CN111143653A

CN111143653A - Credibility verification method for mass science popularization resources

Info

Publication number: CN111143653A
Application number: CN201911374626.8A
Authority: CN
Inventors: 刘小华; 胡文心
Original assignee: Shanghai Qidisheng Network Technology Co ltd
Current assignee: Shanghai Jietu Intelligent Technology Co ltd
Priority date: 2019-12-27
Filing date: 2019-12-27
Publication date: 2020-05-12
Anticipated expiration: 2039-12-27
Also published as: CN111143653B

Abstract

The present invention relates to the field of databases and data analysis. The method of the invention comprises the following steps: the method comprises the steps of combining a data crawling technology to acquire, sort, clean, screen, classify, fuse and the like the existing mass science popularization data to realize knowledge construction, and building different resource directory systems aiming at different science popularization resources to form a resource information base; generating credibility values of the resources aiming at different resources by specific rules according to a series of key information such as the sources, organizations, authoritativeness of authors and the like of the resources in the library; building an expert resource library by means of professional forces of science popularization experts in different fields, and forming characteristic values of corresponding experts; and matching the classified data with popular science experts in the corresponding field by using a collaborative crowdsourcing technology according to the rule, thereby performing cross identification in modes of scoring, evaluating and the like. The invention can realize high credibility authentication for mass science popularization resources, thereby building a science popularization professional resource library for collecting public trust resources.

Description

Credibility verification method for mass science popularization resources

Technical Field

The invention relates to the field of databases and data analysis, in particular to a credibility authentication method for mass science popularization resources.

Background

At present, compared with the traditional science popularization mode with more authoritative science popularization content and insufficient updating rate, the science popularization resource released and spread in the internet has the characteristics of larger information amount, quicker updating and wider content. The social network is different from military and is developed vigorously, so that the social network has huge user quantity and becomes one of the most important new media platforms, massive user groups generate massive information which is rapidly spread by virtue of the social network, and the information can be published in the internet at any time no matter professional organizations, related enterprises or individuals, so that the science popularization information in the network is huge in quantity and rich and diverse in content, covers different disciplines, different fields, different regions and different languages, and comprises various science popularization resources such as texts, images, sounds, videos and the like in form. Most information in the network is not authenticated by an authority department or an expert, so that the popular science information in the network has the characteristics of complexity, disorder, uneven quality, weak credibility and the like, and difficulty is caused in information selection. The 'pseudo science popularization' allows social network users to have a large demand for science popularization knowledge, and spreading characteristics of civilization, strong interactivity and the like of social network information production are utilized to flood the network. For individuals, the information causes many personal and property safety losses, and for society, the information also causes many threats in public opinion safety, cultural development and social stability.

A new cloud technology, an AI technology and a data mining and analyzing technology based on big data put forward new requirements for development of science popularization on one hand and promote rapid advance of science popularization upgrading on the other hand. In order to provide scientific, authoritative and accurate science popularization information content for users, in order to eliminate negative effects of 'pseudo science popularization' on individuals and society as far as possible, a large number of science popularization resources in the internet are fully utilized, and a credibility authentication method for researching a large number of science popularization resources is urgently needed.

Disclosure of Invention

The invention aims to provide a method for carrying out credibility certification aiming at mass science popularization resources in a network, and more valuable science popularization resources are effectively screened.

The embodiment of the invention discloses a credibility authentication method for mass popular science resources, which comprises the following steps:

extracting and converting the data crawled by the resources, further cleaning, extracting and the like, acquiring key information such as website source, manufacturing mechanism and author, time attribute, resource theme and the like, performing structured storage, establishing an index according to a management rule, forming a fusion data directory of a resource library, and providing data support for various services, thereby constructing a resource attribute table;

after the resource attribute table is established, the resources are further required to be analyzed, classified and sorted, and credibility values of the corresponding resources are generated according to key information such as authority of sources, organizations and authors in the attribute table according to a certain rule;

based on the data characteristics of science popularization resources in different fields, expert characteristic values are formed according to rules such as expert categories, adept fields and the like by relying on the prior science popularization related professional institutions such as a science popularization expert library and the like and the expert resource library;

and according to a series of matching rules such as the constructed resource attribute table, the credibility value, the expert characteristics and the like, the classified popular science resource data are subjected to cross scoring and evaluation by corresponding popular science experts by using a collaborative crowdsourcing technology, and resource authentication is carried out.

