CN111797210A

CN111797210A - Information recommendation method, device and equipment based on user portrait and storage medium

Info

Publication number: CN111797210A
Application number: CN202010138636.8A
Authority: CN
Inventors: 于其位
Original assignee: Ping An Life Insurance Company of China Ltd
Current assignee: Ping An Life Insurance Company of China Ltd
Priority date: 2020-03-03
Filing date: 2020-03-03
Publication date: 2020-10-20

Abstract

The invention relates to the technical field of big data, and discloses an information recommendation method based on user portrait, which comprises the following steps: receiving a human-computer conversation chatting record in a first scene in real time through a flash of a log acquisition system; desensitizing the chatting record to obtain first data; performing stop word processing on the first data to obtain second data; extracting keywords of the second data through a preset word graph; obtaining a first label set according to the keyword; removing the duplication of the first label set data to obtain a second label set; generating a user interest portrait based on the second tag set, and storing the user interest portrait in a database; receiving a recommendation instruction, and acquiring a user interest portrait according to the recommendation instruction; and acquiring information to be recommended corresponding to the recommendation instruction according to the user interest portrait. The invention also provides an information recommendation device, equipment and a storage medium based on the user portrait.

Description

Information recommendation method, device and equipment based on user portrait and storage medium

Technical Field

The invention relates to the technical field of big data, in particular to an information recommendation method, device and equipment based on user portrait and a storage medium.

Background

At present, for a customer service and functional robot, user interaction is very purposeful, and even though the current robot cannot meet the requirements of users, certain interests and concerns brought by the users still exist. The existing industry cannot effectively extract user interest tags based on conversation content, and the user interest tags can be applied to practical application scenes of application software, such as: and (4) recommending commodities and advertisements. In the prior art, the user portrait generation mode according to the user interest tags has the phenomenon of inaccurate recommendation when applied to an information recommendation scene, and the fundamental reason is that the accuracy of the existing user portrait is low.

Disclosure of Invention

The invention mainly aims to provide a user portrait-based information recommendation method, device, equipment and storage medium, and aims to solve the technical problem of low user portrait accuracy in the prior art.

Receiving a human-computer conversation chatting record in a first scene in real time through a flash of a log acquisition system;

desensitizing the chat records to obtain first data;

performing stop word processing on the first data to obtain second data;

extracting keywords of the second data through a preset word graph;

obtaining a first label set according to the keyword, wherein the first label set consists of corresponding labels and/or similar labels;

removing the duplication of the first label set data to obtain a second label set;

generating a first user interest representation based on the second set of tags and storing the first user interest representation in a database, the database comprising a data warehouse tool HIVE and a distributed file system HDFS;

and when a recommendation instruction is received, acquiring information to be recommended corresponding to the recommendation instruction according to the first user interest portrait.

Optionally, the desensitizing the chat records to obtain the first data includes the following steps:

filtering machine voice information in the human-computer conversation chatting record through a regular expression;

judging whether sensitive information exists in the received user voice information;

if the received user voice information contains sensitive information, the sensitive information is replaced by a preset symbol through a replacement algorithm to obtain first data.

Optionally, the extracting the keywords of the second data through a preset word graph includes the following steps:

traversing all nodes of the preset word graph, and determining the weight occupied by the keywords according to the occurrence times of the keywords in each node;

and sorting the keywords according to the weight to obtain a target sorting result, and acquiring the keywords of the second data according to the target sorting result.

Optionally, the sorting the keywords according to the weights to obtain a target sorting result, and obtaining the keywords of the second data according to the target sorting result, includes the following steps:

sorting the keywords according to the weights through a bubble sorting algorithm to obtain a target sorting result;

judging whether the accuracy of the target sorting result is greater than or equal to a preset accuracy or not according to a preset target sorting result;

and if the accuracy of the target sorting result is greater than or equal to the preset accuracy, selecting the first N keywords in the target sorting result as the keywords of the second data according to the keywords of the second data obtained from the target sorting result, wherein N is greater than 1.

Optionally, before determining the weight occupied by the keyword according to the occurrence number of the keyword in each node in all the nodes of the traversal of the preset word graph, the method further includes the following steps:

and iterating the weight of each node in the word graph according to a preset rule and training data until the output target ordering result of the keywords corresponding to the training data is greater than or equal to a preset accuracy rate.

