CN116186402A

CN116186402A - Product public opinion analysis method and system for big data platform

Info

Publication number: CN116186402A
Application number: CN202310141558.0A
Authority: CN
Inventors: 陈璇; 吴荣明
Original assignee: Individual
Current assignee: Individual
Priority date: 2023-02-21
Filing date: 2023-02-21
Publication date: 2023-05-30

Abstract

The invention provides a product public opinion analysis method and system for a big data platform, and relates to the technical field of artificial intelligence. In the invention, product public opinion direction data corresponding to each product public opinion data important segment in product public opinion description data is analyzed, and the statistical quantity value of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data is determined; according to the statistics value, analyzing matching product public opinion direction data corresponding to important segments of each product public opinion data; determining target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the statistics value; and analyzing target product public opinion direction data based on the product public opinion data important segments corresponding to the target direction matching parameters larger than the reference direction matching parameters. Based on the above, the reliability of product public opinion analysis can be improved.

Description

Product public opinion analysis method and system for big data platform

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a product public opinion analysis method and system for a big data platform.

Background

Artificial intelligence (Artificial Intelligence, AI for short) is a theory, method, technique and application system that simulates, extends and extends human intelligence, senses environment, obtains knowledge and uses knowledge to obtain optimal results using digital computers or digital computer controlled computations. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence.

The artificial intelligence technology is applied in many ways, for example, it can be used to analyze public opinion data of products, such as public opinion or public opinion. In the prior art, in the process of analyzing public opinion data of products, the problem of low reliability exists.

Disclosure of Invention

Therefore, the present invention aims to provide a product public opinion analysis method and system for a big data platform, so as to improve reliability of product public opinion analysis.

In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme:

a product public opinion analysis method for a big data platform comprises the following steps:

extracting a plurality of product public opinion description data of a target product, wherein each product public opinion description data has at least one product public opinion data important segment and product public opinion direction data, and each product public opinion data important segment belonging to one product public opinion description data corresponds to the same product public opinion direction data, and the product public opinion direction data is used for reflecting the public opinion direction of the target product;

Performing data analysis operation on the plurality of product public opinion description data to output product public opinion direction data corresponding to each product public opinion data important segment in the plurality of product public opinion description data, and determining a statistics value of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data;

according to the statistical magnitude of corresponding product public opinion description data of each product public opinion data important segment under corresponding product public opinion direction data, analyzing matched product public opinion direction data corresponding to each product public opinion data important segment from the corresponding product public opinion direction data;

determining target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the statistics value of corresponding product public opinion description data of each product public opinion data important segment under corresponding product public opinion direction data;

and carrying out product public opinion analysis operation on the product public opinion description data based on the product public opinion data important segments corresponding to the target direction matching parameters larger than the reference direction matching parameters so as to output target product public opinion direction data corresponding to the product public opinion description data, wherein the target product public opinion direction data is at least used for reflecting the public opinion direction of the target product.

In some preferred embodiments, in the product public opinion analysis method for a big data platform, the step of performing product public opinion analysis operation on the product public opinion description data based on the product public opinion data important segments corresponding to the target direction matching parameters greater than the reference direction matching parameters to output target product public opinion direction data corresponding to the plurality of product public opinion description data includes:

determining important fragments of the product public opinion data corresponding to target direction matching parameters larger than the reference direction matching parameters, and marking the important fragments as data fragments to be processed so as to form a data fragment cluster to be processed;

q mapping characteristic representations corresponding to the data segments to be processed in the xth tense are determined, and network tense data of J analysis sub-networks included in the public opinion analysis neural network in the xth-1 tense are determined, wherein the network tense data include intermediate output data of the J analysis sub-networks in the xth-1 tense;

determining a correlation characterization parameter of each analysis sub-network based on the intermediate output data of the J analysis sub-networks in the x-1 time state and the Q mapping characteristic representations, and determining L first analysis sub-networks in the J analysis sub-networks based on the correlation characterization parameter of each analysis sub-network, wherein the number of the first analysis sub-networks is smaller than or equal to that of the analysis sub-networks;

Determining to-be-processed fusion data of a to-be-processed first analysis sub-network, wherein the to-be-processed fusion data is formed by performing decimation and fusion operations on the Q mapping characteristic representations and the intermediate output data of the J analysis sub-networks in the x-1 th tense based on the intermediate output data of the to-be-processed first analysis sub-network in the x-1 th tense, and the to-be-processed first analysis sub-network belongs to any one of the L first analysis sub-networks;

analyzing the intermediate output data of the first analysis sub-network to be processed in the x-th tense based on the fusion data to be processed;

combining the intermediate output data corresponding to each data segment to be processed in the data segment clusters to form a corresponding intermediate output data cluster, wherein the data segment to be processed in the xth tense belongs to one data segment to be processed in the data segment clusters to be processed;

and loading the intermediate output data cluster to the public opinion direction output sub-network included in the public opinion analysis neural network to analyze target product public opinion direction data corresponding to the product public opinion description data.

