查询记录。该候选查询记录与信息查询记录一样，根据信息类型的不同，可以记录有候选对象查询过的信息，例如查询过的单词或浏览过的文章等。
可选地，根据信息查询记录以及候选查询记录计算相似度的方式可以为：终端设备可以分别检测每个候选查询记录中包含的字符在信息查询记录中的出现的次数，并基于该出 现的次数作为该候选查询记录的相似因子，基于所有候选查询记录的相似因子，计算出该候选对象的相似度。
在S103中，基于各个所述候选对象对应的所述相似度，从所述候选对象中选取所述目标对象的关联对象。
在本实施例中，由于相似度可以表示候选对象与目标对象之间的学习领域以及学习阶段越接近，因此终端设备在计算了各个候选对象的相似度后，可以根据相似度数值的大小次序，对各个候选对象进行排序，从而从候选对象中选取出目标对象的关联对象。
在本实施例中，终端设备可以设置有关联个数阈值，在该情况下，终端设备可以根据相似度的数值从大到小进行排序，并选取前N个候选对象作为关联对象，其中N为上述的关联个数阈值；终端设备还可以设置有相似度阈值，终端设备可以选取相似度的数值大于该相似度阈值为关联对象，在该情况下，即关联对象的个数并非固定的。
在S104中，基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息。
在本实施例中，每个对象会有对应的信息库，该信息库用于记录该对象关注的信息、存储的信息和/或浏览过的信息等。因此，在确定了目标对象的关联对象后，可以根据关联对象的对象标识，从数据库中确定该对象标识的关联信息库，该关联信息库用于存储关联对象的所有历史查询信息，从而可以根据历史查询信息，提取关联对象的所关注的信息、所存储的信息等。
在本实施例中，终端设备可以计算该关联信息库中关联对象的历史查询信息的推荐优先级，并基于该推荐优先级从历史查询信息中确定出推荐信息，并将所有推荐信息生成推荐信息列表。
可选地，在本实施例中，计算推荐优先级的方式可以为：终端设备可以识别各个历史查询信息的信息类型，并选取包含的历史查询信息个数最多的信息类型作为主信息类型，并将所有主信息类型的推荐优先级设置为最高级，并将信息个数次多的信息类型作为次信息类型，将所有次信息类型的推荐优先级设置为次高级，如此类推，可以根据各个历史查询信息所属信息类型的层级，确定该历史查询信息的推荐优先级。
在S105中，对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表。
在本实施例中，终端设备可以提取各个关联对象的推荐信息列表包含的推荐信息，并对所有推荐信息进行重新排列，得到关于目标对象的待推送信息列表。终端设备可以根据该待推送列表中包含的推荐信息，向目标对象进行推送操作。其中，终端设备可以根据各 个推荐信息的推荐优先级确定推荐次序，并基于推荐次序依次对目标对象进行信息推送操作。其中，推荐优先级可以参考S104的计算方式，在此不再赘述；或者，该推荐优先级可以与其对应的关联对象的相似度的大小确定，若对应的关联对象与目标对象的相似度越高，则该关联对象的推荐信息的推送优先级越高。
现有技术中，可以基于社会网络构建目标用户的个人中心网，通过单词库计算与参考用户的相似度；当相似性大于相似度阈值时，通过联系数量计算目标用户与参考用户间的亲密；根据相似度与亲密度确定单词评分，然后将评分大于评分阈值的单词推荐给目标用户。而该方式具有以下缺陷：
(1)仅以社会网络上的好友作为单词匹配的参考，忽略了潜在的与用户相似的陌生人，局限了单词匹配的参考对象，限制了对单词匹配准确率的提升；
(2)只是关注用户的兴趣将忽视用户其处于不同学习阶段的事实，匹配和推荐的单词可能不属于其学习阶段，对用户来说可能出现推荐的单词太容易或者太难的情况；
(3)根据单词的查阅记录和推荐记录计算单词评分，相对仅用查阅记录的话评分会出现虚高，影响匹配的准确率，查阅记录才是用户对单词推荐结果、对单词是否感兴趣的直接反馈；
(4)单词的分类条件不明确，单词类别间界限模糊，根据单词的分类、单词分类的评分进行判断用户间兴趣点的相似度，构建个人中心网，存在一定的误差；
本发明实施例提供了一种基于群智感知的信息匹配和推荐方法，引入群智感知的概念，收集所有用户的信息查阅历史并根据信息查阅历史找出与目标用户兴趣相似、学习阶段重合度大的关联用户，从而匹配和推荐目标用户真正感兴趣的信息。其次，本发明实施例可以自动匹配目标用户的学习阶段，在察觉目标用户的学习阶段发生改变时，自动地为其匹配和推荐属于当前学习阶段的单词，更适合用户进行长期的信息获取。本发明实施例中所提出的基于群智感知的信息匹配和推荐方法，准确匹配用户感兴趣信息，同时让信息推荐智能地匹配用户所处的学习阶段，实现信息推送的个性化定制、优化用户学习体验进而提升学习效率。
以上可以看出，本发明实施例提供的一种信息推送的方法通过获取需要进行信息推送的目标对象的查询记录，根据查询记录计算候选对象与目标对象之间的相似度，并根据相似度确定与目标对象存在强关联关系的关联对象，即表示关联对象的浏览习惯以及信息层级与目标对象较为相似，因此可以根据关联对象的关联信息库，生成推荐信息列表，并通过多个关联对象的推荐信息列表，生成目标对象的待推送信息列表，向目标对象推送待推送列表内包含的信息。与现有的信息推送技术相比，本发明不只是仅仅依靠目标对象的查 询记录进行推送信息的确定，而是根据目标对象的查询记录确定与目标对象的对象属性相似度较高的关联对象，并根据关联对象的关联信息库生成推荐信息列表，由于不同的关联对象所浏览的信息领域较多，且关联对象与目标对象的相似度较高，从而可以推断关联对象感兴趣的信息领域也很大可能是目标对象所感兴趣的信息领域，实现了自动扩展推送信息的信息领域的目的，在保证精准推送的同时，也能够提高信息领域的范围，提高目标对象的信息推送效率。
图2示出了本发明第二实施例提供的一种信息推送的方法S102的具体实现流程图。参见图1，相对于图1所述实施例，本实施例提供的一种信息推送的方法S102包括：S1021～S1026，具体详述如下：
进一步地，所述根据所述信息查询记录以及候选对象的候选查询记录，计算所述候选对象与所述目标对象之间的相似度，包括：
