CN113254796A

CN113254796A - Network object label management method and system

Info

Publication number: CN113254796A
Application number: CN202010089351.XA
Authority: CN
Inventors: 林韦廷
Original assignee: Chubibi Co ltd
Current assignee: Chubibi Co ltd
Priority date: 2020-02-12
Filing date: 2020-02-12
Publication date: 2021-08-13

Abstract

The invention is a network article label management method and system, in the method, obtain the public data about particular network article from the network first, every public data has time stamps, through analyzing the public data about this network article, can find out the corresponding label in the meaning, set up the association of the label analyzed with the network article, and then store the public data and label associated with network article based on time stamp with the database, wherein, there is a time threshold, when the time of any public data exceeds the time threshold, obtain the public data about network article again, in order to upgrade, delete, newly-add or maintain the label associated with network article. The system is provided with an intelligent module, and a plurality of labels related to a plurality of network objects are learned from a database by a neural network algorithm so as to establish a model capable of automatically judging the labels related to the public data.

Description

Network object label management method and system

[ technical field ] A method for producing a semiconductor device

The present invention relates to the field of label setting technology for specific articles, and is especially network article label managing method and system capable of setting label related to specific article based on the meaning of public data in network.

[ background of the invention ]

The network objects can refer to various things for evaluation on the network, such as lodging places, delicates, scenic spots and other objects which are disclosed on the network and are interesting to users, and the network platform often sets tags (tag) according to the attributes of various network objects, so that the tags are convenient for users to search, classify and obtain preliminary information. The tag information of various network objects may be set by a network platform operator according to the attributes of the network objects, and taking a lodging place as an example, the related tags have the advantages of novel equipment, convenient traffic, good service attitude, quiet environment, good lighting, tidiness and the like. The labels of the network items can also come from real experiences after the use of the consumers, and the contents of the experiences can be written in blogs, social networks, web messages, forums and the like, so that relevant network owners can analyze the characteristics of accommodation places according to the contents, such as novel (or old) equipment, good (or bad) sound insulation of rooms, rich (or common) dining spots and the like.

The label can be used for judging the attribute of the specific network object, so that the user can conveniently and quickly obtain the evaluation of the interested place. For example, what people arrange for a tour is most important and need to arrange in advance is more than a lodging and three meals, searching for a lodging place and a place to have a meal using a network is already a necessary lesson before people travel, and when searching for data, evaluation of each lodging place or restaurant on the network is very important reference information.

The user can inquire the lodging places by relying on the label information on the network when the lodging places such as hotels, lodging residents, camping places, camping vehicles and the like which provide the lodging for the user, and the user can inquire wrong data if the label information cannot reflect the current situation.

For example, for a new accommodation site, the relevant evaluations may be equipment-good, housing-novelty, but may lack a complete service-wise evaluation; for old accommodation sites, the rating may be good service, good dining, but with the impression of old equipment. However, when such hotel, hotel and lodging pairs are already under-managed and poorly operated, the old high ratings will affect consumer rights and benefits; conversely, when the lodging location is improved from the past, and there is a good management team, the choice is missed if the old rating is unfair and the consumer cannot get the latest information.

Therefore, consumers may need to carefully search for the latest evaluation or spend time confirming the update, and thus, the complete and up-to-date information cannot be obtained effectively.

[ summary of the invention ]

According to the technical purpose of the network object label management method and system provided by the specification, one of the technical purposes is that a reliable label can be set for each network object, so that the situation that a user obtains wrong information due to outdated, deficient or wrong labels is avoided, and the method can enable the user to quickly obtain interesting and accurate contents when inquiring the network object.

In one embodiment, a system for managing network object tags is provided with a database for storing public data about a plurality of network objects obtained from a network and for analyzing each public data to obtain one or more tags associated with each network object, a server for executing a method for managing network object tags, capable of analyzing the public data about each network object, wherein each public data has a timestamp, and for establishing an association between each network object and the analyzed one or more tags according to the corresponding tag obtained from the meaning of each public data, and for storing the public data based on the timestamp and the tags associated with each network object in the database for querying.

