RU2683482C2 - Method of displaying relevant contextual information - Google Patents

Abstract

FIELD: information technology.SUBSTANCE: invention relates to a method and system for displaying relevant data and advertising on computer devices. Method for displaying relevant information based on content on a computer device includes: connecting a computer device to a media server; receiving content from a media server on a computer device; displaying the contents of the computer device user; downloading content to the ad server; establishing connection of the computer device with the advertising server; generating relevant data blocks based on categories, content part and target parameters, where the target parameters include information about the user that contains social profile data of the user on social networks; content analysis by the advertising server; generating attributes associated with content elements, based on content analysis; recognition of a part of the text content selected by the user of the computer device; removal of stop words from the content; identifying attributes associated with the content elements associated with the content and the content selected by the user; determining the potential goals of the user of the computer device; displaying to the user of the computer device potential user targets for user selection; determining the target chosen by the user of the computer device; application of an associative matrix with a predetermined associative model to the target selected by the user of the computer device; selecting at least one relevant data block corresponding to the target of the user of the computer device; loading the identifier of the relevant data block onto the computer device from the advertising server; generating a relevant data block using a predetermined pattern and displaying the relevant data block on a computer device screen.EFFECT: technical result is providing relevant information to the user of the computer device based on the content selected by the user and the user's objectives.15 cl, 14 dwg

Description

The present invention relates to computer and mobile devices, and in particular, to displaying relevant data and advertising on computer devices based on the context and the relevant part of the content selected by the user.

The existing market for advertising on computer and mobile devices was mainly formed by applying the technologies and principles of advertising used on personal computers and laptops to mobile platforms. However, the effectiveness of advertising on mobile devices is much lower compared to advertising on personal computers.

Mobile device users usually run special narrow applications instead of the heavyweight conventional browsers that take up a large amount of system resources. Thus, mobile users do not use regular search almost as often as personal computer users. Accordingly, the most popular methods of monetization on computer devices are relevant information blocks (such as advertising, banners, ad units, links to web pages, text messages, web pages, etc.) and applications embedded in web resources.

Relevant information blocks (relevant data blocks) are small ads that occupy parts of the screens of mobile devices, which are usually very small. Since the relevant information blocks are also very small, they contain very simple advertising data (i.e. short text). This simple advertisement basically annoys users and does not attract attention, and does not stimulate user interest.

The most commonly used method for mobile advertising is the use of advertising along with free media applications that carry ads - that is, advertising is essentially imposed on users who download a free application or game. This ad generates some small profits in line with the growth of the mobile device market, but in general is also ineffective.

The traditional way of advertising makes each advertisement different from the other, even if two advertisements advertise the same type of action, for example, “sell”. This advertising approach can be effective for static situations, when the client has enough time and is in the appropriate state of consciousness to view the advertisement. However, computer device users are on the go. These users do not want to spend a lot of time and effort to realize that the main point of two different advertisements is the same - "selling" products of different brands.

In situations of mobile users “on the go” it is much more important to show them the most relevant ads in a simple and easy to understand manner in order to achieve positive advertising results. Accordingly, it is desirable to display one or more relevant information blocks or banners associated with a particular context of the displayed data for any text or graphic data displayed on a user's computer device.

In addition, it is desirable to have relevant information blocks displayed based on user gender, user geographic location, user browsing history, media site statistics, website content, page metadata, etc. However, existing mobile advertising systems do not use relevant information blocks (or advertisements) based on context, because of the impossibility to extract the context itself, with the exception of browser-based advertising on search websites.

According to an illustrative embodiment, a method and system for displaying contextual advertising / information to mobile device users is provided. The advertisement used on the small screen does not require additional “shelf space”, since it is displayed within the existing GUI controls on the screens of mobile devices or on a separate overlay panel of the user interface.

According to an illustrative embodiment, the system may display one or more relevant information blocks associated with a specific part of the context of the displayed data for any text or graphic data shown to the user. Relevant information blocks can be shown based on target parameters, for example, the user's geographic location, user’s Internet browsing history, media site statistics, website part content, page metadata, public data of the user's social profile (age, gender, position, interests, friendship), tags related to the current user, provided by third-party services (DMP, DSP) and others.

A list of relevant information blocks is automatically generated either based on the content of the web page, or on the basis of the selected part of the text and its content. The text of the web page is pre-scanned and a call to the application programming interface with predefined parameters is added for each element of the content structure (i.e., paragraph, sentence, embedded image or video block). The functions of the application programming interface analyze this part of the content and activate the controls, which, in turn, can trigger the display of the relevant information block (s).

The illustrative embodiment advantageously uses a narrow context of content explicitly selected by the user. The relevant information block is displayed based on the interests of the user indicated by the content part, i.e. text, paragraph, word, image or part of an image, video, hyperlink and their content and context. In an illustrative embodiment, the relevant area adjacent to a portion of the content selected by the user is called a context. The selected content may be a word or a combination of words (expression), an image or video block area, and the context may include a sentence, paragraph or all text (in the case of small text), geometrically close images or video blocks.