Optionally, the obtaining a first tag set according to the keyword, where the first tag set is composed of corresponding tags and/or similar tags, includes the following steps:

if an instruction for acquiring a label according to a keyword is received, judging whether a label matched with the keyword exists in a preset label library or not;

if the preset label library has a label matched with the keyword, acquiring a corresponding label according to a preset mapping relation between the keyword and the label of the preset label library;

if the preset label library does not have a label matched with the keyword, converting the keyword into a corresponding vector;

calculating a cosine included angle between the vector and a preset keyword vector in a preset word bank;

judging whether the cosine included angle is smaller than or equal to a preset included angle or not;

if the cosine included angle is smaller than or equal to a preset included angle, obtaining similar keywords corresponding to the keywords according to a preset keyword vector;

judging whether a label matched with the similar keyword exists in the label system or not;

if the label system has a label matched with the similar keyword, obtaining a similar label;

the corresponding tags and/or the similar tags are grouped into a first set of tags.

Optionally, after the generating a first user interest representation based on the second tag set and storing the first user interest representation in a database, the database including a data warehouse tool HIVE and a distributed file system HDFS, the method further includes the following steps:

receiving a human-computer conversation chat record in a second scene in real time through a flash of a log acquisition system;

extracting a third tag set from the human-computer conversation chatting record in the second scene;

generating a second user representation based on the third set of tags, storing the second user representation in a database, the database comprising a data warehouse tool HIVE and a distributed file system HDFS.

Further, to achieve the above object, the present invention further provides an information recommendation apparatus based on a user profile, including the following modules:

the chat record receiving module is used for receiving the human-computer conversation chat record in the first scene in real time through a log acquisition system flash;

the desensitization processing module is used for desensitizing the chat records to obtain first data;

the stop word processing module is used for performing stop word processing on the first data to obtain second data;

the keyword extraction module is used for extracting keywords of the second data through a preset word graph;

the first label set obtaining module is used for obtaining a first label set according to the keyword, wherein the first label set is composed of corresponding labels and/or similar labels;

the first label set data deduplication module is used for deduplication of the first label set data to obtain a second label set;

a user interest representation storage module, configured to generate a first user interest representation based on the second tag set, and store the first user interest representation in a database, where the database includes a data warehouse tool HIVE and a distributed file system HDFS;

and the information recommending module to be recommended is used for acquiring information to be recommended corresponding to the recommending instruction according to the first user interest image.

Optionally, the desensitization processing module includes the following units:

the filtering unit is used for filtering machine voice information in the human-computer conversation chatting record through a regular expression;

the sensitive information judging unit is used for judging whether the received user voice information has sensitive information or not;

and the replacing unit is used for replacing the sensitive information into a preset symbol through a replacing algorithm to obtain first data if the sensitive information exists in the received user voice information.

Optionally, the keyword extraction module includes the following units:

the traversal unit is used for traversing all nodes of the preset word graph and determining the weight occupied by the keywords according to the occurrence frequency of the keywords in each node;

and the sorting unit is used for sorting the keywords according to the weight to obtain a target sorting result, and acquiring the keywords of the second data according to the target sorting result.

Optionally, the sorting unit is configured to:

sorting the keywords of the weight through a bubble sorting algorithm to obtain a target sorting result;

Optionally, the information recommendation device based on the user profile further includes the following modules:

and the iteration module is used for iterating the weight of each node in the word graph according to a preset rule and training data until the output target ordering result of the keyword corresponding to the training data is greater than or equal to a preset accuracy rate.

Optionally, the first tag set obtaining module includes the following units:

the keyword matching tag judging unit is used for judging whether a tag matched with the keyword exists in a preset tag library or not if an instruction for acquiring the tag according to the keyword is received;

a corresponding tag obtaining unit, configured to obtain a corresponding tag according to a preset mapping relationship between the keyword and a tag in the preset tag library if the tag matching the keyword exists in the preset tag library;

the word vector conversion unit is used for converting the keywords into corresponding vectors if the labels matched with the keywords do not exist in the preset label library;

the calculation unit is used for calculating a cosine included angle between the vector and a preset keyword vector in a preset word bank;

the cosine included angle judging unit is used for judging whether the cosine included angle is smaller than or equal to a preset included angle or not;

a similar keyword obtaining unit, configured to obtain a similar keyword corresponding to the keyword according to a preset keyword vector if the cosine included angle is less than or equal to a preset included angle;

the similar keyword judging unit is used for judging whether a label matched with the similar keyword exists in the label system;

a similar label obtaining unit, configured to obtain a similar label if a label matching the similar keyword exists in the label system;

a combining unit for combining the corresponding label and/or the similar labels into a first label set.