Accommodation sites such as hotels, camping camps, motels, etc. provide sites for users to check in, and public data about accommodation sites such as introductions, shares, and comments about accommodation site names and accommodation sites are made public on the web, which may include positive or negative evaluations.

Furthermore, the system sets a time threshold for each public data, the public data with the time greater than the time threshold is not referred to establish a label, and when the time of any public data of the network object exceeds the time threshold, the public data of the network object can be obtained again so as to update, delete, add or maintain the label associated with the network object.

Preferably, the public data of the network object comprises audio-visual or text contents disclosed in the blog, the community media and the forum, and a network software program running on the network retrieves the audio-visual or text contents disclosed in the blog, the community media and the forum, and can be stored in the database in a structured manner.

Then, a computer software program can analyze the public data of the network object sentence by sentence or paragraph by paragraph, and determine one or more labels corresponding to each sentence or each paragraph, or none of the labels. Or the public data can be analyzed by a natural language semantic analysis method, and the attributes of a plurality of labels provided by the server are compared to judge one or more labels corresponding to each sentence or each paragraph in the public data.

Furthermore, the network object label management system also comprises an intelligent module, wherein a neural network algorithm is operated, and a plurality of labels related to a plurality of network objects can be learned from a database through the neural network algorithm to establish a neural network model.

Further, when a new network object is obtained and includes the associated public data, or new public data about the network object with the tag association established is obtained, the associated tag can be obtained or updated through the neural network model, or the neural network model can be modified according to feedback information.

The specific techniques employed in the present invention will be further illustrated by the following examples and accompanying drawings.

[ description of the drawings ]

FIG. 1 is a diagram of an embodiment of a network object tag management system.

FIG. 2 is a main flow chart of a method for implementing the tag management of network objects.

FIG. 3 depicts a flow diagram of an embodiment of a method for ensuring tag timeliness in network item tag management.

FIG. 4 is a diagram of an embodiment of functional modules of a network object tag management system.

FIG. 5 is a diagram of an embodiment of an application-like neural network of the network object tag management system.

FIG. 6 shows an embodiment of a process for training a neural network, using an accommodation site as an example.

FIG. 7 shows an embodiment of a process for analyzing public data using a trained neural network to generate labels, for example, at a lodging site.

Fig. 8 shows an embodiment of the process for the customer to select the accommodation site.

FIG. 9 shows an embodiment of a process for retraining a neural network-like model for accommodation site tuning.

Description of the main element symbols:

10 network

12 server

14 database

101 character data

103 video and audio data

105 picture data

401 accommodation place information management module

402 accommodation place information module

403 data acquisition module

404 data segmentation module

405 label correspondence module

406 neural network format conversion module

407 authentication module

Model generation module for 408 type neural network

409 type neural network model evaluation module

410 accommodation place label optimization module

411 tag management module

412 tag generation module

413 tag database

421 data

422 disclose data

423 tag adjustment

424 accommodation site

426 accommodation site

51 public data

52 database

Class 53 neural network model

54 tag database

55 network platform

Class 56 neural network algorithm

57 feedback information

501 user

[ detailed description ] embodiments

The specification discloses a network object label management method and system, the network object label management system mainly includes a server, wherein, the software method realized by software means and matching with the computer arithmetic capability can obtain the video and text contents of the related network objects disclosed on the network through the software program, analyze the contents through the software method and compare a plurality of labels provided by the system, or a model capable of automatically judging the meaning and the corresponding label in the public data can be established in an artificial intelligence algorithm to finally form the label of each network object, so that a user can inquire and screen the network objects according to the label, according to an embodiment, a network item may refer to various items on the network that are available for evaluation, such as accommodation locations, so the system may form tags that provide queries and filtering for various accommodation locations.

Accommodation sites such as hotels, camping camps, motels, campers, etc. provide sites for users to check in, while public data about accommodation sites such as introductions, shares, and comments about accommodation site names and accommodation sites are disclosed on the network, which may include positive or negative evaluations.