In the figures:

FIG. 1 illustrates the display of relevant information blocks on a computer device in accordance with an illustrative embodiment;

FIG. 2 illustrates a system architecture in accordance with an illustrative embodiment;

FIG. 3 illustrates a flowchart of an accurate advertising selection method in accordance with an illustrative embodiment;

FIG. 4 illustrates a flowchart of an accurate advertising selection method in accordance with an illustrative embodiment;

FIG. 5, 6, 7, 8 illustrate the display of advertising and organic results on the screen of a computer device;

FIG. 9 illustrates the processing of content on a computer device using a list of actions for predetermined user requests;

FIG. 10 illustrates a flowchart of an accurate advertising selection method in accordance with an illustrative embodiment;

FIG. 11 illustrates screens of computer devices in accordance with an illustrative embodiment;

FIG. 12 illustrates the exact selection of advertising and a graphical representation of targeted advertising to a user;

FIG. 13, 14 illustrate flowcharts of algorithms Page VCD;

The technical result of the present invention is to provide relevant information to a user of a computer device based on user selected content and user goals.

According to an illustrative embodiment, a method and system for displaying context-sensitive additional information blocks or advertising to users of mobile / computer devices is presented. Advertising used on a small screen does not require any additional “shelf space” because it is displayed within the existing GUI controls or overlapping the GUI panel of a computer device’s screen. Unlike regular advertising, which involves a large amount of data traffic, the device library has predefined templates that are displayed to the user.

According to an illustrative embodiment, the system may show one or more relevant information blocks associated with a particular context and fuzzy selected items of displayed data for any text or graphic data shown to the user in FIG. 1. Relevant information blocks can be displayed based on the user's location, browsing history by the user, media site statistics, part of the website’s content, page metadata, public data of the user's social profile (ie age, gender, position, interests, friendship), tags related to the current user, provided by third-party services (DMP, DSP) and others.

It is worth noting that the relevant information unit may be in the form of additional information in various formats, which may be relevant to the user. For example, the relevant information block image may be text blocks (such as advertising banners, advertising relevant information blocks, links to web pages), media data, web pages, etc. Relevant information blocks are called advertisements. However, any relevant information may be displayed to the user.

A list of relevant information blocks is automatically generated based on the content of the web page or the selected part of the content (text, image, part of the text, part of the image, multimedia, etc.) and its context. In addition, any combination of target parameters can be taken into account to generate relevant information blocks. The web page context is scanned. The software application interface function calculates the content attributes with predefined parameters that are added for each element of the content structure. It is worth noting that the content attributes can be pre-calculated. The interface function of a software application conducts explicit (or implicit) associations of elements with their attributes. The software application interface analyzes the provided part of the text and activates the controls that initiate the transition to the relevant information block (s).

The illustrative embodiment advantageously uses a narrow context of content explicitly selected by the user. Relevant information blocks are displayed based on the interests of the user, the indicated part of the text or part of the content (or hyperlink and its content) and target parameters. In an illustrative embodiment, the relevant area adjacent to a portion of the text selected by the user is called a context. The selected content may be a word or a combination of words (expression), and the context may include a sentence, paragraph, or the entire text (in the case of short text).

According to an illustrative embodiment, users can notify the system of their interest in one or another part of the content, using, for example, an infrared camera, infrared sensor or video camera, where a sensor recorder (integrated or external device such as Kinect) replaces the physical “touch screen” and recognizes user gestures and associates them with the context on the screen. In addition, text can be selected with the mouse. In the case of using the mouse, the user can select part of the context (that is, part of the text or image) and use the right-click on the selected context and select "show relevant data block" in the menu. Movements and mouse clicks can be implemented using the keyboard. In addition, a special keyboard button can be used to display relevant data blocks. In addition, users can use a microphone, for example, Siri-like models (that is, a sensor with a background voice recognition module, which can be integrated, external or server-side).

The user indicates part of the context and his wishes through voice commentary, a kinesthetic detector (that is, the user uses a joystick), a mouse or wearable sensors (i.e. special gloves), an eye movement detector (the user can use a device, such as Google glasses or a special one) camera), which can recognize part of the context that the user is looking at according to the position of the user's eyes and commands (for example, double blinking) to “activate” this solution. It is worth noting that the illustrative version uses only part of the text, which is considered as an expression of user interest.

An expression may contain a term, a named entity, or a phrase that, when supplemented with the heading of the entire page / paragraph, can be used to narrow down the selection. An additional advantage of such an explicit “activation” is that it identifies a user who is in a state of “inquiry,” “curiosity,” or “demand.” Unlike hyperlinks, any part of the text is clickable (i.e., selectable). A click, touch, eye, kinetic or other action performed on a page element (i.e. text, image, media content, text part, part of an audio / video file) leads to a list of relevant information blocks (pages, page links, links to page elements) that correspond to the content, context, specific choice and user profile and target parameters. Thus, the user is shown the most relevant information block.

According to one illustrative embodiment, some elements may be excluded from the preview, or they may be replaced by common images (or icons). Using icons instead of some less relevant page elements reduces network traffic and the load on your mobile device. It also improves the speed of displaying web pages. Almost any activated (clickable, tangible, etc.) text takes into account the nearby context and leads not to one page, but to a list of relevant information blocks / pages corresponding to the content, context, specific choice and user profile.