In some preferred embodiments, in the product public opinion analysis method for big data platform, the step of determining the to-be-processed fusion data of the to-be-processed first analysis sub-network includes:

based on the intermediate output data of the first analysis sub-network to be processed in the x-1 th tense, determining the correlation characterization parameters corresponding to the Q mapping feature representations, and determining Z mapping feature representations matched with a preset correlation rule from the Q mapping feature representations, wherein the number of the Z mapping feature representations is smaller than or equal to the number of the Q mapping feature representations;

based on the intermediate output data of the first analysis subnetwork to be processed in the x-1 th tense, determining the correlation characterization parameters of the intermediate output data of the J-1 analysis subnetworks other than the first analysis subnetwork to be processed in the x-1 th tense in the J-analysis subnetwork, and determining the intermediate output data of the W analysis subnetworks matched with the set correlation rule in the x-1 th tense in the intermediate output data of the J-1 analysis subnetwork in the x-1 th tense;

and determining the fusion data to be processed of the first analysis sub-network to be processed based on the Z mapping characteristic representations, the intermediate output data of the W analysis sub-networks in the x-1 time and the intermediate output data of the first analysis sub-network to be processed in the x-1 time.

In some preferred embodiments, in the product public opinion analysis method for big data platform, the step of analyzing the output data of the first analysis sub-network to be processed in the x-th temporal based on the fusion data to be processed includes:

loading the network temporal data of the first analysis sub-network to be processed in the x-1 th temporal state and the fusion data to be processed into the first analysis sub-network to be processed, and analyzing corresponding candidate processing data;

determining to-be-processed aggregate data of the to-be-processed first analysis sub-network;

analyzing the intermediate output data of the first analysis sub-network to be processed in the x-th time based on the aggregate data to be processed and the candidate processing data, wherein the aggregate data to be processed is formed by performing decimation and aggregation operations on the Q mapping characteristic representations and the intermediate output data of the J-1 analysis sub-networks, except the first analysis sub-network to be processed, in the J analysis sub-networks in the x-1 time based on the intermediate output data of the first analysis sub-network to be processed in the x-1 time, and the decimation and aggregation operations are consistent with the processing mode of the decimation and aggregation operations.

In some preferred embodiments, in the product public opinion analysis method for a big data platform, the step of determining the target direction matching parameter of each product public opinion data important segment under the corresponding matching product public opinion direction data according to the statistics of the corresponding product public opinion description data under the corresponding product public opinion direction data of each product public opinion data important segment includes:

analyzing the probability parameters of each product public opinion data important segment appearing under the corresponding product public opinion direction data according to the statistics value of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data;

according to the probability parameters of each product public opinion data important segment under the corresponding product public opinion directional data, analyzing the matching characterization parameters of each product public opinion data important segment, and analyzing the parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion directional data, wherein the parameter aggregation coefficient is used for reflecting the parameter aggregation degree of the product public opinion data important segment under the corresponding matching product public opinion directional data;

And analyzing target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the multiplication operation result of the matching characterization parameters of each product public opinion data important segment and the parameter aggregation coefficients of each product public opinion data important segment under corresponding matching product public opinion direction data.

In some preferred embodiments, in the product public opinion analysis method for a big data platform, the product public opinion direction data corresponding to the product public opinion data important segment includes first product public opinion direction data for reflecting a product profit direction and second product public opinion direction data for reflecting a product profit direction;

the step of analyzing the matching characterization parameters of each product public opinion data important segment according to the probability parameters of each product public opinion data important segment appearing under the corresponding product public opinion direction data comprises the following steps:

determining the possibility parameters of each product public opinion data important segment under the corresponding first product public opinion directional data, determining the possibility parameters of each product public opinion data important segment under the corresponding second product public opinion directional data, and calculating the subtraction operation result among the possibility parameters;

And marking the larger one of the subtraction operation result and the pre-configured reference matching parameter to be the matching characterization parameter of each product public opinion data important segment.

In some preferred embodiments, in the product public opinion analysis method for a big data platform, the step of analyzing the matching characterization parameters of each product public opinion data important segment according to the probability parameters of each product public opinion data important segment appearing under the corresponding product public opinion direction data includes:

based on a predetermined matching characterization parameter determining rule, determining a matching characterization parameter corresponding to each product public opinion data important segment, wherein the matching characterization parameter determining rule is as follows:

and for each product public opinion data important segment, determining the possibility parameter of the product public opinion data important segment under each product public opinion direction data corresponding to and used for reflecting the product interest direction, calculating the difference value between the possibility parameter and a first preset value, carrying out summation operation on the difference value corresponding to the possibility parameter under each product public opinion direction data, and determining the matching characterization parameter corresponding to the product public opinion data important segment based on the summation operation result and a second preset value.