The network object tag management system is configured on a network system, and is mainly configured as shown in fig. 1, and includes a server 12, where the server 12 can be implemented by a computer system to implement a network object tag management method, and includes a processor and a memory, where the processor executes the network object tag management method, and a main flow embodiment is shown in fig. 2. In this system architecture, the network object tag management system is provided with a database 14 for storing public data about a plurality of network objects obtained from the network 10, and for analyzing each public data to obtain one or more tags associated with each network object.

The public data about the network object can be various text data 101, video data 103 and picture data 105, and the public data can be analyzed by a software method to obtain the content and meaning. Particularly, as the text data 101, the meaning can be analyzed by using a natural language semantic analysis method, so that the labels which can be represented in the text data and the corresponding labels can be analyzed. For example, the video data 103 and the picture data 105 can be analyzed by a software method to obtain the content, and the tags thereof can be analyzed.

According to the main flowchart of fig. 2, a network software program (e.g. web crawler program) is used to obtain various public data on the network, and the public data is recorded in a database of a specific server, and then the public data about the network object is parsed sentence by sentence or paragraph by a computer software program in step S203, so as to determine one or more labels corresponding to each sentence or each paragraph, and the public data is classified, and the linked network object (e.g. lodging site) is determined according to the content, so that information such as related time and place can be obtained from forums, websites or blogs where the public data are posted, and the public data has a time stamp (time stamp) indicating the public time of the public data.

In the server, a plurality of tags (tag) representing various information may be preset, and the computer software program may compare the meaning represented by each tag according to the meaning of each public data obtained by parsing, as in step S205, except for determining the associated network object, one or more corresponding tags may be obtained, or no corresponding tag may be obtained, and the association between each network object and the parsed one or more tags is established. Thereafter, in step S207, the public data based on the time stamp and one or more tags associated with each network item are stored in a database, and then the tags associated with a specific network item can be published and queried.

Fig. 3 is a flowchart illustrating an embodiment of ensuring the timeliness of the tag in the tag management method of the network object, wherein the process is executed continuously or periodically by the tag management system of the network object to ensure the timeliness of the tag associated with the network object.

In step S301, one or more tags associated with each network object and corresponding public data in the database are retrieved, and then in step S303, after the difference between the current time and the time of the public data is obtained, a time threshold is compared, in step S305, it can be determined whether the public data exceeds the time threshold set by the system, and when the time of any public data related to the network object does not exceed the time threshold, it indicates that the public data is still reliable information, i.e. it is not processed, and the process returns to step S301 and other processes for ensuring the temporal nature of the tags; otherwise, if the time of the public data exceeds the time threshold by the current time, in step S307, it is determined whether there is any newer public data related to the network object, and the content in the public data can be retrieved and analyzed to update, delete, add or maintain one or more tags associated with the network object.

According to the embodiment of the network object tag management method, the network object tag management system is mainly realized by a network object tag management system which is mainly realized by a server and comprises a plurality of various software functional modules which are realized by software and matched with hardware processing capacity, as shown in fig. 4, the embodiment shows the tag management system of the accommodation place.

The body of the system is an application that maintains the operation of the overall network item tag management system in accommodation locations, including but not limited to hotels, camping camps, motels, etc. where users live.

The system obtains public data about data 421 of various accommodation places from the network, manages names, addresses, coordinates (such as global positioning coordinates), telephone, e-mail, official website information of the accommodation places through an accommodation place information management module 401, and obtains the information from public data by means such as an accommodation place information retrieving module 402, retrieves information about each accommodation place by a computer software program according to keywords of country, city, region, etc., and establishes information about the accommodation places.

The system continuously or periodically retrieves public data 422 about various accommodation places from the network by using the public data retrieving module 403, such as API concatenation, web crawler, manual operation or batch import, where the retrieved items include but are not limited to websites, blogs, forums, community websites and other systems or platforms for users to share information, evaluation, opinion, discussion and mind about accommodation places, and similarly, the system establishes association with each accommodation place.

Since the data captured by the public data capturing module 403 comes from various platforms and systems, and may be provided in various formats such as HTML, TEXT, CSV, JSON, etc., the system performs data cleansing on the public data obtained from the network through the data segmenting module 404, the data cleansing includes removing meaningless symbols, links, movies, pictures, words, etc., setting different information cleansing rules according to the public data from different sources (web pages, blogs, social networks, forums, etc.), and periodically maintaining the information cleansing manner with the information format change of the target website. And after the information is cleaned, segmenting the data into sentences, and storing the segmented sentences into a database.