the system comprises a man-machine conversation receiving module under a second scene, a log acquisition module and a chat module, wherein the man-machine conversation receiving module is used for receiving a man-machine conversation chat record under the second scene in real time through a log acquisition system flash;

the third tag set extraction module is used for extracting a third tag set from the human-computer conversation chatting record in the second scene;

and the second user portrait generation module is used for generating a second user portrait based on the third label set and storing the second user portrait in a database, wherein the database comprises a data warehouse tool HIVE and a distributed file system HDFS.

Further, to achieve the above object, the present invention also provides a user representation-based information recommendation apparatus, which includes a memory, a processor and a user representation-based information recommendation program stored in the memory and executable on the processor, wherein the user representation-based information recommendation program, when executed by the processor, implements the steps of the user representation-based information recommendation method according to any one of the above items.

Further, to achieve the above object, the present invention provides a storage medium having a user-representation-based information recommendation program stored thereon, wherein the user-representation-based information recommendation program, when executed by a processor, implements the steps of the user-representation-based information recommendation method according to any one of the above aspects.

The invention filters out the privacy information of the user through desensitization processing, so that the finally generated user portrait does not contain sensitive information, and obtains key words in the man-machine conversation process through a word map mode, thereby ensuring that the generated user portrait is more accurate, in addition, one or more user interest labels can be obtained according to the mapping relation between the key words and a preset label system, the user interest portrait can be generated based on the user interest labels, and the generated user portrait is stored in a database, when an instruction of recommending advertisements or products exists, the user interest portrait can be obtained from the database, and the advertisements or products can be accurately pushed for the user by taking the user portrait as a basis, because each generated user portrait can be stored in the database at any time, the mode not only can fetch the historical user at any time, but also can fetch the real-time user portrait, the accuracy of the user portrait is improved.

Drawings

FIG. 1 is a schematic diagram of an operating environment of an information recommendation device based on a user representation according to an embodiment of the present invention;

FIG. 2 is a flowchart illustrating a user portrait based information recommendation method according to a first embodiment of the present invention;

FIG. 3 is a detailed flowchart of one embodiment of step S20 in FIG. 2;

FIG. 4 is a detailed flowchart of one embodiment of step S40 in FIG. 2;

FIG. 5 is a detailed flowchart of one embodiment of step S402 in FIG. 4;

FIG. 6 is a flowchart illustrating a second embodiment of a user portrait based information recommendation method according to the present invention;

FIG. 7 is a detailed flowchart of one embodiment of step S50 in FIG. 2;

FIG. 8 is a flowchart illustrating a third embodiment of a user portrait based information recommendation method according to the present invention;

FIG. 9 is a block diagram of an embodiment of a user profile-based information recommendation device according to the present invention.

The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.

Detailed Description

It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The information recommendation method based on the user portrait is mainly applied to information recommendation equipment based on the user portrait, and the information recommendation equipment based on the user portrait can be equipment with display and processing functions, such as a PC (personal computer), a portable computer, a mobile terminal and the like.

Referring to fig. 1, fig. 1 is a schematic diagram of a hardware structure of an information recommendation device based on a user profile according to an embodiment of the present invention. In an embodiment of the present invention, the user portrait based information recommendation apparatus may include a processor 1001 (e.g., a CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used for realizing connection communication among the components; the user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard); the network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface); the memory 1005 may be a high-speed RAM memory, or may be a non-volatile memory (e.g., a magnetic disk memory), and optionally, the memory 1005 may be a storage device independent of the processor 1001.

Those skilled in the art will appreciate that the hardware configuration shown in FIG. 1 does not constitute a limitation of a user representation-based information recommendation device, and may include more or fewer components than those shown, or some components in combination, or a different arrangement of components.

With continued reference to FIG. 1, memory 1005, which is one type of storage medium in FIG. 1, may include an operating system, a network communication module, and a user profile-based information recommender.

In fig. 1, the network communication module is mainly used for connecting to a server and performing data communication with the server; and the processor 1001 may call the user-portrait-based information recommendation program stored in the memory 1005 and execute the user-portrait-based information recommendation method according to the embodiment of the present invention.