在S1021中，从所述信息查询记录中提取所述目标对象的多个第一信息，以及从所述候选查询记录中提取所述候选对象的多个第二信息。
在本实施例中，终端设备在获取了目标对象的信息查询记录中，可以从每个信息查询记录中提取得到目标对象查询过的信息，即上述的第一信息，对每个信息查询记录均执行上述操作，即可以得到多个第一信息。需要说明的是，一个信息查询记录中可以包含有一个第一信息，还可以包含有两个或以上的第一信息，在该情况下，目标用户在编写查询关键词时，可以通过缺省查询或批量查询的方式对信息进行查询，从而一条信息查询记录中可以输出多个查询结果，从而一个查询记录可以对应多个第一信息。
可选地，在本实施例中，若信息查询记录的信息类型为文章信息类型，在该情况下，终端设备可以获取目标对象浏览过的文章，并对该文章进行语义分析，提取出该文章对应的核心关键词，例如统计各个候选关键词在文章中的出现词频，并基于该出现词频选取出核心关键词，并将各个核心关键词识别为该信息查询记录对应的第一信息。
在本实施例中，从候选查询记录中提取第二信息的具体操作过程与提取第一信息的步骤相同，可以参考第一信息的提取过程，在此不再赘述。
在1022中，分别将各个所述第一信息词向量转换模型，生成所述第一信息对应的第一词向量，以及将各个所述第二信息导入到所述词向量转换模型，生成所述第二信息对应的第二词向量。
在本实施例中，终端设备可以存储有一个词向量转换模型，可以将提取得到的信息转换为在同一量纲表示的词向量，从而能够通过词向量之间的距离，确定两个信息之间的关联度。基于此，终端设备可以将第一信息导入该词向量转换模型，计算得到第一信息对应 的第一词向量；以及将第二信息导入该词向量转换模型，计算得到第二信息对应的第二词向量。
可选地，在本实施例中，终端设备可以根据多个预设的词性维度，确定第一信息在各个词性维度的维度值，从而根据各个词性维度的维度值构建该第一信息的第一词向量，其中该第一词向量包含的元素个数与词性维度的维度数一致。同样地，对于第二信息也可以通过上述方式进行词向量的转换操作。
可选地，在本实施例中，终端设备可以通过Word2Vec语义模型，将第一信息以及第二信息导入到训练好的Word2Vec语义模型中，得到第一词向量以及第二词向量。
在S1023中，将所述第一词向量以及所述第二词向量导入到预设的第一相似因子计算模型，计算所述候选对象以及所述目标对象之间的第一相似因子。
在本实施例中，由于在S1022中，已经将第一信息以及第二信息转换为通过统一量纲表示的第一词向量以及第二词向量，终端设备可以根据第一词向量以及第二词向量之间的向量距离，确定不同信息之间的关联度，若两个向量距离越小，则表示两个信息之间关联度越大；反之，若两个向量距离越大，则表示两个信息之间的关联度越小。
可选地，在本实施例中，终端设备可以分别计算各个第一词向量与第二词向量之间的向量距离，并根据各个向量距离进行加权求和，将加权求和的数值作为候选对象与目标对象之间的第一相似因子。
在S1024中，获取所述目标对象的目标属性，以及获取所述候选对象的候选属性。
在本实施例中，终端设备可以根据目标对象的对象标识，从对象数据库中提取关于目标对象的目标属性，该目标属性可以包括有目标对象的年龄、性别、住址、教育背景、学历等与目标对象相关的信息。同样地，终端设备也可以通过候选对象的对象标识，从对象数据库中提取关于候选对象的候选属性。
在S1025中，根据所述目标属性以及所述候选属性，计算所述目标对象以及所述候选对象之间的第二相似因子。
在本实施例中，目标属性可以对目标对象所感兴趣的领域和/或学习阶段进行判定，因此若目标属性与候选属性之间的相似度越大，则表示候选对象与目标对象之间的关联度越大，基于此，可以根据目标属性以及候选属性计算两个对象之间的第二相似因子。
可选地，在本实施例中，终端设备可以统计目标属性以及候选属性中参数值相同的属性项的个数，并将相同属性项的个数导入预设的哈希转换函数，计算出该相同属性项个数对应的哈希值，将该哈希值识别为目标对象与候选对象之间的第二相似因子。若该属性项的个数越大，则对应的第二相似因子越大；反之，若该属性的个数越小，则对应的第二相 似因子越小。
在S1026中，将所述第一相似因子以及所述第二相似因子导入预设的相似度计算模型，确定所述目标对象以及所述候选对象之间的所述相似度；所述相似度计算模型具体为：
Sim(A,X)＝W IC·Sim IC(A,X)+W SC·Sim SC(A,X)
其中，Sim(A,X)为所述候选对象与所述目标对照之间的所述相似度；Sim IC(A,X)为所述第一相似因子；Sim SC(A,X)为所述第二相似因子；W IC以及W SC为预设系数。
在本实施例中，终端设备将计算得到第一相似因子以及第二相似因子导入到相似度计算模型，通过第一相似因子以及第二相似因子对应的加权权重，对两个相似因子进行加权求和，将求和后的值作为目标对象与候选对象之间的相似度。
在本发明实施例中，由于候选对象不仅包含目标对象的认识的对象还包括有陌生对象，从而能够更大的数据体量提高了匹配和推荐的准确性，进一步而言，引入额外的用户信息，即上述的目标属性以及候选属性，表示学习阶段特征，在原有的浏览记录的基础上，针对不同的学习阶段，匹配和推荐既符合兴趣又符合学习水平的信息，提高了推送信息的准确性。
图3示出了本发明第三实施例提供的一种信息推送的方法S1023的具体实现流程图。参见图3，相对于图2所述的实施例，本实施例提供的一种信息推送的方法S1023包括：S301～S302，具体详述如下：
进一步地，所述将所述第一词向量以及所述第二词向量导入到预设的第一相似因子计算模型，计算所述候选对象以及所述目标对象之间的第一相似因子，包括：
在S301中，分别将各个所述第二词向量以及所述第一词向量导入到预设的相似距离计算模型，确定所述第二词向量的相似距离；所述相似距离计算模型具体为：
Figure PCTCN2019130897-appb-000001
其中，v 2i为第i个所述第二词向量的所述相似距离；
Figure PCTCN2019130897-appb-000002
为所述第i个所述第二词向量；
Figure PCTCN2019130897-appb-000003