In one embodiment, after various data about the accommodation site is obtained by the data segmentation module 404, the conversion and segmentation may be performed according to the size, length or format of the information content. The purpose of segmentation is to divide the lengthy information into several segments, which has the advantage of facilitating the correspondence of the tags. For example, it may not be a good way to deliver all 5 ten thousand words to the neural network algorithm for training, and in the data segmenting module 404, the commonly used symbols (such as periods, line-feed symbols, blanks, and straight-line symbols) of the 5 ten thousand words may be used as the basis for segmentation. Furthermore, segmentation may also be based on word count, possibly close to N words as a way of segmentation. In addition to these segmentation methods, it is also possible to determine when to segment and segment the data by using semantic correlation between the context of each paragraph, and to save the data in the database or specific file in the system for the convenience of subsequent operations.

The tag mapping module 405 is used to perform tag mapping management on the sentences stored in the database, and give one or more tags to each sentence, or do not have any tag. The method of setting labels for each sentence can be obtained by comparing the label attributes according to keywords by a software program, or after the meaning of each sentence is obtained by analyzing the semantic meaning of natural language, the associated labels can be obtained by comparing the attributes of each label, especially one or more labels can be compared for each sentence or each paragraph in public data, or the labels associated with each sentence can be judged by a neural network-like model trained by a neural network-like algorithm, so that the labels can be automatically generated by the software methods, and if the accuracy is improved, the generated labels can be audited by manual reference. Further, adjustments made by manual review may form a feedback neural network-like algorithm to update the established neural network-like model.

For example, the tag association module 405 establishes an association between a sentence and a tag, which may be performed by a software program or assisted by manual correction. For example, the system may provide an operation interface to display all tags, and the personnel responsible for modifying and setting the tags may assign one or more tags to each sentence, or may not have the appropriate tags, and then save the set result in a database or a specific file for other modules, such as the neural network format conversion module 406.

When various public data retrieved from the network are stored in the database, the system divides the public data into sentences to be stored in the database, and the data for setting the accommodation place tags are also stored in the database. The data in the database is then exported and passed through the neural network format conversion module 406 to form the format required for training the neural network, such as CSV (text format with data separated by punctuation marks, TAB, space, etc.) or plain text format.

For the labels needed by the accommodation places stored in the database, the system maintains the correctness of the labels through a software method, so that a mechanism for adding, removing or modifying the labels is provided, and the mechanism comprises system updating, manual updating or correction through a neural network training result. When a label is set, when there is a need for label adjustment 423, for example, a new label is added or an old label is to be deleted, the adjusted label may be input to the label corresponding module 405 through a label management module 411, and may be handed to the neural network training and verifying module 407 through the neural network format conversion module 406, so as to retrain the neural network model, so that the neural network may correctly output the required label.

According to one embodiment, after the neural network format conversion module 406 generates the data for training the neural network, the data may be provided to the neural network training and verification module 407 for training to establish the neural network model by the neural network model generation module 408. If the training data is not easy to obtain, part of the data generated by the neural network format conversion module 406 may be used as training data, and the rest may be used as test data, and the purpose of the test data is to provide the data to the neural network model evaluation module 409 for use after the neural network training and verification module 407 completes the training of a neural network model, so as to test the accuracy of the generated model.

Further, in the neural network training and verifying module 407, a neural network system is set up by a software method, the required neural network algorithm, word bank, preprocessed data, program code, storage space, computing resources, network resources and the like are prepared and transmitted to the neural network, the neural network model generating module 408 trains the neural network model, after the training is finished, test data are provided for the trained neural network, feedback information generated after the user uses the training data is added, the parameters of the neural network are continuously adjusted, the accuracy of the neural network model for the test data can meet the requirement as confirmed by the neural network model evaluation module 409, the mode for adjusting the neural network membrane type can also comprise a manual mode, namely whether the label generated by the neural network for the input sentence meets the requirement or not can be checked according to the feedback information of the user.