Optionally, the science popularization resource collection and data fusion, and the establishment of the resource attribute table include:

when massive Chinese and Chinese science popularization resources which are scattered on the Internet nationwide or all over the world and produced by a large number of science popularization professional institutions or network platforms, enterprises and public institutions, scientific research institutions, experts from media and the like are subjected to network crawling, institutions, categories, authors, release time and article sources of the resource website are recorded;

establishing a label attribute table according to the acquired organization and author of the popular science resource, the organization to which the author belongs and the article source;

and establishing a corresponding relation between attributes in the resource attribute table according to the overall distribution condition of the collected science popularization resources.

Optionally, the establishing of the correspondence between the attributes in the resource attribute table includes:

counting the attribute distribution of popular science resources and the distribution condition of labels by means of big data analysis and professional classification;

and establishing a relation chart among the labels according to the resource attributes.

Optionally, generating a trustworthiness value of the corresponding resource includes:

obtaining authority values of resources according to authority degrees of tag values such as source mechanisms, provided authors and the like of the science popularization information;

and generating a credibility value of the resource according to the authority value corresponding to the resource.

Optionally, the creating an expert resource library, and the forming of the expert attribute table of the category and the excellence area includes:

collecting, sorting and classifying information such as expert academic calendars, work experiences, work fields, papers, monographs and publication situations and the like to construct an expert resource library;

and setting labels of corresponding categories, good areas and the like for experts in the expert database according to the expert database access standard which is unified with the science popularization resource database.

Optionally, crowd-sourcing the resource to the expert in the expert resource library for authenticating according to the resource attribute table and the credibility value comprises:

pushing the resource information with the credibility value of the popular science resource lower than 80% to corresponding experts in an expert resource library for cross authentication in modes of scoring, evaluating and the like;

and determining the number of the persons pushing the expert identification according to the credibility value, wherein the credibility value is inversely proportional to the number of the experts. .

Compared with the prior art, the implementation mode of the invention has the main differences and the effects that:

according to the invention, through analyzing the resource sources, the related science popularization resources which are originally created or recommended by an authority, a professional expert or a scholars are identified as credible resources, so that the user can use the resources at ease.

The method analyzes the resource sources, crowdedly packages the popular science resources made by non-authoritative organizations and non-professionals to professionals for authentication, highlights accurate and excellent resources, marks wrong and fake popular science resources and enables users to be identified.

Drawings

FIG. 1 illustrates a flow diagram of a method of trustworthiness authentication in accordance with an embodiment of the present invention;

FIG. 2 shows a flow diagram of resource acquisition and fusion.

Detailed Description

In order to make the purpose and technical solution of the embodiments of the present invention clearer, the technical solution of the embodiments of the present invention will be clearly and completely described below with reference to the drawings of the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the described embodiments of the invention without any inventive step, are within the scope of protection of the invention.

It should be noted that, although a logical order is illustrated in the flow chart, in some cases, the steps shown or described may be performed in an order different than that illustrated.

Fig. 1 shows a flowchart of a credibility certification method according to an embodiment of the present invention.

Wherein the method 100 comprises:

step 101, carrying out knowledge construction in the modes of acquisition, arrangement, fusion and the like of mass science popularization data, and building different resource directory systems to form a resource attribute table;

102, generating credibility values of the resources aiming at different resources by specific rules according to a series of key information of the resources in the library;

103, building an expert resource library by means of the professional power of science popularization experts, and forming characteristic values of corresponding experts;

and 104, matching the classification data with popular science experts in the corresponding field by using a crowdsourcing technology according to the rule, and performing cross identification in modes of scoring, evaluating and the like.