为所述第j个所述第一词向量；L 1为所述第一词向量的总个数；L 2为所述第二词向量的总个数；Max{x}为最大值选取函数。
在本实施例中，终端设备可以通过余弦相似度计算两个向量之间的相似距离，并从第 一词向量中，选取与第i个第二词向量向量距离最短的一个第一词向量，将该距离最近的第一词向量与第二词向量之间的余弦相似度作为该第二词向量的相似距离。
可选地，在本实施例中，终端设备可以根据第一信息以及第二信息生成一个信息字典，该信息字典中为每个第一信息以及第二信息配置对应的信息编号，终端设备可以根据信息编号配置一个长度为L的距离向量，即上述的v 2，其中L为该信息字典包含的信息总数。在该情况下，L 1＝L 2＝L。
在S302中，根据各个所述第二词向量的所述相似距离，计算所述候选对象与所述目标对象之间的所述第一相似因子；计算所述相似度的模型具体为：
Figure PCTCN2019130897-appb-000004
其中，L为预设的词向量长度。
在本实施例中，终端设备在计算了各个第二词向量的相似距离后，可以根据该相似距离计算出候选对象与目标对象之间的第一相似因子。由于目标对象与自身的距离为0，因此可以用一个全为1的数组标识目标对象与自己之间的相似距离。终端设备将所有第二词向量的相似距离进行封装，构成了关于候选对象的一个L维的距离向量，基于候选对象的距离向量以及目标对象的原生向量计算两个向量之间的向量距离，当然，在S302中计算向量距离也可以采用余弦相似度进行计算，从而将计算得到的结果作为目标对象与候选对象之间的第一相似因子。
在本发明实施例中，通过计算出各个第二词向量与最接近的第一词向量之间的相似距离，并构建一个关于候选对象的距离向量，从而计算出目标对象与候选对象在查询记录上的相似度，提高了第一相似因子计算的准确性。
图4示出了本发明第四实施例提供的一种信息推送的方法S1025的具体实现流程图。参见图4，相对于图2所述实施例，本实施例提供的一种信息推送的方法S1042包括：S401～S403，具体详述如下：
进一步地，所述目标属性包括所述目标对象的目标年龄以及目标地址；所述候选属性包括所述候选对象的候选年龄以及候选地址；所述根据所述目标属性以及所述候选属性，计算所述目标对象以及所述候选对象之间的第二相似因子，包括：
在S401中，计算任意两个所述候选地址之间的第一距离值，以及计算任一所述候选地址与所述目标地址之间的第二距离值。
在本实施例中，目标属性包含有目标对象的所在的目标地址，而每个候选对象的候选属性也包含有候选对象所在的候选地址。终端设备可以计算对目标地址以及所有候选地址之间进行两两组合，从而计算得到关于任意两个候选地址之间的第一距离值，以及任一候选地址与目标地址之间的第二距离值。
在S402中，从所有所述第一距离值以及所述第二距离值中选取距离最大值，并将所述距离最大值识别为距离基准值。
在本实施例中，终端设备可以根据第一距离值以及第二距离值，选取距离数值最大的一个距离值，将该数值最大的距离值识别为距离基准值，从而可以确定最大地理位置距离。
在S403中，将所述候选对象的候选地址以及所述候选年龄和所述目标对象的目标地址以及所述目标年龄导入到第二相似因子计算模型，计算所述候选对象与所述目标对象之间的所述第二相似因子；所述第二相似因子计算模型具体为：
Figure PCTCN2019130897-appb-000005
其中，age A为所述目标年龄；age X为所述候选年龄，Add A为所述目标地址；Add X为所述候选地址；Max_Range为所述距离基准值；α和β为预设系数；Min(age A,age X)为最小值选取函数；Range(Add A,Add X)为距离计算函数
在本实施例中，学习阶段可以从用户的年龄和位于的地理位置间接表示。年龄越相近，位于的地理位置更接近，学习阶段更为相近。年龄越小，年龄差异所造成的影响越明显；年龄越大，这种差异反而变得越小。对于地理位置上的差异，同一座城市内并不会有很大的差异，差异更多是在城市与城市、省内较大的区域以及省与省之间。
在本发明实施例中，通过年龄以及地理位置，确定两个对象之间学习阶段的相似度，从而计算出第二相似因子，能够提高后续计算两个用户之间的相似度的准确性。
图5示出了本发明第五实施例提供的一种信息推送的方法S104的具体实现流程图。参见图5，相对于图1-4所述实施例，本实施例提供的一种信息推送的方法S104包括：S1041～S1044，具体详述如下：
进一步地，所述基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息，包括：
在S1041中，基于各个所述历史查询信息的查询时间，选取所述查询时间在预设的有效时间范围内的所述历史查询信息作为有效查询信息。
在本实施例中，终端设备在确定了关联对象后，需要根据关联对象确定推送给目标对 象的推荐信息，基于此，终端设备可以检测该关联对象各个历史查询信息的查询时间，并选取在预设的有效时间范围内的历史查询信息作为有效查询信息。该有效时间范围可以为以当前时刻至T时刻之间，从而能够获取得到关联对象最近浏览过的历史查询信息，从而能够确定该关联对象当前最感兴趣的信息，提高了推荐信息的时效性以及准确性。
在S1042中，基于所述有效查询信息对应的第二信息，统计各个所述第二信息的第一出现次数。
在本实施例中，终端设备从各个有效查询信息中提取对应的第二信息，并统计各个第二信息在所有有效查询记录中的第一出现次数，若该出现次数越多，则表示该第二信息对应的推荐度越高，目标对象对该第二信息感兴趣的概率越大。
在S1043中，将所述第二信息的所述出现次数以及第二词向量导入到推荐度计算模型，计算所述第二信息的推荐度；所述推荐度计算模型具体为：
Figure PCTCN2019130897-appb-000006
其中，
Figure PCTCN2019130897-appb-000007
为所述推荐度；N wi为所述第一出现次数；
Figure PCTCN2019130897-appb-000008
为所述第二词向量；
Figure PCTCN2019130897-appb-000009
为所述目标对象第j个已查询的目标信息对应的第一目标向量；L 1为所述目标信息的总数。