In some preferred embodiments, in the product public opinion analysis method for big data platform, the step of analyzing the parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion direction data includes:

according to the probability parameter and a third preset value of each product public opinion data important segment under the corresponding product public opinion directional data, and based on a predetermined aggregation coefficient determining rule, determining a parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion directional data;

the aggregation coefficient determination rule is as follows:

and determining the probability parameters of the product public opinion data important segments under the corresponding product public opinion direction data for reflecting the product interest direction for each product public opinion data important segment, and determining the parameter aggregation coefficient of the product public opinion data important segments under the corresponding matched product public opinion direction data based on the maximum value of the probability parameters under each product public opinion direction data and the third preset value, wherein the parameter aggregation coefficient and the maximum value and the third preset value have positive correlation corresponding relations respectively.

In some preferred embodiments, in the product public opinion analysis method for a big data platform, the step of analyzing matching product public opinion direction data corresponding to each product public opinion data important segment from the corresponding product public opinion direction data according to the statistics of corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data includes:

and marking the product public opinion direction data with the largest product public opinion description data in the statistics value of the corresponding product public opinion description data under the corresponding product public opinion direction data for each product public opinion data important segment so as to mark the matched product public opinion direction data corresponding to each product public opinion data important segment.

The embodiment of the invention also provides a product public opinion analysis system for the big data platform, which comprises a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program so as to realize the product public opinion analysis method for the big data platform.

The product public opinion analysis method and system for the big data platform provided by the embodiment of the invention can analyze the product public opinion direction data corresponding to each product public opinion data important segment in the product public opinion description data, and determine the statistics value of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data; according to the statistics value, analyzing matching product public opinion direction data corresponding to important segments of each product public opinion data; determining target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the statistics value; and analyzing target product public opinion direction data based on the product public opinion data important segments corresponding to the target direction matching parameters larger than the reference direction matching parameters. Based on the foregoing, before the product public opinion analysis operation, the product public opinion data important segments are analyzed and screened, so that the reliability of the product public opinion data important segments for the product public opinion analysis operation is higher, and the reliability of the product public opinion analysis is improved.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

Fig. 1 is a block diagram of a product public opinion analysis system for a big data platform according to an embodiment of the present invention.

Fig. 2 is a flowchart illustrating steps involved in a product public opinion analysis method for a big data platform according to an embodiment of the present invention.

Fig. 3 is a schematic diagram of each module included in a product public opinion analysis device for a big data platform according to an embodiment of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

As shown in FIG. 1, the embodiment of the invention provides a product public opinion analysis system for a big data platform. The product public opinion analysis system for the big data platform can comprise a memory, a processor, other possible devices and the like.

In detail, the memory and the processor are electrically connected directly or indirectly to realize transmission or interaction of data. For example, electrical connection may be made to each other via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute the executable computer program stored in the memory, so as to implement the product public opinion analysis method for big data platform provided by the embodiment of the present invention.

It should be appreciated that in some possible embodiments, the Memory may be, but is not limited to, random access Memory (Random Access Memory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like.

It should be appreciated that in some possible embodiments, the processor may be a general purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a System on Chip (SoC), etc.; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

It should be appreciated that in some possible embodiments, the product public opinion analysis system for a big data platform may be a server with data processing capabilities.

With reference to fig. 2, the embodiment of the invention further provides a product public opinion analysis method for a big data platform, which can be applied to the product public opinion analysis system for the big data platform. The method steps defined by the flow related to the product public opinion analysis method for the big data platform can be realized by the product public opinion analysis system for the big data platform.

The specific flow shown in fig. 2 will be described in detail.

Step S110, extracting a plurality of product public opinion description data of the target product.

In the embodiment of the invention, the product public opinion analysis system for the big data platform can extract a plurality of product public opinion description data of the target product, wherein the plurality of product public opinion description data can be public opinion descriptions of the target product, for example, public opinion descriptions of different platforms, different users or different time periods, such as public opinion descriptions in a first time period, public opinion descriptions in a second time period, public opinion descriptions in a third time period and public opinion descriptions in a fourth time period, and the main evaluation of the target product in the A-group is XXX. Each product public opinion description data has at least one product public opinion data important segment and product public opinion direction data, each product public opinion data important segment in one product public opinion description data corresponds to the same product public opinion direction data, the product public opinion direction data is used for reflecting the public opinion direction of the target product, the product public opinion data important segment can be one or more important segments in a plurality of product public opinion data segments included in the product public opinion description data, the product public opinion data segments can refer to a sentence and the like, specific granularity can be configured according to actual requirements, in addition, the product public opinion data important segments can be marked based on manual analysis or not marked and need subsequent analysis mining. The specific definition of the product public opinion direction data is not limited, and can be configured according to actual needs, for example, the product is good, bad, or neutral evaluation of the product, or specific degree or degree grade of the product is good, bad, and other complex definitions can be obtained.