When crawling resources, the method records key information such as website source, production organization, author, time attribute and the like, fuses popular science resource information of multi-channel source through data fusion processing, extracts symptom information, matches the symptom with knowledge in a popular science resource library through designing a system inference machine, and makes data classification decision. And a self-learning module can be added in the classification system based on the resource information fusion, the classification information is fed back to the resource base through the self-learning module, so that the resource base is updated, meanwhile, the self-learning module can carry out reasoning according to knowledge in the resource base and dynamic response of a user to system questions, so as to obtain new knowledge, summarize new experience, continuously expand the knowledge base and realize the self-learning function of the resource system.

In addition, the system also needs to perform relevant analysis on the resource issuing main body to judge whether the source belongs to a professional institution or not, whether the resource issuing main body passes through a platform authenticated by the institution or not, and the like. A uniform data access rule is required to be established, and a first class credibility value is formed by automatically scoring the resource source mechanism; the invention further analyzes the manufacturing unit of the resource information, if the science popularization information issued by a department with the name of a normal hospital institution is deployed, a uniform data access rule is also required to be established, and system scoring is carried out according to the level of the department of the unit and the subordinate medical and health institutions from which the resource information comes, so as to form a second type credibility value; the invention also analyzes the provider of the resource information, for example, whether the collected information has authority and influence, and the third credibility value of the popular science resource can be formed by correctly scoring and evaluating the collected information by a uniform access rule through a big data analysis technology.

For resources with high comprehensive credibility, the system automatically marks the high-credibility resources. For the resources with low comprehensive credibility, the system crowdsourcing the resources to the experts in the corresponding fields in the constructed expert resource library for cross authentication. The crowdsourcing system has scientific task design and distribution strategies and crowdsourcing quality control strategies, and can enable experts to be efficiently communicated with a system platform according to the flow and the progress of crowdsourcing activities, so that the crowdsourcing system integrates the resultant force of field experts, and realizes the identification and the authentication of popular science resources.

For a particular authentication task, it is desirable to obtain high quality task results as quickly as possible with as little effort as possible. Around this problem, two aspects are considered in the collaborative crowd sourcing of techniques for trusted authentication. Firstly, from the perspective of the quality of the resource authentication result, the result quality of the authentication task is predicted through a model and an algorithm, and the working quality of crowdsourced experts in the task is automatically judged, so that the results such as false authentication and the like caused by the intention or the accident of some experts in the task can be avoided. The angle is actually to evaluate the answers returned by the workers and also to score the quality of the questions made by the workers themselves. And secondly, from the perspective of designing crowdsourcing tasks, the question-making quality and the question-making interest of workers are improved by carrying out more popular and understandable description on the certification tasks, designing a better user interface, designing an effective incentive mechanism and the like, so that the efficiency of the expert certification process is improved in the crowdsourcing certification process, and the certification quality is ensured.

The invention also has special strategy for resource distribution in crowdsourcing. For the resource with relatively high credibility judged by the system, less experts are allocated for performing authentication tasks, and for the resource with relatively low credibility, more experts are allocated and cross authentication is performed.

According to the invention, the crowdsourcing quality is controlled by asking part of tasks of marking gold labeling data sets by experts in advance, the gold standard data sets are used as test problems, the problems are inserted into the tasks to control the authentication quality of the experts, and for the working results provided by the experts with low working quality, a mode of multiple times of authentication can be adopted to reduce the error rate. For tasks that do not have a "gold annotation dataset," the simplest approach is to assign a question to an odd number of experts and then decide which is the correct answer by majority voting. Most voting principles, however, assume that the answer accuracy rates of each worker are consistent, and in fact the answer accuracy rates of the workers are often different. In order to improve the task completion quality, the invention adopts a quality control algorithm combining a plurality of control methods, integrates an error loss value calculation method based on a minimum expected error and a vector, predicts the task completion quality according to the uncertainty of a worker on task judgment, and achieves the control effect from the aspect of the worker.