When the generation result of the label by the neural network model evaluation module 409 does not meet the requirement, and the accuracy of the model is not sufficient, the label needs to be optimized by the accommodation site label optimization module 410, and the optimized label is provided to the label management module 411 to adjust the label, and the label corresponding module 405 is performed, and the subsequent action is performed. If the labels are too close to cause inaccurate classification, similar labels need to be merged; if the training sentence database required by the label is lacked, sentences required by more labels can be prepared; if the required information cannot be grabbed, the label needs to be considered to be removed.

Then, the system generates labels used by the accommodation places through the label generation module 412, and transmits the public data about the accommodation places to the trained neural network-like model, so as to obtain the labels corresponding to each accommodation place, thereby obtaining the corresponding relevance between the hotel and the labels, and storing the relevance in the database, such as the label database 413 in this example.

After the neural network-like model generation module 408 calculates the accuracy of the test data, if the accuracy effect is not satisfactory, the algorithm and parameters of the neural network may be adjusted at module 7, or the tags may be adjusted at the accommodation site tag optimization module 410, or more information may be obtained by the public data acquisition module 403, or other segmentation methods may be used at the data segmentation module 404, or the corresponding links between sentences and tags may be changed at the tag correspondence module 405.

On the other hand, when the neural network-like model generated by the neural network model generation module 408 meets the requirements, the model is trained. The requirement is not a certain value, and the user can customize the target, for example, the target can be met only when the order accuracy exceeds 95%, or the target can be met only when 90% is reached. The trained neural network-like model may be provided to the tag generation module 412 for use in generating a correspondence table of accommodation locations and tags stored in the tag database 413.

Furthermore, after the neural network-like model generated by the neural network-like model generation module 408 is completed, if there is new comment information, the new comment can be segmented into sentences by the data segmentation module 404, and the sentences need not be input into the label correspondence module 405, but the sentences are input into the neural network-like model generated by the neural network-like model generation module 408, so as to directly generate labels corresponding to the new comment, and the labels are stored in the label database 413.

It should be noted that the label database 413 records a positive label and a negative label of each network item. Taking the accommodation place as an example, the front labels are such as "" sound insulation good "", "" service good "", "" near airport "", and "" near business circle ""; negative labels are, for example, "" poor sound insulation "", "" poor service "", "" poor equipment "", and "" smart "".

After completing each network object and the corresponding label, the user can obtain the front label of each network object through a software program (such as APP) in a platform or a mobile device, and when the user selects the front label, the user indicates that the network object to be displayed needs to have the labels, which becomes a way of screening the network objects.

In this example, the user is shown as a lodging place, when looking for the lodging place 424, the user can use the tags to filter the lodging places 425, so that the user can set some filtering conditions, such as duration of stay, country, city, price, facility (these conditions can be all the components of part of tags), the lodging places can be filtered by the tags provided by the system, if the tags are positive tags, the user only displays the lodging places with the positive tags; on the contrary, for the negative label, for example, "demon", there is no label "without demon" as training data for the neural network when training the neural network, so the neural network can only help to stick the label "demon" to the lodging location with the problem of demon, but can not help to stick the label "without demon" or "very safe" to the lodging location without the problem of demon. Therefore, the platform provides a "" no-exception "" option to the user, which indicates that the system will exclude the lodging places with "" exception "" tags when the user clicks "" no-exception "".

Then, the user selects various screening conditions including location, country, budget, date, score and label on the online booking platform to select the accommodation location 426 meeting the conditions, and the online booking process is completed.

Since the evaluation of a lodging location may change over time, for example, a freshly opened lodging location, a consumer may evaluate that "the device is new, but the service personnel are not trained enough, and many devices are not yet opened", but after 10 years of operation at the same lodging location, the consumer may evaluate that "the device is a little old, the wallpaper is dropped, but the service personnel is well-attitude, and the devices are old, but each item can be used". Therefore, the system needs to continuously capture the evaluation, discussion, sharing or rating related to the accommodation location update, and the system can continuously or periodically retrieve the public data 422 related to various accommodation locations from the network through the public data retrieving module 403 to adjust the tags of the accommodation locations in the tag database 413.