Step S120, performing data analysis operation on the plurality of product public opinion description data to output product public opinion direction data corresponding to each product public opinion data important segment in the plurality of product public opinion description data, and determining a statistics value of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data.

In the embodiment of the invention, the product public opinion analysis system for the big data platform can perform data analysis operation on the plurality of product public opinion description data to output product public opinion direction data (as described above, the product public opinion direction data can be determined based on labels) corresponding to each product public opinion data important segment in the plurality of product public opinion description data, and determine the statistics of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data. For example, the product public opinion directional data 1 has a product public opinion data important segment a and corresponds to the product public opinion directional data a, the product public opinion directional data 2 has a product public opinion data important segment a and corresponds to the product public opinion directional data a, the product public opinion directional data 3 has a product public opinion data important segment a and corresponds to the product public opinion directional data a, and based on this, the statistics of product public opinion description data corresponding to the product public opinion data important segment a under the corresponding product public opinion directional data a is 3.

Step S130, analyzing matching product public opinion directional data corresponding to each product public opinion data important segment from the corresponding product public opinion directional data according to the statistics value of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion directional data.

In the embodiment of the invention, the product public opinion analysis system for the big data platform can analyze matching product public opinion directional data corresponding to each product public opinion data important segment from the corresponding product public opinion directional data according to the statistical magnitude of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion directional data.

Step S140, determining target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the statistics value of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data.

In the embodiment of the invention, the product public opinion analysis system for the big data platform can determine the target direction matching parameter of each product public opinion data important segment under the corresponding matching product public opinion direction data according to the statistics value of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data.

And step S150, carrying out product public opinion analysis operation on the product public opinion description data based on the product public opinion data important segments corresponding to the target direction matching parameters larger than the reference direction matching parameters so as to output target product public opinion direction data corresponding to the product public opinion description data.

In the embodiment of the invention, the product public opinion analysis system for the big data platform can perform product public opinion analysis operation on the product public opinion description data based on the product public opinion data important segments corresponding to the target direction matching parameters larger than the reference direction matching parameters so as to output target product public opinion direction data corresponding to the product public opinion description data. The target product public opinion direction data is at least used for reflecting the public opinion direction of the target product. The specific value of the reference direction matching parameter is not limited, the configuration can be performed according to actual application requirements, the target product public opinion direction data can be the final determined public opinion direction of the target product, and the product public opinion direction data corresponding to the important segments of the product public opinion data can be initially confirmed, such as manually confirmed.

Based on the foregoing, before the product public opinion analysis operation, the product public opinion data important segments are analyzed and screened (for example, by determining the corresponding target direction matching parameter, the product public opinion data important segments corresponding to the target direction matching parameter greater than the reference direction matching parameter can be screened), so that the reliability of the product public opinion data important segments for the product public opinion analysis operation is higher, and the reliability of the product public opinion analysis is improved.

It should be understood that, in some possible embodiments, for the step S130, that is, the step of analyzing, from the corresponding product public opinion directional data, the matching product public opinion directional data corresponding to each product public opinion data important segment according to the statistics of the corresponding product public opinion description data corresponding to each product public opinion data important segment, the following detailed description may further be included:

and marking the largest product public opinion direction data in the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data so as to mark the matched product public opinion direction data corresponding to each product public opinion data important segment, for example, the product public opinion data important segment 1 under the corresponding product public opinion direction data 1 corresponds to the product public opinion description data with the product public opinion data important segment 5, and the product public opinion data important segment 1 under the corresponding product public opinion direction data 2 corresponds to the product public opinion description data with the product public opinion direction data 8, so that the matched product public opinion direction data corresponding to the product public opinion data important segment 1 is determined to be the product public opinion direction data 2.

It should be understood that, in some possible embodiments, for the step S140, that is, the step of determining the target direction matching parameter of each product public opinion data important segment under the corresponding matching product public opinion direction data according to the statistics of the corresponding product public opinion description data under the corresponding product public opinion direction data of each product public opinion data important segment, the following detailed description may be further included:

according to the statistical magnitude of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data, analyzing the possibility parameter of each product public opinion data important segment under the corresponding product public opinion direction data, for example, the possibility parameter of the product public opinion data important segment 1 under the corresponding product public opinion direction data 1 is equal to the sum of the statistical magnitude of the corresponding product public opinion description data of the product public opinion data important segment 1 under the corresponding product public opinion direction data 1 divided by the statistical magnitude of the corresponding product public opinion description data of the product public opinion data important segment 1 under the corresponding product public opinion direction data (such as the product public opinion direction data 1, the product public opinion direction data 2 and the product public opinion direction data 3);

according to the multiplication operation result of the matching characterization parameter of each product public opinion data important segment and the parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion direction data, the target direction matching parameter of each product public opinion data important segment under the corresponding matching product public opinion direction data is analyzed, for example, the multiplication operation result of the matching characterization parameter and the parameter aggregation coefficient can be used as the corresponding target direction matching parameter.