在本实施例中，终端设备可以将第二信息转换为第二词向量，具体转换为第二词向量的方式可以参照S1022的相关描述，在此不再赘述。在确定了第二词向量以及第一出现次数后，可以将上述两个参数导入到推荐度计算模型，从而计算各个第二信息的推荐度。其中，出现次数越大，则对应的推荐度越大；若该第二信息的第二词向量与目标对象的第一词向量之间的相似度越大，则该第二信息的推荐度越大。
在S1044中，基于各个所述第二信息的推荐度，生成所述推荐信息列表。
在本实施例中，终端设备可以根据各个第二信息的推荐度的数值，从大到小对各个第二信息进行排序，并生成关联对象对应的推荐信息列表。可选地，终端设备可以设置有最大推荐个数，在该情况下，终端设备根据推荐度的数值，从大到小对各个第二信息进行排序，并选取前N个第二信息作为推荐信息，其中N的职位上述的最大推荐个数。当然，终端设备还可以设置有推荐阈值，将所有推荐度大于该推荐阈值的第二信息识别为推荐信息，并生成推荐信息列表。优选地，终端设备可以设置有最大推荐个数以及推荐阈值，选取推荐度大于推荐阈值的第二信息作为候选信息，并选取前N个候选信息作为推荐信息，并生成推荐信息列表。
在本发明实施例中，通过计算各个第二信息的推荐度，从而选取出需要进行推送的推荐信息，生成推荐信息列表，提高了推荐信息的准确率以及与目标用户的相关度，实现了 精准信息推送的目的。
图6示出了本发明第六实施例提供的一种信息推送的方法S105的具体实现流程图。参见图6，相对于图1-4所述实施例，本实施例提供的一种信息推送的方法S105包括：S1051～S1053，具体详述如下：
进一步地，所述对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表，包括：
在S1051中，基于各个所述推荐信息列表包含的推荐信息，统计各个所述推荐信息在所有所述推荐信息列表中的第二出现次数。
在本实施例中，终端设备可以根据各个推荐信息列表中包含的推荐信息，分别统计各个推荐信息在所有推荐信息列表中的第二出现次数，若该出现次数越多，则表示该推荐信息对应的推荐优先级越高，目标对象对该推荐信息感兴趣的概率越大。
在S1052中，根据所述第二出现次数确定各个所述推荐信息的推送优先级。
在本实施例中，终端设备可以根据第二出现次数，分别计算各个推荐信息的推送优先级，例如将第二出现次数直接作为推荐信息的推送优先级。可选地，可以参考S1043的计算公式，将各个推荐信息转换为推荐词向量，并基于推荐词向量以及第二出现次数导入到推荐度计算模型，计算出各个推荐信息的推送优先级。
在S1053中，基于所述推送优先级生成所述待推送信息列表。
在本实施例中，终端设备根据推送优先级的数值，由大到小对所有推荐信息进行排序，从而生成关于目标对象的待推送信息列表，并根据该待推送信息列表内的推送次序，依次向目标对象进行信息推送操作。
在本发明实施例中，通过统计各个推荐信息的第二出现，从而确定各个推荐信息的推送优先级，并生成带有推送优先级次序的待推送信息列表，能够有序地向用户进行推送操作，提高了推送操作准确性。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本发明实施例的实施过程构成任何限定。
图7示出了本发明一实施例提供的一种信息推送的设备的结构框图，该信息推送的设备包括的各单元用于执行图1对应的实施例中的各步骤。具体请参阅图1与图1所对应的实施例中的相关描述。为了便于说明，仅示出了与本实施例相关的部分。
参见图7，所述信息推送的设备包括：
信息查询记录获取单元71，用于获取目标对象的信息查询记录；
相似度计算单元72，用于根据所述信息查询记录以及候选对象的候选查询记录，计算 所述候选对象与所述目标对象之间的相似度；
关联对象选取单元73，用于基于各个所述候选对象对应的所述相似度，从所述候选对象中选取所述目标对象的关联对象；
推荐信息列表生成单元74，用于基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息；
待推送信息列表生成单元75，用于对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表。
可选地，所述相似度计算单元72包括：
信息提取单元，用于从所述信息查询记录中提取所述目标对象的多个第一信息，以及从所述候选查询记录中提取所述候选对象的多个第二信息；
词向量转换单元，用于分别将各个所述第一信息词向量转换模型，生成所述第一信息对应的第一词向量，以及将各个所述第二信息导入到所述词向量转换模型，生成所述第二信息对应的第二词向量；
第一相似因子计算单元，用于将所述第一词向量以及所述第二词向量导入到预设的第一相似因子计算模型，计算所述候选对象以及所述目标对象之间的第一相似因子；
对象属性获取单元，用于获取所述目标对象的目标属性，以及获取所述候选对象的候选属性；
第二相似因子计算单元，用于根据所述目标属性以及所述候选属性，计算所述目标对象以及所述候选对象之间的第二相似因子；
相似度计算单元，用于将所述第一相似因子以及所述第二相似因子导入预设的相似度计算模型，确定所述目标对象以及所述候选对象之间的所述相似度；所述相似度计算模型具体为：
Sim(A,X)＝W IC·Sim IC(A,X)+W SC·Sim SC(A,X)
其中，Sim(A,X)为所述候选对象与所述目标对照之间的所述相似度；Sim IC(A,X)为所述第一相似因子；Sim SC(A,X)为所述第二相似因子；W IC以及W SC为预设系数。
可选地，所述第一相似因子计算单元包括：
相似距离计算单元，用于分别将各个所述第二词向量以及所述第一词向量导入到预设的相似距离计算模型，确定所述第二词向量的相似距离；所述相似距离计算模型具体为：
Figure PCTCN2019130897-appb-000010
其中，v 2i为第i个所述第二词向量的所述相似距离；
Figure PCTCN2019130897-appb-000011
为所述第i个所述第二词向量；
Figure PCTCN2019130897-appb-000012