Furthermore, the system will continuously update the database, when a new accommodation place is added (421), the public data of the accommodation place is obtained through the software modules, and is provided for the neural network model, the corresponding label is obtained, and the label is stored in the database, so as to provide the latest and most applicable label information for the user.

In other cases, when there is a need for tag adjustment 423, the system has to perform sentence-to-tag re-correspondence and retrain the neural network with the new tags. For example, when 1 tag is added, since the original neural network has no correspondence between sentences and new tags, the tag correspondence module 405 is executed after the new tag is added, and the subsequent actions are executed. After deleting 1 tag, the stored database or specific file can be removed, and the data is exported to the neural network-like format conversion module 406, and the subsequent actions are executed. Alternatively, if the neural network is to avoid attaching the deleted tag to the sentence, the tag correspondence module 405 may be executed, and the subsequent action may be executed.

According to the embodiment of the network object tag management system applied to the neural network shown in fig. 5, the network object tag management system continuously or periodically obtains the public data 51 corresponding to various network objects by a software method through a network software program, for example, capturing the video or text contents of the relevant accommodation places disclosed in the websites, blogs, community media and discussion areas, and the public data 51 includes the evaluation, rating, discussion and introduction corresponding to each network object and is structurally stored in the database 52.

The public data 51 stored in the database 52 can be input into the trained neural network-like model 53, and one or more tags associated with new network objects can be automatically obtained to form a tag database 54, so that the network object tag management system can query the user 501 through the network platform 55 and screen interesting network objects through tag attributes, such as lodging places as proposed in the above embodiments.

On the other hand, if the public data 51 stored in the database 52 is new or updated, the public data can be inputted into the neural network algorithm 56 for calculation, and the labels associated with the network objects, including one or more labels associated with the network objects for updating, deleting, adding or maintaining, are learned from the database 52, so as to verify and optimize the neural network model 53, and maintain the labels provided by the system in the correct state.

Correspondingly, when the analog neural network model 53 established by the analog neural network generates a label judgment result, whether the result meets the expectation or not can be judged through manual intervention, when the result does not meet the requirement, the analog neural network model 53 needs to be adjusted, and the related optimization process can optimize the result in a mode of adjusting an information source, adjusting a word segmentation algorithm, adjusting a label, adjusting an analog neural network algorithm 56 in the analog neural network and the like. By using the neural network model 53, taking the accommodation place as an example, the mechanism of providing the label by the network object label management system can be provided for the user to query and filter the accommodation place by using the label, and the system can filter the accommodation place according to the label and other basic conditions (such as the number of rooms, the number of people, the check-in period, the price and the equipment) and display the accommodation place meeting the user requirement.

Further, in addition to optimizing the neural network model 53 with new network objects and their associated public data or new public data related to network objects, when a user queries and filters network objects through the network platform 55, if feedback information 57 generated by the user is received, such as providing tag suggestions or error reports, the feedback information 57 will form a reference for modifying the neural network model 53.

According to the embodiment, taking the accommodation place as an example, the usage of the neural network-like model 53 completed by the system training includes: firstly, when a user queries a specific network object through the network platform 55, the user can input a natural sentence through a user interface provided by the network platform 55, the system transmits the natural sentence input by the user to the neural network model 53, and the neural network model 53 can transmit a corresponding label back through the network platform 55; secondly, the platform provides tags as screening conditions, and after the user checks the tags, the lodging places meeting or not meeting the tags can be screened; thirdly, as new public data such as comments, opinions, discussions, shares and the like are continuously provided, the public data are extracted and transmitted to the neural network model 53 in the system, and tags corresponding to the new data are obtained, and at the moment, the system gives new tags to the accommodation places. The evaluation of the accommodation places can lead consumers who live in at different times to have different evaluation and feelings due to various factors such as management teams, equipment maintenance, equipment increase or decrease, service adjustment, personnel transaction, position relocation, emergencies, activities and the like of the accommodation places, and the system obtains the latest tags by continuously capturing new information to the neural network model 53, and then leads the specific accommodation places to be linked to the latest tags, so that when the users screen the accommodation places by the tags, the screening condition is the latest evaluation opinions of the consumers in the accommodation places; fourth, the network platform 55 uses the tag as a lodging location screening method, which solves the conventional problems of too much information, too much noise, difficulty in reading, no concentration of information, flooding of false evaluation, outdated information, unfixed information format, and the like.