FIG. 2 shows a flow diagram of a data fusion method according to an embodiment of the invention.

Wherein, the method 200 comprises:

step 201, adopting a common web page crawling engine for a static web page, and adopting a data crawling engine combining various crawling modes such as a deep network and the like for a dynamic web page to obtain target resource data;

step 202, extracting symptom information of resource data, matching the symptom information with the existing resources through an inference engine, and making a data classification decision;

step 203, feeding back the classified resource information to the existing resource library through the self-learning module, modifying the corresponding confidence coefficient factor, updating the resource library, and performing a series of processing to complete data fusion;

and a common web page crawling engine is adopted for the static web pages, and a data crawling engine which combines various crawling modes such as a deep network and the like is adopted for the dynamic web pages, so that target resource data are obtained.

In the network directional crawling technology, a common web crawler crawls common web pages and selects one of three modes of depth priority, breadth priority and optimal priority according to different target websites. When the depth priority is adopted, the web crawler tracks the links one by one from the initial URL, and then transfers to the next initial page after the link is processed, and continues to track the links. When breadth is first, the web crawler searches the next level after finishing the current level of search, or crawls the web page by the breadth first strategy, and then filters the web page irrelevant to the web page. And when the best priority is adopted, predicting the similarity between the candidate URL and the target webpage or the correlation between the candidate URL and the target webpage according to a certain webpage analysis algorithm, and selecting one or more URLs with the best evaluation for crawling. The information crawling strategy of the science popularization resource library needs to be improved by combining with specific application so as to call out a local optimal point. And filtering the links irrelevant to the subject according to a certain webpage analysis algorithm, reserving useful links, and putting the useful links into a URL queue waiting for grabbing. Then, it will select the next web page URL from the queue according to a certain search strategy, and repeat the above process until reaching a certain condition of the system. The deep network crawling mainly crawls high-quality information, and has the characteristics of huge quantity, good quality and high value. Through deep network crawling, the efficiency and the quality of information acquisition can be greatly improved, so that the technology combining various crawling technologies is the strategy required for constructing the original information of the popular science resource library. And finally, all the information captured by the crawling engine is stored by the system, certain analysis and filtering are carried out, and the information is structurally stored in the database according to different resource fields and the establishment of response indexes.

The invention collects, transmits, synthesizes, filters, correlates and synthesizes useful information given by various information sources through a data fusion technology so as to assist people in situation/environment judgment, planning, detection, verification and diagnosis. And extracting symptom information of the resource data, and matching the symptom information with the existing resources through an inference machine to make a data classification decision.

The method carries out processing through data fusion technologies such as Bayes inference, voting, D-S inference, neural network fusion and the like according to different conditions.

The classification information of the invention is fed back to the existing resource library through the self-learning module, and the corresponding confidence coefficient factor is modified, the resource library is updated, and the data fusion is completed.

It is noted that in the claims and the description of the patent, relational terms such as first, second, and third, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, the use of the verb "comprise a" to define an element does not exclude the presence of another, same element in a process, method, article, or apparatus that comprises the element.

While the invention has been shown and described with reference to certain preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.

Claims

1. A credibility verification method for mass science popularization resources is characterized by comprising the following steps:

the method comprises the following steps of carrying out knowledge construction in the modes of acquisition, arrangement, fusion and the like of mass science popularization data, and building different resource directory systems to form a resource attribute table;

generating credibility values of the resources aiming at different resources by specific rules according to a series of key information of the resources in the library;

constructing an expert resource library and forming a characteristic value of a corresponding expert;

and matching the classified data with popular science experts in the corresponding field by using a crowdsourcing technology according to the rule, thereby performing cross identification in modes of scoring, evaluating and the like.