为所述第j个所述第一词向量；L 1为所述第一词向量的总个数；L 2为所述第二词向量的总个数；Max{x}为最大值选取函数；
相似距离导入单元，用于根据各个所述第二词向量的所述相似距离，计算所述候选对象与所述目标对象之间的所述第一相似因子；计算所述相似度的模型具体为：
Figure PCTCN2019130897-appb-000013
其中，L为预设的词向量长度。
可选地，所述目标属性包括所述目标对象的目标年龄以及目标地址；所述候选属性包括所述候选对象的候选年龄以及候选地址；所述第二相似因子计算单元包括：
距离值计算单元，用于计算任意两个所述候选地址之间的第一距离值，以及计算任一所述候选地址与所述目标地址之间的第二距离值；
距离基准值计算单元，用于从所有所述第一距离值以及所述第二距离值中选取距离最大值，并将所述距离最大值识别为距离基准值；
用户信息导入单元，用于将所述候选对象的候选地址以及所述候选年龄和所述目标对象的目标地址以及所述目标年龄导入到第二相似因子计算模型，计算所述候选对象与所述目标对象之间的所述第二相似因子；所述第二相似因子计算模型具体为：
Figure PCTCN2019130897-appb-000014
其中，age A为所述目标年龄；age X为所述候选年龄，Add A为所述目标地址；Add X为所述候选地址；Max_Range为所述距离基准值；α和β为预设系数；Min(age A,age X)为最小值选取函数；Range(Add A,Add X)为距离计算函数。
可选地，所述推荐信息列表生成单元74包括：
有效信息选取单元，用于基于各个所述历史查询信息的查询时间，选取所述查询时间在预设的有效时间范围内的所述历史查询信息作为有效查询信息；
第一出现次数统计单元，用于基于所述有效查询信息对应的第二信息，统计各个所述第二信息的第一出现次数；
推荐度计算单元，用于将所述第二信息的所述出现次数以及第二词向量导入到推荐度计算模型，计算所述第二信息的推荐度；所述推荐度计算模型具体为：
Figure PCTCN2019130897-appb-000015
其中，
Figure PCTCN2019130897-appb-000016
为所述推荐度；N wi为所述第一出现次数；
Figure PCTCN2019130897-appb-000017
为所述第二词向量；
Figure PCTCN2019130897-appb-000018
为所述目标对象第j个已查询的目标信息对应的第一目标向量；L 1为所述目标信息的总数；
推荐信息选取单元，用于基于各个所述第二信息的推荐度，生成所述推荐信息列表。
可选地，所述待推送信息列表生成单元75包括：
第二出现次数统计单元，用于基于各个所述推荐信息列表包含的推荐信息，统计各个所述推荐信息在所有所述推荐信息列表中的第二出现次数；
推送优先级计算单元，用于根据所述第二出现次数确定各个所述推荐信息的推送优先级；
推送信息选取单元，用于基于所述推送优先级生成所述待推送信息列表。
因此，本发明实施例提供的信息推送的设备同样不只是仅仅依靠目标对象的查询记录进行推送信息的确定，而是根据目标对象的查询记录确定与目标对象的对象属性相似度较高的关联对象，并根据关联对象的关联信息库生成推荐信息列表，由于不同的关联对象所浏览的信息领域较多，且关联对象与目标对象的相似度较高，从而可以推断关联对象感兴趣的信息领域也很大可能是目标对象所感兴趣的信息领域，实现了自动扩展推送信息的信息领域的目的，在保证精准推送的同时，也能够提高信息领域的范围，提高目标对象的信息推送效率。
图8是本发明另一实施例提供的一种终端设备的示意图。如图8所示，该实施例的终端设备8包括：处理器80、存储器81以及存储在所述存储器81中并可在所述处理器80上运行的计算机程序82，例如信息推送的程序。所述处理器80执行所述计算机程序82时实现上述各个信息推送的方法实施例中的步骤，例如图1所示的S101至S105。或者，所述处理器80执行所述计算机程序82时实现上述各装置实施例中各单元的功能，例如图7所示模块71至75功能。
示例性的，所述计算机程序82可以被分割成一个或多个单元，所述一个或者多个单元被存储在所述存储器81中，并由所述处理器80执行，以完成本发明。所述一个或多个单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述所述计算机程 序82在所述终端设备8中的执行过程。例如，所述计算机程序82可以被分割成信息查询记录获取单元、相似度计算单元、关联对象选取单元、推荐信息列表生成单元以及待推送信息列表生成单元，各单元具体功能如上所述。
所述终端设备8可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。所述终端设备可包括，但不仅限于，处理器80、存储器81。本领域技术人员可以理解，图8仅仅是终端设备8的示例，并不构成对终端设备8的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述终端设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器80可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器81可以是所述终端设备8的内部存储单元，例如终端设备8的硬盘或内存。所述存储器81也可以是所述终端设备8的外部存储设备，例如所述终端设备8上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。进一步地，所述存储器81还可以既包括所述终端设备8的内部存储单元也包括外部存储设备。所述存储器81用于存储所述计算机程序以及所述终端设备所需的其他程序和数据。所述存储器81还可以用于暂时地存储已经输出或者将要输出的数据。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
以上所述实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围，均应包含在本发明的保护范围之内。