It should be understood that, in some possible embodiments, the product public opinion data corresponding to the product public opinion data important segments includes first product public opinion data for reflecting the product public opinion direction and second product public opinion data for reflecting the product public opinion direction, based on which, for the likelihood parameter according to which each product public opinion data important segment appears under the corresponding product public opinion data, the step of analyzing the matching characterization parameter of each product public opinion data important segment may further include the following detailed description:

Determining the possibility parameters of each product public opinion data important segment under the corresponding first product public opinion direction data, determining the possibility parameters of each product public opinion data important segment under the corresponding second product public opinion direction data, and calculating the subtraction operation result (namely the difference value of the two possibility parameters) between the possibility parameters;

and marking the larger one of the subtraction operation result and the pre-configured reference matching parameters to be the matching characterization parameter of each product public opinion data important segment, wherein the specific numerical value of the reference matching parameter is not limited, and the configuration can be carried out according to actual requirements, such as 0, 0.015 and the like.

It should be understood that, in other possible embodiments, for the likelihood parameter according to which each product public opinion data important segment appears under the corresponding product public opinion directional data, the step of analyzing the matching characterization parameter of each product public opinion data important segment may further include the following detailed description:

For each important piece of product public opinion data, determining a possibility parameter of the important piece of product public opinion data under each corresponding product public opinion direction data for reflecting the product interest direction, calculating a difference value between the possibility parameter and a first preset value, carrying out summation operation on the difference value corresponding to the possibility parameter under each product public opinion direction data, and determining a matching characterization parameter corresponding to the important piece of product public opinion data based on a result of the summation operation and a second preset value, wherein specific values of the first preset value and the second preset value are not limited, can be configured according to actual requirements, for example, the first preset value can be 1, the second preset value can be 0, and for example, a larger value in the result of the summation operation and the second preset value can be marked to mark the matching characterization parameter corresponding to the important piece of product public opinion data.

It should be understood that, in some possible embodiments, for the step of analyzing the parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion direction data, the following detailed description may be further included:

According to the probability parameter and a third preset value of each product public opinion data important segment under the corresponding product public opinion directional data, and based on a predetermined aggregation coefficient determining rule, determining a parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion directional data; the aggregation coefficient determination rule is as follows:

for each product public opinion data important segment, determining a probability parameter of the product public opinion data important segment appearing under each product public opinion direction data corresponding to and used for reflecting the product interest direction, and determining a parameter aggregation coefficient of the product public opinion data important segment under the corresponding matching product public opinion direction data based on the maximum value of the probability parameter appearing under each product public opinion direction data and the third preset value, wherein the parameter aggregation coefficient and the maximum value and the third preset value have positive correlation corresponding relations respectively, for example, a difference value between the fourth preset value and the maximum value can be calculated firstly, then a ratio between the third preset value and the difference value can be calculated to obtain the parameter aggregation coefficient, the specific numerical value of the third preset value and the fourth preset value is not limited, for example, the third preset value can be a numerical value larger than 0, and the fourth preset value can be a numerical value configured according to the real requirements of 2, 2.5 and the like.

It should be understood that, in some possible embodiments, for the step S150, that is, the step of performing the product public opinion analysis operation on the product public opinion descriptive data based on the product public opinion data important segments corresponding to the target direction matching parameters greater than the reference direction matching parameters, to output the target product public opinion direction data corresponding to the plurality of product public opinion descriptive data, the following detailed description may be further included:

determining Q mapping feature representations corresponding to the data segments to be processed in the x-th tense, and determining network tense data of J analysis sub-networks in the x-1 th tense, where the network tense data includes intermediate output data of the J analysis sub-networks in the x-1 th tense, where, by way of example, the Q mapping feature representations corresponding to the data segments to be processed may refer to performing feature space mapping processing on the data segments to be processed by Q different mapping modes to obtain Q mapping feature representations in the Q different mapping modes, where the dimension numbers between the Q mapping feature representations may be consistent, but belong to different feature spaces, and a specific mapping mode may be configured according to actual requirements, for example, a plurality of mapping functions may be configured, parameters in each mapping function may be formed in a training process of the corresponding neural network, and the network tense data may also include other data, and it is required that, for an LSTM neuron, two pieces of information are transferred backward in a time dimension; (1) other data (2) intermediate output data, wherein the intermediate output data is obtained by passing other data through a neuron and an "output gate", so that memory contained in the intermediate output data is substantially that after other data has decayed, and further, the other data is transferred along the time axis in a channel having less decay, and information having a larger time span is retained much more than the intermediate output data, so that substantially "recent memory" is stored in the intermediate output data, mainly "long-term memory", refer to the related description of LSTM specifically, and in addition, it should be noted that, the xth tense may refer to the xth moment or time step on the time axis, and when x is equal to 0, i.e. for the 0 th tense, the network tense data may be configured, for example, formed during the training process of the corresponding neural network;

combining intermediate output data corresponding to each data segment to be processed in the data segment cluster to form a corresponding intermediate output data cluster, wherein the data segment to be processed in the x-th temporal belongs to one data segment to be processed in the data segment cluster to be processed, and illustratively, the data segments to be processed can be arranged in sequence in the data segment cluster to be processed, for example, the data segments to be processed are arranged in sequence according to corresponding time, so that for a first temporal, the corresponding data segment to be processed can be the data segment to be processed with earliest time, and for a last temporal, the corresponding data segment to be processed can be the data segment to be processed with latest time;