According to another illustrative embodiment, content-based on-demand dynamic advertising is implemented. Any type of content (e.g. video, audio, text, etc.) tends to arouse or increase the desire of people viewing (or listening to) the content. There is a direct link between content and some desires caused by content. According to an illustrative embodiment, this relationship is used for targeted advertising.

Any part of the content (and the corresponding context) that creates certain desires and associations can be used to target advertisements (relevant information blocks) to the user. However, the main problem is that the elements in the Text → Desire → Satisfaction chain described in terms of word vectors are not geometrically close to each other. This means that the page or product descriptions that satisfy a specific desire may not contain the words of the original Text and not name or implicitly relate to the desire itself. Thus, some analysis of Text-Desire-Satisfaction associations is needed. With the exception of obvious cases, associations between Text and Satisfaction can be made statistically based on user transitions.

It is worth noting that when advertising is relevant in terms of user desires and expectations, it is no longer considered by the user as advertising. Instead, the user perceives targeted advertising as useful information. According to an illustrative embodiment, the user is given the opportunity to explicitly indicate the part of the content (for example, a paragraph within the text or the name of the product) that provokes interest (or desire). A set of user goals / desires (content elements that indicate potential user interest) is implemented as a set of element attributes (information block attributes) automatically generated for this content. The system determines the most appropriate content element attributes for a particular user.

According to an illustrative embodiment, the attributes of the content element describe the user's intentions with respect to the topics disclosed in the content and / or related topics. Content element attributes can be implemented as a ranked list of keywords, bigrams, topic descriptors, etc., extracted from the context and expanded by a synonym and statistically related descriptors. Descriptors are simple text labels that expand the set of well-known words. For example, a process can be described as follows:

the text "SVR - Singular Values Decomposition - Linear Algebra Algorithm", after lexical analysis and stemming are marked with the following initial list of descriptors:

[ks | svr, ks | singular, ks | quantities, ks | decomposition, ks | lin, ks | algebras, ks | algorithm, bg | singular + quantities, bg | quantities + decomposition, bg | lin + algebra, bg | algebra + algorithm].

Next, the list is sent for categorization. After detecting directories (marking), the list is expanded based on category dictionaries with the following labels:

[cat | computer + technology, cat | computers + devices, cat | education].

After receiving a specific part of the sentence specified by the user, for example, “cfr”, the algorithm is able to detect (from statistics) that a search with the keyword “cfr” on the subject of computer technology is not leading to an operation (i.e., it does not lead to a purchase or registration, or to something commercially significant). Thus, the possible requirements associated with these categories (and context) are limited to “Learn more.” Therefore, a new label is added to the context descriptors:

[us | learn + more]. Here, the prefixes cc |, bg |, cat | and us | denote the keyword, bigram, category and intention, respectively.

All content is analyzed. When the user selects a small part of the content by clicking on it, the attributes of the content element reflecting this part are generated and used to target the user to the corresponding advertisement. To do this, the system recognizes parts of the content, parts of speech, the meaning of words, etc. Using OEY technology (Natural Language Processing) or semantic analysis of the content and context.

Given that OYE processing can take considerable time depending on the size of the content, delays can affect the usability of the system. The attributes of the content element of the content itself can be pre-generated and cached, and hashed for later use. Thus, the user is shown highly relevant advertising on demand, where advertising is provided in response to a direct client request (as opposed to a compulsory regime), and this advertising is presented as useful recommendations, and not as direct advertising. This ad is more effective than randomly displayed static ad or flashing relevant information blocks. In other words, the user points to a certain part of the content and explicitly indicates what he wants to see.

This has advantages over regular Google ™ advertising, which requires the user to switch to Google ™ search and enter a search string. This is especially important on computer devices (mobile phones, laptops, desktop computers, game consoles, televisions, Internet tablets, tablet computers, ultra-mobile computers, mobile Internet devices, e-books, smartphones, etc.) because search (using Google ™ or other means) is strongly attached to the input ability, which is very limited in relation to computer devices. On stationary computers, the “Read → Select → Search Choice” case is much simpler than on computer devices. An illustrative example provides users with a quick call: "Read → Specify → Search."

It is worth noting that OEYA technology provides an automated analysis of parts of the text selected by the user, and determination of the content of the text. OEYA also defines the main objects, such as associations, entities, geographical objects, quantity / volume, money / currency, interest, company data, etc. The list of objects can be expanded using element attributes and target data (for example, time, geographical location, personal preferences, user associations, registrations and membership in social networks, gender, language, etc.). According to an illustrative embodiment, the NPL helps the user when the user has difficulty remembering and entering search text, or selecting, copying and pasting the desired text, or highlighting the text and starting the search.

According to an illustrative embodiment, the system uses OEY and “guessing” parts of the text that may be of interest to the user based on the approximate location in the text. Well-known Essence Retrieval methods (IP algorithms) can be used along with a statistical analysis of the degree of unusual words in a part of speech (CR) for a particular paragraph, text and its category. Unusual words (for example, words that provoke a user's interest in a certain part of the content) can be grouped by user interests. Unusual words are used as a "highlight" - items of interest that the system can offer the user.