The embodiment of the invention provides an information recommendation method based on a user portrait.

Referring to fig. 2, fig. 2 is a flowchart illustrating a user portrait-based information recommendation method according to a first embodiment of the present invention. In this embodiment, the information recommendation method based on the user portrait includes the following steps:

step S10, receiving a human-computer conversation chat record in a first scene in real time through a Flume log acquisition system;

in this embodiment, the Flume is a distributed mass log collection system, and the chat records of the real-time human-computer conversation can be received through the log collection system Flume.

Step S20, desensitizing the chat records to obtain first data;

in this embodiment, the chat log of the human-computer conversation includes sensitive information of the user, for example, the sensitive information includes: the mobile phone number, the bank card number, the policy number and the address information can be replaced by a replacement algorithm in order to prevent the sensitive information of the user from being leaked, for example, the mobile phone number, the bank card number and the policy number adopt a method of matching continuous numbers, if the continuous 4-digit heavy rain is matched, the number is changed into a number, and then a mark is added, for example, the bank card number is changed into bankNum. The geographic information changes provinces such as provinces, cities, districts and the like into provinces according to the regular matching, for example, Guangdong province becomes province. The information security of the user is ensured by the unrecoverable desensitization process.

Step S30, performing stop word processing on the first data to obtain second data;

in this embodiment, in order to improve the efficiency of extracting the tag subsequently, the first data is first processed to remove stop words, so as to obtain the second data.

Step S40, extracting the key words of the second data through the preset word graph;

in this embodiment, different data may be input into different nodes of the word graph to obtain the word graph conforming to the current application scenario, and when some data repeatedly appears in a node, the weight occupied by the node storing the data is increased, thereby achieving the purpose of extracting the keyword.

The vocabulary, i.e. Lattice, is a speech recognition result expression form, and a plurality of decoded candidate results are expressed on a directed acyclic graph to extract keywords in the second data.

Step S50, obtaining a first label set according to the keyword, wherein the first label set is composed of corresponding labels and/or similar labels;

in this embodiment, mapping processing of the keyword and the tag system is performed before, and a corresponding relationship between the keyword and the first tag set is established. For example, the received text data of the user is: the 'running' in the text data can be extracted by 'running for half an hour', and the 'running' and the 'healthy' label system have a mapping relation, so that the label of 'healthy' can be obtained, the corresponding label refers to a label which has a corresponding relation with the keyword, the similar label refers to a label which has no corresponding relation with the keyword, but has a corresponding relation with a word similar to the keyword.

Step S60, the data of the first label set are deduplicated to obtain a second label set;

in this embodiment, the first tag set may refer to one tag or a plurality of tags, and under a scenario where there are a plurality of tags, there is a possibility that repeated tags occur, so all tags may be sequentially checked in a traversal manner, and if there is a repetition, a deduplication process is performed to obtain a second tag set.

Step S70, generating a first user interest portrait based on the second label set, and storing the first user interest portrait in a database, wherein the database comprises a data warehouse tool HIVE and a distributed file system HDFS;

in this embodiment, all the second tab sets are combined together, so that a user image of the user in a man-machine conversation scene can be formed. HIVE is a data warehouse tool based on Hadoop, can map structured data files into a database table, provides a complete sql query function, and can convert sql statements into MapReduce tasks for operation. Hadoop, a distributed file system, may be used to store data.

And step S80, acquiring information to be recommended corresponding to the recommendation instruction according to the first user interest image.

In this embodiment, the execution subject of the recommendation information is a server, and the recommendation information can be obtained according to a pre-established correspondence between the user interest representation and the information to be recommended. The information to be recommended is preset, can be commodity information, and can also be data to be responded of a machine in a man-machine conversation scene.

By desensitization processing, private information of a user is filtered out, so that a finally generated user portrait does not contain sensitive information, a keyword in a man-machine conversation process is obtained in a word map mode, the generated user portrait can be ensured to be more accurate, in addition, one or more user interest tags can be obtained according to a mapping relation between the keyword and a preset tag system, the user interest portrait can be generated based on the user interest tags, the generated user portrait is stored in a database, when an instruction of recommending advertisements or products exists, the user interest portrait can be obtained from the database, the advertisements or the products are accurately pushed for the user by taking the user portrait as a basis, and because each generated user portrait is stored in the database, the mode not only can call historical user portrait at any time, but also can call real-time user portrait, the accuracy of the user portrait is improved.