And loading the intermediate output data cluster to the public opinion direction output sub-network included in the public opinion analysis neural network to analyze target product public opinion direction data corresponding to the product public opinion description data, wherein the public opinion analysis neural network can be formed by training based on the sample to-be-processed data fragment cluster and the corresponding public opinion direction data label.

Wherein it should be understood that, in some possible embodiments, for the step of determining the correlation characterization parameter of each analysis sub-network based on the intermediate output data of the J analysis sub-networks in the x-1 th tense and the Q mapping feature representations, the following detailed description may be further included:

calculating the number product of the intermediate output data of each analysis sub-network in the x-1 time state and the Q mapping feature representations to output Q number products corresponding to each analysis sub-network, for example, for the analysis sub-network A, calculating the number product of the intermediate output data of the analysis sub-network A in the x-1 time state and each mapping feature representation in the Q mapping feature representations can be performed, so that Q number products corresponding to the analysis sub-network A can be obtained;

And carrying out accumulation operation on the Q number products corresponding to each analysis sub-network so as to output the correlation characterization parameter corresponding to each analysis sub-network, for example, for the analysis sub-network A, the Q number products corresponding to the analysis sub-network A can be subjected to addition operation, and thus, the correlation characterization parameter corresponding to the analysis sub-network A can be obtained.

Wherein it should be understood that, in some possible embodiments, for the correlation characterization parameter based on each analysis sub-network, determining L first analysis sub-networks in the J analysis sub-networks may further include the following detailed description:

marking L analysis subnetworks with the largest correlation characterization parameters among the J analysis subnetworks so as to be marked as a first analysis subnetwork; or, marking L analysis sub-networks with correlation characterization parameters larger than or equal to preset reference correlation characterization parameters in the J analysis sub-networks to be marked as a first analysis sub-network, wherein the specific values of the reference correlation characterization parameters and L are not limited, and the configuration can be carried out according to actual application scenes.

It should be understood that, in some possible embodiments, for the step of determining the fusion data to be processed of the first analysis sub-network to be processed, the following detailed description may be included:

determining correlation characterization parameters corresponding to the Q mapping feature representations respectively based on the intermediate output data of the first analysis sub-network to be processed in the x-1 th temporal, and determining Z mapping feature representations matched with a preset correlation rule in the Q mapping feature representations, wherein the number of the Z mapping feature representations is smaller than or equal to the number of the Q mapping feature representations, for example, the correlation characterization parameters, such as cosine similarity between feature representations and vectors, of each of the Q mapping feature representations and the intermediate output data of the first analysis sub-network to be processed in the x-1 th temporal can be calculated first, and then, the mapping feature representations, such as the cosine similarity, of which the corresponding correlation characterization parameters are larger than the first reference correlation characterization parameters, can be determined, so that Z mapping feature representations matched with the preset correlation rule can be obtained;

based on the intermediate output data of the first analysis subnetwork to be processed in the x-1 th time, determining the correlation characterization parameters of the intermediate output data of the J-1 analysis subnetworks other than the first analysis subnetwork to be processed in the x-1 th time respectively, and determining the intermediate output data of the W analysis subnetworks matched with the set correlation rule in the x-1 th time in the intermediate output data of the J-1 analysis subnetwork in the x-1 th time, for example, determining the intermediate output data of the analysis subnetwork with the corresponding correlation characterization parameters larger than the second reference correlation characterization parameters in the x-1 th time;

Wherein it should be understood that, in some possible embodiments, for the step of determining the fusion data to be processed of the first analysis sub-network to be processed based on the Z mapping feature representations, the intermediate output data of the W analysis sub-networks in the x-1 th time and the intermediate output data of the first analysis sub-network to be processed in the x-1 th time, the method may further include the following detailed description:

optimizing Q-Z mapping feature representations other than the Z mapping feature representations in the Q mapping feature representations, wherein for each of the Q-Z mapping feature representations, it may be determined whether the mapping feature representation is consistent with any one of the Z mapping feature representations, if so, the mapping feature representation itself is regarded as an optimized mapping feature representation, and if not, a preset feature representation may be regarded as an optimized mapping feature representation, the preset feature representation may be configured according to actual requirements, for example, parameters of the preset feature representation may all be 0;