According to one illustrative embodiment, instead of a search query, content and related attributes of a content element based on an explicit user choice can be used. Content element attributes are a set of formalized properties arising from the meaning of content. Content element attributes (metadata) are presented in the form of a machine-readable language, including keywords, bigrams, n-grams, detected named entities, expected categories, as well as user categories and tags based on frequency distributions. In other words, all possible data is used that can be used to determine the relevance of the advertisement.

According to an illustrative embodiment, two-component targeting (or two-level representation of the request) of an advertisement is used. Advertising is aimed at the user based on a combination of contextual attributes of the content element and the choice or explicit indication of words. This new approach provides advantages over search-based advertising or on relevant information blocks that use either the general theme of the page or an explicit search query.

The specialist will take into account that targeted advertising on demand is advantageous for computer devices having a small screen size. Constantly hanging relevant information blocks (advertising) takes up space on the screen and require changes to the original content. According to an illustrative embodiment, the content (i.e., an object, such as a sentence or part of a sentence, part of an image, paragraph of an article, video fragment, image, three-dimensional model, etc.) specified by a user is used for advertising instead of a single search term torn out out of context.

Thus, according to an illustrative embodiment, irrelevant advertising will be completely eliminated. Advertising is shown on request, instead of relevant information blocks that occupy screen space. Advertising is displayed according to a portion of content explicitly selected by the user. Thus, an advertisement displays only relevant information based on a portion of the content, in contrast to some general advertising based on the entire content.

Knowing the set of appeals of "desired content" an advertisement can be generated and displayed. For example, a hotel - booking, reviews; restaurant - menu, table reservation; product - buy, reviews, etc. According to an illustrative embodiment, a list of key objects and their categories are automatically generated using Natural Language Processing technology.

For example, if an article is devoted to a ski resort and the user clicks on the part describing one of the resort hotels, then there is a high probability that the user wants to learn more about this hotel, book a room there or buy tickets for the ski lift, etc. however, if the user clicks on new skis, then most likely he wants to see some reviews and find out where he can buy or rent them. To implement this scenario, the system uses a predefined associative model that defines a list of possible actions for each type of object. An associative model is formed on the basis of user feedback (i.e., user actions). The associative model can be adjusted based on collected statistics and marketing research.

In the case where the part of the content selected by the user is complex and contains several potential goals and wishes, the process of displaying advertising is performed in two steps. In the first step, the system determines the wishes of the user. In the second step, an advertisement is generated based on the data from the first step and displayed to the user. In the first step, the user is presented with a list (words and pictograms) of possible goals. The user is asked what is most interesting to him at the moment. The user then selects one or more targets. This data is used in the second step, and an advertisement is displayed to the user relevant to the selected user goals.

According to one illustrative embodiment, the entire screen of a computer device (or part thereof) can be converted into one active hyperlink, instead of a large number of hyperlinks (most of which are not used). This avoids constantly displayed ads and relevant information blocks that occupy a small screen of a computer device. In addition, the user can see an unlimited amount of advertising, since the relevant information blocks are shown on demand and do not take up space on the screen for a long time.

According to an illustrative embodiment, any textual content can be used for targeted advertising. The system can display to users several relevant information blocks based on the respective context. The on-screen interface of the illustrative embodiment may use controls (panels, input windows, buttons, check boxes, etc.) for additional operations. The user can override the request using the list of options. For example, the system may provide a button “I'm disappointed” or “Not what I'm looking for,” which means that the customer is dissatisfied with the proposed advertising.

Also, the system can provide a way to choose advertising that is close (or very close) to the needs of the client, but does not fully meet his requirements. This provides the user with the opportunity to rate ads directly on the site using the “like”, “dislike” buttons, etc. Further, this information can be used as feedback from advertisers to target ads and encourage users. It also provides customers with a “gamification” effect for customers, for example, enhances customer interaction with advertising and involves internal motivators such as a sense of independence and control. Active control along with relevant information blocks is depicted in FIG. one.

According to an illustrative embodiment, targeted advertising on demand is only visible temporarily, in contrast to the obsessive constant relevant information blocks. Thus, advertising predominantly does not require “shelf space” and does not destroy the original content. The user can be displayed a large amount of advertising, enhancing the effect of advertising. Ads do not annoy the user (or at least annoy the user less than regular ads), because they are generated based on user preferences. Advertising is aimed at a specific user based on semantic contextual analysis, if necessary, without tracking user actions. Feedback from content users is provided.

Users (customers) can click on the part of the content that somehow stimulates their feelings - this provides additional feedback on the quality of the content. Users can tell the media resource directly - "this is interesting" and "this is not interesting." The history of clicks and analysis of the “clicked” parts of the content makes it possible to recognize patterns that can help make the content more attractive to the user. Advertising that receives more than a certain number of negative user reviews over a period of time T is automatically excluded from the submission process, and the advertiser receives a notification.

In an illustrative embodiment, multiple quality notification thresholds are used with predefined actions for negative and positive reviews. Reviews are stored on the server in the form of links: <advertisement, timestamp, review>. Only the server owner has direct access to all data for further analysis. Feedback data is available to users in aggregated form (for example, amount, average, etc.)