2. The method according to claim 1, wherein the knowledge construction is performed in the modes of acquisition, arrangement, fusion and the like of mass science popularization data, different resource directory systems are set up, and a resource attribute table is formed, and the method comprises the following steps:

the method comprises the steps of utilizing a deep network crawling technology to obtain a large amount of Chinese and Chinese popular science contents which are distributed on the Internet nationwide or all over the world and produced by a large number of popular science institutions or network platforms, enterprises and public institutions, scientific research institutions and experts through media, carrying out big data analysis and professional classification through sorting, cleaning, screening, classifying, fusing and the like, and establishing a resource attribute table.

3. The method of claim 2, wherein the science popularization resource collection and data fusion, establishing a resource attribute table, comprises:

combining multiple Web crawling technologies such as a deep Web crawler and an incremental Web crawler, collecting science popularization webpage information, downloading a Web page, tracking hyperlink pages contained in the webpage layer by layer, acquiring and recording key information such as a mechanism to which a resource website belongs, a category to which the resource website belongs, an author, release time, an article source and the like, and filtering, classifying and clustering the information to realize construction of original information of a science popularization resource library;

establishing a popular science resource label attribute table according to the key information in the constructed original resource library;

and establishing the corresponding relation between the attributes in the attribute table according to the overall distribution condition of the resources.

4. The method of claim 3, wherein establishing the corresponding relationship between the attributes in the attribute table according to the overall distribution of the resources comprises:

taking source mechanisms of all resources, mechanisms providing authors and affiliated mechanisms of the authors, article sources, article keywords and the like as tags for statistics;

and establishing a relation chart of the labels according to the resource attributes.

5. The method of claim 1, wherein generating a confidence level value for a resource in the library for different resources according to a set of key information about the resource with a specific rule comprises:

and further analyzing, classifying and sorting the resource attribute table, and generating a credibility value of the corresponding resource according to key information such as the authority of the source, the organization and the author in the attribute table by a certain rule.

6. The method of claim 5, wherein generating a trustworthiness value for a resource comprises:

according to the authority degree of the tag values such as sources, organizations and authors, carrying out quantification to generate authority values;

and according to the authority value quantized by the resource attribute and the label, generating a credibility value of the resource.

7. The method of claim 1, wherein constructing an expert resource library, forming expert eigenvalues according to rules, comprises:

based on the data characteristics of science popularization resources in different fields, expert characteristic values are formed according to rules such as expert categories, adept fields and the like by relying on the prior science popularization related professional institutions such as a science popularization expert library and the expert resource library.

8. The method of claim 6, wherein constructing an expert repository to form expert eigenvalues according to rules comprises:

acquiring data such as expert academic calendars, work experiences, work fields, papers, monographs and publication situations and the like, and constructing an expert resource library through sorting and fusion;

and (4) establishing an expert resource library access standard which is unified with the science popularization resource library, and setting labels such as categories, areas of good interest and the like for experts in the expert library.

9. The method of claim 1, wherein the cross-discriminating between the classification data and the science popularization experts in the corresponding field by using a crowdsourcing technique according to the rule to perform scoring, evaluation and the like comprises:

according to the collaborative crowdsourcing mode, the joint force of experts in the corresponding field is collected through a task design and distribution strategy and a crowdsourcing quality control strategy, so that the identification and authentication of popular science resources are realized;

and according to a series of matching rules such as the constructed resource attribute table, the credibility value, the expert characteristics and the like, cross scoring and evaluating the classified popular science resource data by using a crowdsourcing technology for the corresponding popular science experts, and carrying out resource authentication.

10. The method of claim 1, wherein the step of cross-scoring and evaluating the classified science popularization resource data by a crowd sourcing technique for corresponding science popularization experts to perform resource authentication comprises:

the science popularization professional resource library platform automatically pushes resources with the credibility value lower than 80% to corresponding experts in the expert resource library for authentication according to the matching of the attribute of the resources and the characteristics of the experts;

and determining the number of the experts in classified pushing and authentication according to the credibility value, wherein the credibility value of the resource is inversely proportional to the number of the experts.