Optimizing the intermediate output data of the J- (1+m) -th analysis sub-network outside the W-th analysis sub-networks in the x-1 th time, and for each intermediate output data of the J- (1+m) -th analysis sub-network in the x-1 th time, determining whether the intermediate output data is consistent with the intermediate output data of any one of the W-th analysis sub-networks in the x-1 th time, if so, taking the intermediate output data as the optimized intermediate output data, and if not, taking a preset feature representation as the optimized intermediate output data, wherein the preset feature representation is consistent with or not consistent with the preset feature representation;

and carrying out fusion operation on the Z mapping feature representations, the optimized Q-Z mapping feature representations, the intermediate output data of the W analysis subnetworks in the x-1 th tense, the intermediate output data of the optimized J- (1+M) analysis subnetworks in the x-1 th tense and the intermediate output data of the first analysis subnetwork to be processed in the x-1 th tense (cascade combination can be carried out on feature representation dimensions so as to increase the number of dimensions), and forming the fusion data to be processed of the first analysis subnetwork to be processed.

It should be understood that, in some possible embodiments, for the step of analyzing the intermediate output data of the first analysis sub-network to be processed in the x-th temporal based on the fusion data to be processed, the following detailed description may be further included:

loading the network temporal data of the x-1 th temporal state of the first analysis sub-network to be processed and the fusion data to be processed into the first analysis sub-network to be processed, and analyzing corresponding candidate processing data, wherein the network temporal data and the fusion data to be processed can be calculated based on network parameters in the first analysis sub-network to be processed, such as forgetting door parameters, input door parameters, output door parameters and the like, so as to obtain corresponding candidate processing data, and specific processing procedures, such as weighting, summation and the like, can refer to the processing procedures of a long-term and short-term memory neural network, and are not described in detail herein;

Wherein it should be understood that, in some possible embodiments, for the step of determining the aggregate data to be processed of the first analysis sub-network to be processed, the following detailed description may be further included:

optimizing Q-Z mapping feature representations out of the Z mapping feature representations in the Q mapping feature representations, as previously described in relation;

optimizing intermediate output data of J- (1+M) analysis subnetworks outside the W analysis subnetworks in the J-1 analysis subnetwork in the x-1 time state, as described in the related description;

and carrying out fusion operation on the Z mapping characteristic representations, the optimized Q-Z mapping characteristic representations, the intermediate output data of the W analysis subnetworks in the x-1 time and the intermediate output data of the optimized J- (1+M) analysis subnetworks in the x-1 time, so as to form the to-be-processed aggregation data of the to-be-processed first analysis subnetwork.

Wherein it should be understood that, in some possible embodiments, for the step of analyzing the intermediate output data of the first analysis sub-network to be processed in the x-th temporal based on the aggregate data to be processed and the candidate processing data, the method may further include the following detailed description:

Combining the intermediate output data of the first analysis sub-network to be processed in the x-1 time state with the candidate processing data to form corresponding combined data to be processed, wherein the combined operation can be consistent with the fusion operation, namely cascade combination in dimension is performed;

the method comprises the steps of performing linear mapping operation on the to-be-processed aggregate data to output corresponding linear mapping aggregate data, performing multiplication operation on the linear mapping aggregate data and the to-be-processed combined data to output correlation parameter distribution corresponding to the to-be-processed combined data, wherein the linear mapping operation can be exemplified by multiplying a target linear mapping parameter distribution and the to-be-processed aggregate data to obtain corresponding linear mapping aggregate data, and further performing normalization processing on a result of the multiplication operation after the multiplication operation is performed on the linear mapping aggregate data and the to-be-processed combined data to obtain corresponding correlation parameter distribution;

Based on the correlation parameter distribution corresponding to the to-be-processed combined data, updating the to-be-processed combined data, outputting the intermediate output data of the to-be-processed first analysis sub-network in the x-th tense, and by way of example, multiplying the correlation parameter distribution corresponding to the to-be-processed combined data by the to-be-processed combined data, so that the intermediate output data of the to-be-processed first analysis sub-network in the x-th tense can be output.

Wherein it should be understood that in some possible embodiments, the network temporal data of each analysis sub-network except the L first analysis sub-networks in the J analysis sub-networks in the x-1 th temporal may be marked as the network temporal data of the analysis sub-network in the x-th temporal, that is, the network temporal data of the analysis sub-network is maintained in the x-th temporal.

Based on this, the public opinion analysis neural network may determine an actual processing object corresponding to each analysis sub-network based on the characteristics of each analysis sub-network, that is, the mapping feature representation and the correlation characterization parameter of the analysis sub-network, for example, the actual processing object of the analysis sub-network 1 is the mapping feature representation 2 and the mapping feature representation 5, the actual processing object of the analysis sub-network 2 is the mapping feature representation 1 and the mapping feature representation 2, the actual processing object of the analysis sub-network 3 is the mapping feature representation 3 and the mapping feature representation 5, and the actual processing object of the analysis sub-network 4 is the mapping feature representation 2, the mapping feature representation 4 and the mapping feature representation 5. Therefore, the data fragments to be processed, which are actually loaded to each analysis sub-network, can be richer, and the reliability of subsequent processing is further ensured.