The user initiates the feedback process by clicking / touching the feedback button. Each button is associated with an identifier of a relevant information block (web page identifier), which is transmitted to the feedback server. Each review is registered / saved in the database in the form of a record containing:

- identifier of the relevant information block;

- user signature;

- timestamp; and

- type of feedback (for example, like, dislike, useless, not relevant to the topic, expensive, complaint, etc.).

A user signature is a hash of browser user data that does not contain any personal information. The identifier of the relevant information block is the identifier of the advertising message. A timestamp is the current time accurate to milliseconds.

According to an illustrative embodiment, each user has access only to:

- aggregated records, filtered about the user signature and aggregated by the identifier of the relevant information block;

- aggregated records that are not filtered by the identifier of the relevant information block.

Thus, no personal data contained in reviews of other users is available to the user.

For example, if the object is a new car, the list of associations may include finding a dealer, ordering a test drive, searching for a new car on credit, etc. Thus, an advertisement offering these services is generated and displayed to the user. The system architecture is shown in FIG. 2. The system includes an ad server 240. An ad server 240 processes the content, stores and selects an ad. Special advertising libraries 230 and 250 combine ad server 240 with media resources and applications installed on computer devices.

The ad libraries 230 and 250 display the ad and process user reaction to the ad. For example, a user may click on a relevant information block, ignore it, request a redirect, or express disappointment with an advertisement. Advertising libraries 230 and 250 are designed to work with computer devices 260 and media sites 210 (sites are a set of web pages (with scripts, files, links, etc.) displayed to the user via a web browser).

According to an illustrative embodiment, the content is stored on a remote media server (a third-party server that hosts a honey resource for applications or sites consumed by a user) 220 and is provided to media applications on computer devices 260 or media sites 210. The media server is a third-party server , which hosts media resources for applications or sites consumed by the user. A publisher is a person who owns a media resource.

The integration system with the media server 220 is carried out by the advertising libraries 230 and 250. The content analysis is carried out on the media server 220, so that the computational load on the clients 210 and 260 is minimal.

According to an illustrative embodiment, the advertisement is stored on the advertising server 240. The advertising server 240 is accessible through universal control portals 270 accessible by computer devices 260 and media sites 210. After the user clicks on parts of the content, the OEY module determines objects within the text, classifies objects and defines relations between objects. As a result, user wishes (desires) are associated with sentences and paragraphs of text. According to statistics, up to 80% of words are used as connecting elements that do not reflect user wishes. Thus, objects that actually cause user wishes are rather small.

Next, an associative model with a matrix is created. The associative matrix defines the relationships between objects and user wishes caused by these objects. The associative matrix also defines the boundaries of these relationships. In other words, the relationships between the desire object and the application restrictions for a given pair are determined. For example, gender, season, social events, etc. are important restrictions on the behavior and desires of people.

It is worth noting that associations are based on various factors and mainly on the type of object. For example, a hotel object has one set of associations, a restaurant object has a different set of associations, and a car object has another associative set, and so on. The associative matrix inherits some principles from the concept of the “graph of interests” and expands them with the implementation of some new features.

For example, Amy Joe Kim's classification of the behavior of gamers (compete, explore, interact, express) can be used to recognize words in the content that are associated with each type of behavior (for example, the words “design, create, build” are related to mood “express "), and associate the user's mood with the services and activities suitable for that mood. Maslov’s pyramid can be used to represent the prioritization of desires. In addition, links between advertising categories of services / products can be used, which are based on the statistical distribution (popularity) of classifications in resources (such as Pinterest). The classification of external motivators (status, access, authority, material) and internal motivators (competence, independence, kinship) can also be applied to the classification of content and advertising.

As a result of the association, a list of advertising goals is created. The associative model is formed on the basis of sociological and statistical data. The associative model is constantly updated (trained) in the process of system deployment. Since the content determines and stimulates the wishes (desires) of the user, the content can be analyzed by OEYA only once. Then, a universal model of user wishes can be created by applying OEY to an associative matrix. Advertising can be classified in accordance with its relation to user desires, instead of the classification of products / services.

In addition to keywords (for example, the brand or name of a product or service), each object in the content can be associated with several possible activities that can be performed with the object, for example, “buy”, “reservation”, “order”, “ listen "," watch ", etc. In addition, each object can be associated with some desire, for example," security "," hunger "," thirst "," curiosity "," love ", etc. The content also Keywords of interaction may be contained, for example, “build”, “win”, “like”, “collect” and others. This can help determine the current state of the client to better target ads.

In FIG. 3 shows a flow diagram of a method for accurately selecting ads in accordance with an illustrative embodiment. In step 310, the publisher adds new content to the media resource. In step 320, the metadata of the content and the web page (i.e., the web page parameters displayed to the viewer. The HTML tags defining the page metadata are called meta tags) are downloaded from the media source to the ad server. In step 330, the content is analyzed in order to determine the user's goals (desires) (to generate the attributes of the content element, CD). Then, the user selects a part of the content in step 340, and the part of the content and the corresponding goals (i.e., user goals, attributes of the content element) are identified in step 350, and the process proceeds to step 360.