Referring to fig. 3, fig. 3 is a detailed flowchart of an embodiment of step S20 in fig. 2. In this embodiment, in step S20, performing desensitization processing on the chat record to obtain first data, includes the following steps:

step S201, filtering machine voice information in the chat records of the man-machine conversation through a regular expression;

in this embodiment, since it is desired to obtain a user portrait rather than a machine portrait, it is necessary to filter out voice information of a machine.

Step S202, judging whether sensitive information exists in the received user voice information;

in this embodiment, in order to protect the privacy of the user, after the received user voice information, it is necessary to determine whether sensitive information exists. Wherein the sensitive information includes: a mobile phone number, a bank card number, a policy number, and address information.

Step S203, if the sensitive information exists in the received user voice information, the sensitive information is replaced by a preset symbol through a replacement algorithm to obtain first data.

In this embodiment, if there is sensitive information, the sensitive data and the address information need to be converted into preset symbols, for example, the mobile phone number, the bank card number, and the policy number information in the sensitive information are replaced with the preset symbols through a replacement algorithm, and the address information in the sensitive information is replaced with the second preset symbols through a regular method.

Referring to fig. 4, fig. 4 is a detailed flowchart of an embodiment of step S40 in fig. 2. In this embodiment, in step S40, extracting the keyword of the second data through the preset word graph includes the following steps:

step S401, traversing all nodes of a preset word graph, and determining the weight occupied by the keywords according to the occurrence frequency of the keywords in each node;

in this embodiment, the weight occupied by the word segmentation result may be determined by the number of times that the traversed word segmentation result appears in the word graph node, for example, the weight is higher when the number of times of occurrence is higher.

And S402, sorting the keywords according to the weight to obtain a target sorting result, and acquiring the keywords of the second data according to the target sorting result.

In this embodiment, after the word segmentation results with weights are ranked, a target ranking result may be obtained, for example, outputting the keywords ranked in the first three digits may be preset to obtain a final keyword.

Referring to fig. 5, fig. 5 is a schematic view of a detailed flow of an embodiment of step S402 in fig. 4. In this embodiment, the step S402 of ranking the keywords according to the weights to obtain a target ranking result, and obtaining the keywords of the second data according to the target ranking result includes the following steps:

s4021, sorting the weighted keywords through a bubble sorting algorithm to obtain a target sorting result;

step S4022, judging whether the accuracy of the target sorting result is greater than or equal to the preset accuracy according to the preset target sorting result;

step S4023, if the accuracy of the target sorting result is greater than or equal to the preset accuracy, selecting the first N keywords in the target sorting result as the keywords of the second data according to the keywords of the second data obtained from the target sorting result, wherein N is greater than 1.

In this embodiment, two-by-two comparison is performed starting from the head of the unordered sequence, and positions are exchanged according to size until the largest (small) data element is exchanged to the tail of the unordered queue, thereby becoming a part of the ordered sequence; this process continues the next time until all data elements are in order. The keywords meeting preset conditions can be obtained through the sorting process, for example, only the first 3 keywords in the sorted queue are obtained in a preset mode, in order to enable the result to be more accurate, the keywords are sorted manually before the result is, then the keywords sorted through bubbling are compared, the accuracy rate of the bubbling sorting is calculated, in order to enable the result of the bubbling sorting to be more accordant with the artificially preset rule, and therefore if the accuracy rate is not more than or equal to the preset accuracy rate, the weights of certain keywords can be properly reduced or increased, and the keywords accordant with the preset rule are obtained.

Referring to fig. 6, fig. 6 is a flowchart illustrating a user portrait-based information recommendation method according to a second embodiment of the present invention. In this embodiment, in step S401, before traversing all nodes of the preset word graph and determining the weight occupied by the keyword according to the occurrence number of the keyword in each node, the method further includes the following steps:

and step S90, iterating the weights of each node in the word graph according to a preset rule until the target ordering result of the keywords output by the word graph is greater than or equal to the preset accuracy preset rule and the weights of each node in the training data iteration word graph until the target ordering result of the keywords corresponding to the training data output by the word graph is greater than or equal to the preset accuracy.