Claims (10)

  1. 一种信息推送的方法，其特征在于，包括：
    获取目标对象的信息查询记录；
    根据所述信息查询记录以及候选对象的候选查询记录，计算所述候选对象与所述目标对象之间的相似度；
    基于各个所述候选对象对应的所述相似度，从所述候选对象中选取所述目标对象的关联对象；
    基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息；
    对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述信息查询记录以及候选对象的候选查询记录，计算所述候选对象与所述目标对象之间的相似度，包括：
    从所述信息查询记录中提取所述目标对象的多个第一信息，以及从所述候选查询记录中提取所述候选对象的多个第二信息；
    分别将各个所述第一信息词向量转换模型，生成所述第一信息对应的第一词向量，以及将各个所述第二信息导入到所述词向量转换模型，生成所述第二信息对应的第二词向量；
    将所述第一词向量以及所述第二词向量导入到预设的第一相似因子计算模型，计算所述候选对象以及所述目标对象之间的第一相似因子；
    获取所述目标对象的目标属性，以及获取所述候选对象的候选属性；
    根据所述目标属性以及所述候选属性，计算所述目标对象以及所述候选对象之间的第二相似因子；
    将所述第一相似因子以及所述第二相似因子导入预设的相似度计算模型，确定所述目标对象以及所述候选对象之间的所述相似度；所述相似度计算模型具体为：
    Sim(A,X)＝W IC·Sim IC(A,X)+W SC·Sim SC(A,X)
    其中，Sim(A,X)为所述候选对象与所述目标对照之间的所述相似度；Sim IC(A,X)为所述第一相似因子；Sim SC(A,X)为所述第二相似因子；W IC以及W SC为预设系数。
  3. 根据权利要求2所述的方法，其特征在于，所述将所述第一词向量以及所述第二词向量导入到预设的第一相似因子计算模型，计算所述候选对象以及所述目标对象之间的第一相似因子，包括：
    分别将各个所述第二词向量以及所述第一词向量导入到预设的相似距离计算模型，确定所述第二词向量的相似距离；所述相似距离计算模型具体为：
    Figure PCTCN2019130897-appb-100001
    其中，v 2i为第i个所述第二词向量的所述相似距离；
    Figure PCTCN2019130897-appb-100002
    为所述第i个所述第二词向量；
    Figure PCTCN2019130897-appb-100003
    为所述第j个所述第一词向量；L 1为所述第一词向量的总个数；L 2为所述第二词向量的总个数；Max{x}为最大值选取函数；
    根据各个所述第二词向量的所述相似距离，计算所述候选对象与所述目标对象之间的所述第一相似因子；计算所述相似度的模型具体为：
    Figure PCTCN2019130897-appb-100004
    其中，L为预设的词向量长度。
  4. 根据权利要求2所述的方法，其特征在于，所述目标属性包括所述目标对象的目标年龄以及目标地址；所述候选属性包括所述候选对象的候选年龄以及候选地址；
    所述根据所述目标属性以及所述候选属性，计算所述目标对象以及所述候选对象之间的第二相似因子，包括：
    计算任意两个所述候选地址之间的第一距离值，以及计算任一所述候选地址与所述目标地址之间的第二距离值；
    从所有所述第一距离值以及所述第二距离值中选取距离最大值，并将所述距离最大值识别为距离基准值；
    将所述候选对象的候选地址以及所述候选年龄和所述目标对象的目标地址以及所述目标年龄导入到第二相似因子计算模型，计算所述候选对象与所述目标对象之间的所述第二相似因子；所述第二相似因子计算模型具体为：
    Figure PCTCN2019130897-appb-100005
    其中，age A为所述目标年龄；age X为所述候选年龄，Add A为所述目标地址；Add X为所述候选地址；Max_Range为所述距离基准值；α和β为预设系数；Min(age A,age X)为最 小值选取函数；Range(Add A,Add X)为距离计算函数。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，所述基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息，包括：
    基于各个所述历史查询信息的查询时间，选取所述查询时间在预设的有效时间范围内的所述历史查询信息作为有效查询信息；
    基于所述有效查询信息对应的第二信息，统计各个所述第二信息的第一出现次数；
    将所述第二信息的所述出现次数以及第二词向量导入到推荐度计算模型，计算所述第二信息的推荐度；所述推荐度计算模型具体为：
    Figure PCTCN2019130897-appb-100006
    其中，
    Figure PCTCN2019130897-appb-100007
    为所述推荐度；N wi为所述第一出现次数；
    Figure PCTCN2019130897-appb-100008
    为所述第二词向量；
    Figure PCTCN2019130897-appb-100009
    为所述目标对象第j个已查询的目标信息对应的第一目标向量；L 1为所述目标信息的总数；
    基于各个所述第二信息的推荐度，生成所述推荐信息列表。
  6. 根据权利要求1-4任一项所述的方法，其特征在于，所述对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表，包括：
    基于各个所述推荐信息列表包含的推荐信息，统计各个所述推荐信息在所有所述推荐信息列表中的第二出现次数；
    根据所述第二出现次数确定各个所述推荐信息的推送优先级；
    基于所述推送优先级生成所述待推送信息列表。
  7. 一种信息推送的设备，其特征在于，包括：
    信息查询记录获取单元，用于获取目标对象的信息查询记录；
    相似度计算单元，用于根据所述信息查询记录以及候选对象的候选查询记录，计算所述候选对象与所述目标对象之间的相似度；
    关联对象选取单元，用于基于各个所述候选对象对应的所述相似度，从所述候选对象中选取所述目标对象的关联对象；
    推荐信息列表生成单元，用于基于所述关联对象的关联信息库，生成与所述目标对象匹配的推荐信息列表；所述关联信息库包含有所述关联对象的历史查询信息；
    待推送信息列表生成单元，用于对所有所述关联对象的所述推荐信息列表进行合并，生成所述目标对象的待推送信息列表。
  8. 根据权利要求6所述的生成设备，其特征在于，所述相似度计算单元包括：
    信息提取单元，用于从所述信息查询记录中提取所述目标对象的多个第一信息，以及从所述候选查询记录中提取所述候选对象的多个第二信息；
    词向量转换单元，用于分别将各个所述第一信息词向量转换模型，生成所述第一信息对应的第一词向量，以及将各个所述第二信息导入到所述词向量转换模型，生成所述第二信息对应的第二词向量；
    第一相似因子计算单元，用于将所述第一词向量以及所述第二词向量导入到预设的第一相似因子计算模型，计算所述候选对象以及所述目标对象之间的第一相似因子；
    对象属性获取单元，用于获取所述目标对象的目标属性，以及获取所述候选对象的候选属性；
    第二相似因子计算单元，用于根据所述目标属性以及所述候选属性，计算所述目标对象以及所述候选对象之间的第二相似因子；
    相似度计算单元，用于将所述第一相似因子以及所述第二相似因子导入预设的相似度计算模型，确定所述目标对象以及所述候选对象之间的所述相似度；所述相似度计算模型具体为：
    Sim(A,X)＝W IC·Sim IC(A,X)+W SC·Sim SC(A,X)
    其中，Sim(A,X)为所述候选对象与所述目标对照之间的所述相似度；Sim IC(A,X)为所述第一相似因子；Sim SC(A,X)为所述第二相似因子；W IC以及W SC为预设系数。
  9. 一种终端设备，其特征在于，所述终端设备包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时如权利要求1至6任一项所述方法的步骤。
  10. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1至6任一项所述方法的步骤。
PCT/CN2019/130897 2019-04-09 2019-12-31 一种信息推送的方法及设备 WO2020207074A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910279149.0 2019-04-09
CN201910279149.0A CN110162695B (zh) 2019-04-09 2019-04-09 一种信息推送的方法及设备