If in step 360 more than one potential user target is identified, all potential goals are displayed in step 365 so that the user can select one of them. Next, the process proceeds to step 370, where the goal is selected by the user, and the process proceeds to step 375, where the advertisement corresponding to the goal and target parameters is selected. The process then proceeds to step 380, where the free (organic content) is selected. Next, the process proceeds to step 390, where the location of the paid content and / or organic content on the screen of the computer device is determined. The relevant information block (s) and / or organic content is displayed in step 395 and the process ends in step 398.

It is worth noting that the part of the content selected by the user is determined at the initial stage of the process. The system processes user errors caused by the incorrect position of the user's finger on the screen of a computer device by approximating the touch point with several "predicted" points using a random shift [0-X]. In one illustrative embodiment, the minimum target area is a proposal. If a finger touch touches more than one sentence, the system automatically includes the entire paragraph. If a finger touch occurs between paragraphs, both nearby sentences are used for OEY analysis. In the case of computer devices with larger screens, the position within the sentence is used for analysis.

Paragraphs, for example, in HTML format are geometric objects with known boundaries. This allows you to define a paragraph within the text selected by the user of the computer device. The geometric boundaries of sentences or other parts of the text can also be calculated to determine the specific part of the text selected by the user. For example, the text:

"UCWeb named India its second headquarters in April. At that time, co-founder and President Xiaopeng Hye said the company will seek partnerships in the spheres and verticals to build a thriving ecosystem in the Indian market, where IC Browser was first launched in 2011. Ma on board will be useful as a UCWeb input worker in various sectors. "

A user interested in learning more about "UC Browser" may touch the point next to the phrase, but not exactly on it. An application showing this phrase interprets the touch event as associated with the word “where” (before “UC Browser”). The illustration algorithm helps you find the most likely points of interest. For an illustrative sentence, the words "zest": "Xiaopeng" (probably a named entity), "Indian market", "prosperous ecosystem" and "UC Browser". Thus, the user can easily refine his search for "UC Browser" only by selecting a term from a list of assumptions.

In an illustrative embodiment, media resources are used to accurately target ads. An illustrative embodiment uses a resource directory that indicates content and user history for a resource. For example, some directories used for targeted advertising attract a specific audience - for example, each of the “financial markets”, “banking news”, “travel”, “technical news”, “business schools”, “small businesses”, “ the development of "," gardening "," real estate "," cars ", etc. have their own audience. The user story collected on the ad server can be used to expand / refine the user profile in more detail. However, the user story serves only as additional (optional) information.

In FIG. 4 shows a flowchart of a method for accurately selecting ads according to an illustrative embodiment. In step 410, a portion of the content selected by the user is recognized. The metadata (goals) associated with the selected piece of content is determined in step 420. Next, the metadata of the entire content is applied in step 430. This metadata is metadata of the content context, in other words, metadata is additional data that describes the content (keywords, categories , titles and other links).

At step 440, additional restrictions apply (gender, site, device type, location, language, age, gender, etc.). If more than one potential user target is found in step 450, the possible goals are shown to the user to select one in step 460, and the process proceeds to step 470. Otherwise, the associative model (matrix) is applied to the target in step 470. Subsequently, the constraints (geolocation data, gender, time zone, type of user device, website, etc.) are applied to user data in step 480. Then there is the most relevant advertisement in step 490.

For example, when a person clicks on a sentence describing a restaurant and a musical group that plays in that restaurant that night, in the first step, the algorithm must choose between “food” and “music”, and then the client selects the category “food”, and it turns out that the restaurant has a different geographical location than the user (for example, the user is in the USA, and the restaurant is in France), and there is no need to offer this person to reserve a table in this restaurant, but it would make more sense to offer him a local restaurant with the same type kitchen.

For example, the content may be related to automobiles, and the context may include the categories “purchase”, “repair”, “sale” and “sale.” The system takes into account the user history of visited sites, the type of operating system of the computer device, the model of the computer device, financial user status, user preferences.The system also takes into account the metadata of the current page or the metadata of visited pages, user search queries, mail messages and messages and comments in social networks.

According to one illustrative embodiment, relevant advertising can be selected based on keywords and placement on the page currently viewed by the user. Context is a hierarchy of a group of words or topics corresponding to the content (text) located around the selected word, sentence, paragraph or part of the text. In an illustrative embodiment, the context is analyzed and the relevant advertisement is shown depending on the weight of the text and metadata, including additional information for targeting the advertisement. It is worth noting that all text is used as a meta link. Advertising is generated on the advertising server and is displayed without changes in the original page structure. The area selected by the user is highlighted as a hyperlink.

In one illustrative embodiment, an ad that has not been paid for is not shown. Paid advertising is tracked on the ad server that sends the notifications. Paid content is stored as a set of records that contain data about:

- impressions (during the campaign, month, day);

- user visits after clicking on the relevant information block (during the campaign, month, day);

- operations (sales, calls, CMC, etc.);

- the estimated cost of showing ads, moving to the site, operations (during the campaign, month, day);

- the maximum and minimum number of ad impressions, the transition to the site, operations (during the campaign, month, day);

- maximum and minimum prices for ad impressions, transition to the site, operations (during the campaign, month, day);

- cumulative price limit;

- cumulative time limit.