The flow shown in the following figures takes the accommodation place as an example, and the operation manner of the adopted software function module can refer to the content in fig. 4, which is not described herein again.

Fig. 6 is a flowchart illustrating an embodiment of a process for training a neural network, taking as an example a tag management system for an accommodation site.

Starting from the basic data of the network objects, taking the accommodation places as an example, the network objects include data 421 of the accommodation places and software programs for acquiring data about each accommodation place at least by the accommodation place information management module 401 and the accommodation place information extraction module 402, and establish information about each accommodation place in the databases in each system.

The stage of processing public data follows, where one of the main sources of public data about the accommodation sites is public data 422 generated by the accommodation sites themselves, and what can be evaluated is that the software program implemented by the public data retrieval module 403 obtains information about each accommodation site from each website. The data segmentation module 404 is configured to preliminarily process the captured public data, perform semantic analysis on each sentence by segmentation, and perform tag check by the tag correspondence module 405, so as to give a meaningful sentence-corresponding tag, and thus establish a corresponding association between a sentence and a tag. The neural network format conversion module 406 then converts the data stored in the database into a format required by the training neural network to generate data for the training neural network.

Thereafter, at the stage of processing the tags, the tag setting correspondingly completed by the tag correspondence module 405, including the adjusted tags, is generated by the tag management module 411 to form tags for each accommodation place.

In the stage of the neural network, the data formed by the neural network format conversion module 406 is sent to the neural network training and verification module 407 for training the neural network model, and the related data includes the training data formed by the neural network format conversion module 406 and can be converted into the testing data. The neural network-like model generation module 408 is used to build a neural network-like model, which is then provided to the neural network-like model evaluation module 409 for testing, and repeated testing can be performed according to the feedback (e.g., manual inspection) of the result to generate a model with an accurate judgment tag. Finally, the neural network-like model is optimized by the software method of the accommodation site tag optimization module 410.

FIG. 7 illustrates an embodiment of a process for generating labels using trained neural network analysis of public data, using the label management system of an accommodation site as an example.

After the neural network-like model is completed, the public data 422 of the accommodation place and the public data obtained from various public sources by the public data acquisition module 403 can be used, similarly, segmentation, paragraph analysis and semantic analysis of natural language are performed by the data segmentation module 404, the neural network-like model generation module 408 establishes a neural network-like model for determining paragraph labels in the public data, the labels used by the accommodation place are generated by the label generation module 412, and then the data related to the accommodation place and the labels are stored in the label database 413.

After the tag database 413 is completed, the relevant application implements an embodiment of the process for the customer to select the accommodation site, as illustrated in fig. 8 by taking the tag management system of the accommodation site as an example.

When a user searches for an accommodation site (424) through the network article tag management system disclosed in the present specification, the system queries various tags about an interested accommodation site in the tag database 413, so that information about the accommodation site evaluated by various public data can be obtained, and the accommodation site (425) can be screened by using the tags to select an accommodation site (426) meeting the conditions.

On the other hand, the network object tag management system still continuously performs the operations of retrieving public data, analyzing semantics, updating tags, and the like, and particularly, automatically determines and establishes a tag database by using the neural network-like model, and fig. 9 illustrates an embodiment of a process in which the neural network-like model needs to be retrained when the tag management system of the accommodation site is taken as an example for tag adjustment.

When the system determines that there is a need for label adjustment (423), for example, when it determines that data is outdated and new information needs related to the lodging location are obtained, a need for deleting an old label and a new label may be generated, at this time, the system still takes out the label to be adjusted and corresponding data through the label management module 411, inputs the label and the corresponding data into the label corresponding module 405, passes through the neural network format conversion module 406 and then passes to the neural network training and verification module 407, so as to retrain the neural network model, the neural network model generation module 408 updates the neural network model, and provides the model accuracy for the neural network model evaluation module 409 to test, confirms that the set accuracy has been reached, and then optimizes the neural network model by the software method of the lodging location label optimization module 410.