Publications (1)

Publication Number Publication Date
WO2020207074A1 true WO2020207074A1 (zh) 2020-10-15

Family

ID=67638537

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/130897 WO2020207074A1 (zh) 2019-04-09 2019-12-31 一种信息推送的方法及设备

Country Status (2)

Country Link
CN (1) CN110162695B (zh)
WO (1) WO2020207074A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112862600A (zh) * 2021-03-04 2021-05-28 中国银行股份有限公司 银行网点的查询方法及装置、电子设备及存储介质
CN113609385A (zh) * 2021-07-21 2021-11-05 深圳供电局有限公司 一种浏览信息推送方法及其系统、装置
CN114038528A (zh) * 2021-11-20 2022-02-11 深圳市伊登软件有限公司 基于ai的健康数据处理方法及相关产品
CN115865712A (zh) * 2022-11-15 2023-03-28 重庆邮电大学 一种基于时效亲密模型的边缘数据存储方法
CN116662654A (zh) * 2023-05-26 2023-08-29 惠州市西子湖畔网络有限公司 一种基于大数据的相亲匹配系统及方法

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110162695B (zh) * 2019-04-09 2022-04-26 中国科学院深圳先进技术研究院 一种信息推送的方法及设备
CN111126049B (zh) * 2019-12-14 2023-11-24 中国科学院深圳先进技术研究院 对象关系预测方法、装置、终端设备及可读存储介质
CN111177570B (zh) * 2020-01-13 2023-10-24 中国传媒大学 对象推荐方法及系统
CN111429204A (zh) * 2020-03-10 2020-07-17 携程计算机技术（上海）有限公司 酒店推荐方法、系统、电子设备和存储介质
CN112085566B (zh) * 2020-09-07 2023-04-28 中国平安财产保险股份有限公司 基于智能决策的产品推荐方法、装置及计算机设备
CN112785369A (zh) * 2020-12-31 2021-05-11 网银在线（北京）科技有限公司 用户所属人群的确定方法、装置、电子设备及存储介质
CN113076450B (zh) * 2021-03-15 2024-03-22 北京明略软件系统有限公司 一种目标推荐列表的确定方法和装置
CN113254833B (zh) * 2021-06-07 2021-10-29 深圳市中元产教融合科技有限公司 一种基于产教融合的信息推送的方法及服务系统
CN113343090B (zh) * 2021-06-18 2023-07-28 北京百度网讯科技有限公司 用于推送信息的方法、装置、设备、介质和产品
CN114329189B (zh) * 2021-12-13 2022-10-14 北京五八信息技术有限公司 内容信息的推荐方法、装置、电子设备及可读介质
CN115314550B (zh) * 2022-08-17 2023-08-25 常州市儿童医院(常州市第六人民医院) 基于数字化的智慧医疗信息推送方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017121244A1 (zh) * 2016-01-12 2017-07-20 腾讯科技（深圳）有限公司 信息推荐方法、系统以及存储介质
CN108319603A (zh) * 2017-01-17 2018-07-24 腾讯科技（深圳）有限公司 对象推荐方法和装置
CN109086389A (zh) * 2018-07-26 2018-12-25 国信优易数据有限公司 一种信息查询方法、推送方法、装置及电子设备
CN110162695A (zh) * 2019-04-09 2019-08-23 中国科学院深圳先进技术研究院 一种信息推送的方法及设备