According to one illustrative embodiment, advertiser accounts are incremented and compared with the limit, and the relevant information block is marked as inactive and is no longer shown if the limit is reached. If the account balance increases, the relevant information block is displayed again. In one illustrative embodiment, the content is a meta link. The user can get to the linked page by touching, gesture, eye movement, voice, etc. Clicking is performed as a special gesture: moving left or right on a paragraph to move between paragraphs. Link is built dynamically based on the results of semantic analysis.

The context of a given text expression (which may consist of a word) in this text document is a text document that includes the expression and the surrounding text. The range of surrounding text is determined by the level of context, which can take on a value from a part of a sentence, sentence, paragraph, article, set of domain names, etc. The context of a certain level of context C is considered significant if the set of its values A obtained using the function M: A = M (C) is not empty, where M is the mapping of values. The mapping of values is an approximation of a function defined on the domain of the content structure and having its range in a limited set of user reactions or intentions (curiosity, learn more, wish to have, etc.).

The display of values shows potential user interest that can be switched when meeting certain parts of the text. The context level is defined for a particular text expression of a particular level. Sentence: to distinguish the following levels of expressions, from small to large elements: Word, Group of words, Component of sentences, (sequence of words between commas, etc.), Sentence, Group of sentences, Paragraph. The following contextual levels for a Word: Group of words, Component of sentences, Sentence, Group of sentences, paragraph, Group of paragraphs, Article, Group of articles, Topic in section, Subject area. For a group of words: Component of a sentence, Proposal, Group of sentences, Paragraph, etc.

The keyword extraction algorithm can be used in this implementation of the invention as follows:

Input: text element E, quantile q

Output: S - list of keywords / bigrams in the form of CD (Contextual Descriptor)

To describe keywords, a special form is used as a CD. A simple KD is a pair: (KD_type, KD_value), where KD_type takes the value from {keyword, bigrams, ...} and KD_value is a normalized (stamped, lemmatized, "lower case") word or bigram.

1. Extract all text expressions that satisfy the following patterns:

<article> [adjective] {noun / number} [verb]

[adjective] {noun / number} [verb]

...

Designations:

<.> - exactly one element

{.} - one or more elements

[.] - zero or more elements

Add all of these expressions to multi-set R.

2. Remove stop words (if they are part of the template) from the expressions. Stop words are words of a language that are considered low informative, garbage or noise, such as articles, sentences, etc.

3. Count the frequency of words and bigrams included in expressions.

4. Build a distribution. Sort words and bigrams by this distribution. Select words and bigrams with the maximum frequency using the given quantile q.

Add these words and bigrams to the returned set S. Only meaningful content is analyzed. The utility algorithm can be used as follows:

Input: P webpage, q1, q2 quantiles

Output: A set of significant elements M

1. Consider this web page as a set of HTML headers + HTML elements, P = {h1, ..., hN, e1, ..., eM} and meta-information (meta-description + meta-content) M = {m1, ..., mK}.

2. Remove elements with a low value for the length of the text and elements with a large value of the relative concentration of links from R.

3. Stem words in P and M, delete stop words.

4. A multinet of significant elements S = {}

5. To make a text vectorizer based on meta-information m1, ..., mK ', V [m1, ..., mK'] (x) is a vector function. It shows how many words from {m1, ..., mK '} are contained in the text element x.

6. Add to S such elements e from P that satisfy the condition sum (V [m1, ..., mK '] (e)) / lengths (V [m1, ..., mK'])> = q1, where sum (V [m1, ..., mK '] (e)) is the sum of the elements of the vector, length (V [m1, ..., mK']) is the length of the vector (which is equivalent to the "number of function arguments").

7. Find the headers h1, ..., hL such that:

h: Sum (V [h] (x)) -> max x from P, where max is a vector of maximum length. Let's call them base headers. Make a vectorizer V [h1, ..., hL] (x)

8. Add to S such elements e from P: sum (V [h1, ..., hL] (e)) / len (V [h1, ... hL])> = q2.

9. Add to S elements that satisfy certain geometric conditions. For example, if the basic set of headings is {hi}, add in S elements that are located below this heading in the text heading on the web page.

10. A set of significant elements M = Unique elements from S.

Semantic analysis is performed when content is published in real time, or when pages are polled. The user does not require any additional software (everything works in the browser, like regular links). All calculations are performed on the server side. The server side consists of the following services:

Semantic Content Server - is responsible for the preliminary processing of content components and for storing them in a machine-readable format;

Advertising Features Serever - is responsible for the preliminary processing of the resulting texts and for storing them in a machine-readable format;

Target Server - is responsible for the analysis of reader requests for content and the search for matching results, the most relevant points of view of the user's intention.

The types of results can be used in an illustrative version: the promoted result, taken from the advertiser, the advanced, always up-to-date, organic content is shown when there is no relevant selling content.

An example of organic results: reviews about an object, opinions from forums (mechanisms of reliable advertisers), geographical points, price and sales information, similar objects. Depending on the categories found, organic results can be processed from GOOGLE, WIKIPEDIA, ETSY, etc. Organic or Promoted results can be stratified according to user intentions, they are designed to be used in relation to part of the content: Learn more about, I wonder what it is, What about this new, what people say about it, where can I get it, how much does it cost, I want to buy it. It is worth noting that organic and promoted results are shown simultaneously.