In summary, the network object tag management system disclosed in the specification can perform tag management of the accommodation site, and can also be applied to other network objects, and the implementation method is not limited to the above embodiment.

Although the present invention has been described with reference to particular embodiments, it will be understood by those skilled in the art that various changes in form, construction, features, methods and quantities may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A network object label management method is characterized in that the network object label management method comprises the following steps:

analyzing public data about a network object, wherein each piece of public data has a timestamp;

obtaining one or more corresponding labels according to the meaning of each public data, and establishing the association between the network object and the one or more analyzed labels; and

storing the public data based on the timestamp and the one or more tags associated with the network object in a database for querying;

wherein a time threshold is set, and when the time of any public data related to the network object exceeds the time threshold, the public data related to the network object is retrieved to update, delete, add or maintain the one or more labels associated with the network object.

2. The method as claimed in claim 1, wherein the public data of the network object includes video or text content published in websites, blogs, social media and forums.

3. The method as claimed in claim 2, wherein a network software program running on a network retrieves video or text content published in websites, blogs, social media and forums, and stores them in the database in a structured manner.

4. The method as claimed in claim 3, wherein a computer software program parses the public data of the network object sentence by sentence or paragraph by paragraph, and determines whether the one or more tags correspond to each sentence or each paragraph, or whether none of the tags correspond to any of the tags.

5. The method as claimed in claim 4, wherein the public data is analyzed by a natural language semantic analysis method, and the attributes of the tags are compared to determine the one or more tags corresponding to each sentence or paragraph in the public data.

6. A method for label management of network objects according to any of claims 1 to 5, wherein a class of neural network models are created by learning from said database a plurality of said labels associated with a plurality of said network objects with a class of neural network algorithms.

7. The method as claimed in claim 6, wherein when public data of a new network object is obtained, the public data of the new network object is inputted into the neural network-like model, and the one or more tags associated with the new network object are automatically obtained.

8. The method of claim 6, wherein the one or more tags associated with the network object are updated by the neural network-like model when new public data about the network object is retrieved.

9. The method of claim 6, wherein when a new network object and its associated public data are obtained or new public data about the network object are obtained, one or more tags associated with the new network object are obtained or updated by the neural network-like model, and the neural network-like model is modified according to a feedback message.

10. A network article tag management system, the network article tag management system comprising:

a database for storing public data obtained from a network and relating to a plurality of network objects, and analyzing each public data to obtain one or more labels related to each network object; and

a server including a processor and a memory, the processor executing a network object tag management method, comprising:

parsing the public data for each of the network items, wherein each of the public data has a timestamp;

obtaining the corresponding one or more labels according to the meaning of each public data, and establishing the association between each network object and the analyzed one or more labels; and

storing the public data based on the time stamp and the one or more tags associated with each of the network items in the database for querying;

11. The system as claimed in claim 10, wherein the public data of the network object includes video or text content published in the website, blog, community media and forum, and a network software program running on the network retrieves the video or text content published in the website, blog, community media and forum and is structurally stored in the database.

12. The system of claim 11, wherein a computer software program parses the public data about the network object sentence-by-sentence or paragraph-by-paragraph to determine whether each sentence or paragraph corresponds to the one or more tags or whether none of the tags corresponds to the one or more tags.

13. The system of claim 12, wherein the public data is parsed by a natural language semantic analysis method, and the one or more tags corresponding to each sentence or paragraph in the public data are determined by comparing the attributes of the tags provided by the server.

14. A system for network item tag management according to any one of claims 10 to 13, wherein said system further comprises an intelligent module, wherein a neural network-like algorithm is run, and a neural network-like model is created by learning from said database a plurality of said tags associated with a plurality of said network items by said neural network-like algorithm.

15. The system according to claim 14, wherein when a new network object and its associated public data are obtained, or new public data about the network object are obtained, one or more tags associated with the new network object are obtained by the neural network-like model, or the one or more tags associated with the network object are updated, and the neural network-like model is modified according to a feedback message.