In this embodiment, sometimes a word with a high frequency of occurrence is not necessarily a keyword, and therefore a preset mechanism is required to train a weight coefficient of each node, for example, the frequency of occurrence of the two keywords "run" and "lose weight" is very high, but a tag matching with "run" does not exist in a preset word library, but a tag matching with "lose weight" exists, so that the real intention of a user may be more prone to lose weight, and therefore, the weight occupied by the node for storing "run" and "lose weight" needs to be adjusted until the real requirement of the user is met, and when a word graph receives the keyword in such a scene again, the keyword "lose weight" can be directly output.

Referring to fig. 7, fig. 7 is a detailed flowchart of an embodiment of step S50 in fig. 2. In this embodiment, step S50, obtaining a first tag set according to the keyword, where the first tag set is composed of a corresponding tag and a similar tag, and includes the following steps:

in this embodiment, whether an instruction for obtaining a tag according to a keyword is currently received is determined.

Step S501, if an instruction for acquiring a tag according to a keyword is received, judging whether a tag matched with the keyword exists in a preset tag library;

in this embodiment, if an instruction for obtaining a tag according to a keyword is currently received, it is determined whether a tag matching the keyword exists in a tag system.

Step S502, if the preset label library has a label matched with the keyword, obtaining a corresponding label according to the preset mapping relation between the keyword and the label of the preset label library;

step S503, if the preset label library does not have a label matched with the keyword, converting the keyword into a corresponding vector;

in this embodiment, the preset mapping relationship may be generated by clustering the keywords through a clustering algorithm, so as to obtain the category to which each keyword belongs, and obtaining the label according to the category, so as to establish the mapping relationship between the keywords and the label. If the label system has a label matched with the keyword, obtaining a corresponding label according to the mapping relation, and if not, converting the keyword into a corresponding vector.

Step S504, calculating a cosine included angle between the vector and a preset keyword vector in a preset word bank;

step S505, judging whether the cosine included angle is less than or equal to a preset included angle;

in this embodiment, an angle between a vector of a current keyword and a vector of a keyword in a preset lexicon is calculated by a cosine similarity algorithm, a numerical value of the angle may be preset, for example, 20 degrees, when the angle between the vectors is less than 20 degrees, a word similar to the current keyword, that is, a similar keyword, may be obtained, and when the similar keyword exists, a similar label of the similar keyword may be obtained.

Step S506, if the cosine included angle is smaller than or equal to the preset included angle, obtaining similar keywords corresponding to the keywords according to the preset keyword vector;

in this embodiment, a cosine included angle between the vector and a preset keyword vector in a preset lexicon is calculated to obtain a similar keyword corresponding to the keyword. The preset word stock is a library for storing keywords, if a tag matched with a keyword does not exist in the preset tag stock, it is indicated that the current keyword does not have a tag corresponding to the current keyword, in an actual scene, a message is often pushed to a user according to an interest tag of the user, when the tag corresponding to the keyword cannot be obtained, the message cannot be pushed, in order to overcome the defect, in this embodiment, an included angle between a vector of the current keyword and a vector of the keyword in the preset word stock is calculated through a cosine similarity algorithm, a numerical value of the included angle may be preset, for example, 20 degrees, when the included angle between the vectors is less than 20 degrees, a word similar to the current keyword, that is, a similar keyword, and a similar tag of the similar keyword may be obtained.

Step S507, judging whether a label matched with the similar keyword exists in a label system;

step S508, if the label matched with the similar keyword exists in the label system, obtaining a similar label;

in this embodiment, if a tag matching the similar keyword exists in the tag system, a tag matching the similar keyword is obtained. After the similar labels are added on the basis of the corresponding labels, the richness of the labels can be improved, and therefore, the user portrait generation can be more accurate.

In step S509, the corresponding tag and the similar tags are combined into a first tag set.

Similar keywords of the keywords can be obtained by calculating the cosine included angle between the keyword vector and the preset keyword vector, and similar labels can be obtained according to the similar keywords, so that the richness of the labels can be increased.

Referring to fig. 8, fig. 8 is a flowchart illustrating a third embodiment of a user portrait-based information recommendation method according to the present invention. In this embodiment, in step S70, after generating the first user interest representation based on the second tag set and storing the first user interest representation in a database, the database including a data warehouse tool HIVE and a distributed file system HDFS, the method further includes the steps of:

step S100, receiving a human-computer conversation chat record in a second scene in real time through a flash of a log acquisition system;

in this embodiment, the other scenes refer to scenes other than the chat scene of the human-computer conversation.