With reference to fig. 3, the embodiment of the invention further provides a product public opinion analysis device for a big data platform, which can be applied to the product public opinion analysis system for the big data platform. The product public opinion analysis device for the big data platform can comprise the following functional modules:

the public opinion data extraction module is used for extracting a plurality of product public opinion description data of a target product, each product public opinion description data has at least one product public opinion data important segment and product public opinion direction data, each product public opinion data important segment in one product public opinion description data corresponds to the same product public opinion direction data, and the product public opinion direction data is used for reflecting the public opinion direction of the target product;

the public opinion data analysis module is used for carrying out data analysis operation on the plurality of product public opinion description data so as to output product public opinion direction data corresponding to each product public opinion data important segment in the plurality of product public opinion description data and determine the statistics value of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data;

The public opinion data matching module is used for analyzing matched product public opinion direction data corresponding to each product public opinion data important segment from the corresponding product public opinion direction data according to the statistics value of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data;

the target matching parameter determining module is used for determining target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the statistics value of corresponding product public opinion description data of each product public opinion data important segment under corresponding product public opinion direction data;

the public opinion direction determining module is used for carrying out product public opinion analysis operation on the product public opinion description data based on the product public opinion data important segments corresponding to the target direction matching parameters which are larger than the reference direction matching parameters so as to output target product public opinion direction data corresponding to the product public opinion description data, wherein the target product public opinion direction data is at least used for reflecting the public opinion direction of the target product.

In summary, the product public opinion analysis method and system for the big data platform provided by the invention can analyze the product public opinion direction data corresponding to each product public opinion data important segment in the product public opinion description data, and determine the statistics value of the product public opinion description data corresponding to each product public opinion data important segment under the corresponding product public opinion direction data; according to the statistics value, analyzing matching product public opinion direction data corresponding to important segments of each product public opinion data; determining target direction matching parameters of each product public opinion data important segment under corresponding matching product public opinion direction data according to the statistics value; and analyzing target product public opinion direction data based on the product public opinion data important segments corresponding to the target direction matching parameters larger than the reference direction matching parameters. Based on the foregoing, before the product public opinion analysis operation, the product public opinion data important segments are analyzed and screened, so that the reliability of the product public opinion data important segments for the product public opinion analysis operation is higher, and the reliability of the product public opinion analysis is improved.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. The product public opinion analysis method for the big data platform is characterized by comprising the following steps of:

2. The product public opinion analysis method for big data platform of claim 1, wherein the step of performing product public opinion analysis operation on the product public opinion description data based on the product public opinion data important segments corresponding to the target direction matching parameters greater than the reference direction matching parameters to output the target product public opinion direction data corresponding to the plurality of product public opinion description data comprises:

3. The product public opinion analysis method for big data platform of claim 2, wherein the step of determining the to-be-processed fusion data of the to-be-processed first analysis sub-network comprises:

4. The product public opinion analysis method for big data platform of claim 3, wherein the step of analyzing the intermediate output data of the first analysis sub-network to be processed at the x-th tense based on the fusion data to be processed comprises:

5. The product public opinion analysis method for big data platform of any one of claims 1-4, wherein the step of determining the target direction matching parameter of each product public opinion data important segment under the corresponding matching product public opinion direction data according to the statistics of the corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data comprises:

6. The product public opinion analysis method for big data platform of claim 5, wherein the product public opinion direction data corresponding to the important piece of product public opinion data comprises a first product public opinion direction data for reflecting the product interest direction and a second product public opinion direction data for reflecting the product interest direction;

7. The product public opinion analysis method for big data platform of claim 5, wherein the step of analyzing the matching characterization parameters of each product public opinion data important segment according to the probability parameters of each product public opinion data important segment appearing under the corresponding product public opinion direction data comprises:

8. The product public opinion analysis method for big data platform of claim 7, wherein the step of analyzing the parameter aggregation coefficient of each product public opinion data important segment under the corresponding matching product public opinion direction data comprises:

The aggregation coefficient determination rule is as follows:

9. The product public opinion analysis method for big data platform according to any one of claims 1-4, wherein the step of analyzing matching product public opinion direction data corresponding to each product public opinion data important segment from the corresponding product public opinion direction data according to the statistics of corresponding product public opinion description data of each product public opinion data important segment under the corresponding product public opinion direction data comprises:

10. A product public opinion analysis system for a big data platform, comprising a processor and a memory, the memory for storing a computer program, the processor for executing the computer program to implement the method of any of claims 1-9.