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP6008693B2 (ja) * 2012-10-30 2016-10-19 キヤノン株式会社 情報処理装置及びその制御方法、プログラム
CN103886090B (zh) * 2014-03-31 2018-01-02 北京搜狗科技发展有限公司 基于用户喜好的内容推荐方法及装置
CN105868254A (zh) * 2015-12-25 2016-08-17 乐视网信息技术（北京）股份有限公司 信息推荐方法及装置
CN106997358A (zh) * 2016-01-22 2017-08-01 中移（杭州）信息技术有限公司 信息推荐方法及装置
CN107784066A (zh) * 2017-08-25 2018-03-09 上海壹账通金融科技有限公司 信息推荐方法、装置、服务器及存储介质
CN107577807B (zh) * 2017-09-26 2020-11-10 百度在线网络技术（北京）有限公司 用于推送信息的方法和装置
CN107977355A (zh) * 2017-11-17 2018-05-01 四川长虹电器股份有限公司 基于词向量训练的电视节目推荐方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017121244A1 (zh) * 2016-01-12 2017-07-20 腾讯科技（深圳）有限公司 信息推荐方法、系统以及存储介质
CN108319603A (zh) * 2017-01-17 2018-07-24 腾讯科技（深圳）有限公司 对象推荐方法和装置
CN109086389A (zh) * 2018-07-26 2018-12-25 国信优易数据有限公司 一种信息查询方法、推送方法、装置及电子设备
CN110162695A (zh) * 2019-04-09 2019-08-23 中国科学院深圳先进技术研究院 一种信息推送的方法及设备

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112862600A (zh) * 2021-03-04 2021-05-28 中国银行股份有限公司 银行网点的查询方法及装置、电子设备及存储介质
CN113609385A (zh) * 2021-07-21 2021-11-05 深圳供电局有限公司 一种浏览信息推送方法及其系统、装置
CN114038528A (zh) * 2021-11-20 2022-02-11 深圳市伊登软件有限公司 基于ai的健康数据处理方法及相关产品
CN115865712A (zh) * 2022-11-15 2023-03-28 重庆邮电大学 一种基于时效亲密模型的边缘数据存储方法
CN116662654A (zh) * 2023-05-26 2023-08-29 惠州市西子湖畔网络有限公司 一种基于大数据的相亲匹配系统及方法
CN116662654B (zh) * 2023-05-26 2024-04-09 惠州市西子湖畔网络有限公司 一种基于大数据的相亲匹配系统及方法

Also Published As

Publication number Publication date
CN110162695B (zh) 2022-04-26
CN110162695A (zh) 2019-08-23

Similar Documents

Publication Publication Date Title
WO2020207074A1 (zh) 一种信息推送的方法及设备
CN106897428B (zh) 文本分类特征提取方法、文本分类方法及装置
US11314823B2 (en) Method and apparatus for expanding query
WO2020108608A1 (zh) 搜索结果处理方法、装置、终端、电子设备及存储介质
US20210191509A1 (en) Information recommendation method, device and storage medium
CN110569496B (zh) 实体链接方法、装置及存储介质
US9997157B2 (en) Knowledge source personalization to improve language models
WO2017024884A1 (zh) 一种搜索意图识别方法及装置
WO2018072071A1 (zh) 知识图谱构建系统及方法
WO2022095374A1 (zh) 关键词抽取方法、装置、终端设备及存储介质
CN110069709B (zh) 意图识别方法、装置、计算机可读介质及电子设备
CN112148889A (zh) 一种推荐列表的生成方法及设备
US20220083874A1 (en) Method and device for training search model, method for searching for target object, and storage medium
CN111460153A (zh) 热点话题提取方法、装置、终端设备及存储介质
CN110390044B (zh) 一种相似网络页面的搜索方法及设备
CN112052387B (zh) 一种内容推荐方法、装置和计算机可读存储介质
WO2022141876A1 (zh) 基于词向量的搜索方法、装置、设备及存储介质
CN107832338B (zh) 一种识别核心产品词的方法和系统
CN112100396B (zh) 一种数据处理方法和装置
US20230086735A1 (en) Systems and methods for retrieving videos using natural language description
CN111444304A (zh) 搜索排序的方法和装置
WO2017113725A1 (zh) 一种关联信息的获取与排序方法和系统
CN114416998A (zh) 文本标签的识别方法、装置、电子设备及存储介质
CN110083766B (zh) 一种基于元路径引导嵌入的查询推荐方法及装置
WO2022111095A1 (zh) 一种产品推荐方法、装置、计算机存储介质及系统

Legal Events

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

Ref document number: 19924472

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19924472

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 210322)

122 Ep: pct application non-entry in european phase

Ref document number: 19924472

Country of ref document: EP

Kind code of ref document: A1