Step S110, extracting a third label set from the human-computer conversation chatting record in the second scene;

and step S120, generating a second user portrait based on the third label set, and storing the second user portrait in a database, wherein the database comprises a data warehouse tool HIVE and a distributed file system HDFS.

In this embodiment, since the obtained user portrait is obtained in the scene of the human-computer conversation chat, the interest tag of the user in other scenes can also be obtained according to the scheme, and then the user portrait in other scenes can be obtained.

The keywords are obtained in a word graph mode, corresponding labels and similar labels are obtained according to the mapping relation between the keywords and the labels, the number of the labels is greatly increased, user interest portraits are finally obtained according to the user interest labels, and each generated user portraits are stored in a database.

Referring to fig. 9, fig. 9 is a functional block diagram of an embodiment of an information recommendation device based on a user profile according to the present invention. In this embodiment, the information recommendation apparatus based on a user profile includes:

the chat record receiving module 10 is used for receiving the human-computer conversation chat record in the first scene in real time through a log acquisition system Flume;

a desensitization processing module 20, configured to perform desensitization processing on the chat record to obtain first data;

a stop word processing module 30, configured to perform stop word processing on the first data to obtain second data;

a keyword extraction module 40, configured to extract keywords of the second data through a preset word graph;

a first tag set obtaining module 50, configured to obtain a first tag set according to the keyword, where the first tag set is composed of corresponding tags and/or similar tags;

a first label set data deduplication module 60, configured to deduplicate the first label set data to obtain a second label set;

a user interest representation storage module 70 for generating a first user interest representation based on the second set of tags and storing the first user interest representation in a database, the database comprising a data warehouse tool HIVE and a distributed file system HDFS;

and the information to be recommended recommending module 80 is configured to obtain information to be recommended corresponding to the recommending instruction according to the first user interest image.

The invention also provides a storage medium.

In this embodiment, the storage medium stores a user-representation-based information recommendation program, and the user-representation-based information recommendation program, when executed by a processor, implements the steps of the user-representation-based information recommendation method according to any one of the embodiments.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM), and includes instructions for causing a terminal (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.

The present invention is described in connection with the accompanying drawings, but the present invention is not limited to the above embodiments, which are only illustrative and not restrictive, and those skilled in the art can make various changes without departing from the spirit and scope of the invention as defined by the appended claims, and all changes that come within the meaning and range of equivalency of the specification and drawings that are obvious from the description and the attached claims are intended to be embraced therein.

Claims

1. A user portrait based information recommendation method is characterized by comprising the following steps:

desensitizing the chat records to obtain first data;

performing stop word processing on the first data to obtain second data;

extracting keywords of the second data through a preset word graph;

2. The user representation-based information recommendation method of claim 1, wherein desensitizing the chat log to obtain the first data comprises:

3. The user profile-based information recommendation method of claim 1, wherein the extracting the keywords of the second data through the preset word graph comprises:

4. The method as claimed in claim 3, wherein the step of ranking the keywords according to the weights to obtain a target ranking result, and obtaining the keywords of the second data according to the target ranking result comprises:

and if the accuracy of the target sorting result is greater than or equal to the preset accuracy, selecting the first N keywords in the target sorting result as the keywords of the second data, wherein N is greater than 1.

5. The user representation-based information recommendation method of claim 3, wherein before determining the weight of the keyword in each node according to the number of occurrences of the keyword in all nodes of the traversal of the preset word graph, the method further comprises:

6. The method of claim 1, wherein the deriving a first tag set according to the keyword, wherein the first tag set is composed of corresponding tags and/or similar tags, comprises:

7. The user representation-based information recommendation method of any of claims 1-6, wherein after said generating a first user interest representation based on said second set of tags and storing said first user interest representation in a database, said database comprising a data warehouse facility HIVE and a distributed file system HDFS, further comprising:

8. A user profile-based information recommender, comprising:

9. A user representation-based information recommendation apparatus comprising a memory, a processor, and a user representation-based information recommendation program stored on the memory and executable on the processor, the user representation-based information recommendation program when executed by the processor implementing the steps of the user representation-based information recommendation method of any of claims 1-7.

10. A storage medium having stored thereon a user representation-based information recommender, the user representation-based information recommender when executed by a processor implementing the steps of the user representation-based information recommendation method as claimed in any of claims 1 to 7.