In FIG. 5, 6, 7, 8 show advertising and organic results on the screen of a computer device. The size of each object and the number of objects on the screen may vary depending on the screen type of the computer device and the configuration of the mobile application for displaying objects. If there are several objects, the objects can be scrolled up and down in the form of a loop shown in FIG. 5-8. It is worth noting that the user can choose an action for each object (for example, open, vote for / against, send to a friend, etc.)

In FIG. 9 shows the processing of content on a computer device using an action list for predefined user requests. The user can move (up, down, left and right) parts of the content beyond the screen of a computer device using a special gesture. At the remote content location, action buttons may be displayed. Words (goals) from the selected context can also be shown - for example, gas burners, vehicles, etc. The user can select one or more words (goals) and / or actions (from buttons or lists) to specify information, for feedback or to open relevant data information blocks. Thus, the user will be shown the most relevant advertising. It is worth noting that one or more words / goals may include a number of links to various advertising or organic (natural) objects that can be displayed on the screen of a computer device (see FIG. 9). It is worth noting that the selected content can be expanded through animation in such a way that advertising or organic content and a list of words are displayed due to the content. The screen of the mobile device may have a special element that allows you to move parts of the contents of the web page to display goals.

In one illustrative embodiment, the system turns all content into an active link. The system uses an associative model that describes the possible wishes and actions of the user, which can be taken based on the content of the object. In other words, the system provides the user with actions if the object is of interest to the user. For example, the association for the object "hotel" may be "room reservation", for the "restaurant" - "reserve a table", etc.

According to one illustrative embodiment, the relevant data is determined for the user based on the available data (geolocation data, personal data, time zone, type of computer device, etc.). Thus, after creating a model for one user, it can be used for other similar users. This saves a large amount of advertising resources. The user can determine his advertising choice at any time, which reduces the negative effect of regular advertising.

In FIG. 10 is a flowchart of a method for accurately selecting ads according to an illustrative embodiment. The advertiser can create a new relevant information block or target group, or campaign in step 710 (i.e., create a new advertisement) and manually determine the keywords, and he can also manually determine the target parameters. Target parameters is a list of advertising attributes that describes the audience that this advertisement should be displayed, including user’s age, gender, location, user browsing history, media site statistics, website content, page metadata, and any other social information about a user accessible globally, the browsing history or search queries. If the keywords are defined in step 720, then the keywords are converted to CD in step 740, and the process proceeds to step 750. It should be noted that steps 720 and 725 are independent and can be performed in parallel or sequentially. If the target parameters are defined in step 725, then the CDs are created based on the target parameters in step 727, and the process proceeds to step 750. The advertiser can also indicate the Uniform Resource Indices of his product or service. Otherwise, the keywords are created automatically by the PageVKD algorithm in step 730. An advertising category can be automatically determined for each relevant information block in step 750 using the following algorithm:

the CD of the relevant information block is extracted using keywords;

determine the main category or categories for the relevant information block based on the voting algorithm;

the category of the target group is defined as a set of categories of relevant information blocks of the target group, ordered in descending order of frequency;

CD and category are extracted from the context. A heuristic algorithm is used to search for keywords specific to the page and for the entire category. An analysis of part of the speech of the text is performed. CD sets are created and categories are determined by voting.

Relevant contextual advertising is defined as an advertisement that has a category corresponding to a context category. Advertising sorting can be implemented using the voting algorithm: the context CD votes for the relevant information block CD. The number of CD votes is calculated for each relevant information block, and the relevant information blocks are sorted in descending order of votes.

A CD is an entity that has a contextual object (for example, keywords, bigrams) and weights assigned to the object. The more weight, the more important the object. The voting algorithm is implemented as follows. Objects candidates and objects of the participants in the voting have their own properties and weights. Initially, the votes for each candidate are zero. Further, for each candidate and for each property, the weight of the property is added to the candidates with this property. The winning candidates are the objects that collected the largest number of votes.

Next, the process proceeds to step 755. If the exclusion categories are added by the advertiser, the set of main advertising categories is updated in step 760:

adult, cars, business, finance money, entertainment news and gossip, food and drinks, games, health and fitness, movies TV, music, news and weather, shopping, social networks, society - Latin, society - male lifestyle, society - female lifestyle, sports, style and fashion, technology, travel, utilities, etc.

Then, for each category, in step 765, a set of entities of contextual descriptors (CD) is created containing the text features (keywords, bigrams). The CD describes these pages as follows:

- creates a set of search phrases for the category;

- creates a set of search phrases for the category;

- Unified Resource Indexes of the page, which are considered basic for this category, are obtained using the Search Engine Application Programming Interface. An arbitrary number of Common Resource Locators can be obtained.

- Each page is processed by the PageVKD algorithm, which converts the contents of the page into a set of CD. CDs are used as coordinates in context space to measure similarities or correspondences between them. The cosine of the matrix is used to calculate the CD.

A set of CDs ordered in descending order of frequency determines the category. Most frequent CDs describe the category better than less frequent ones. The algorithm for obtaining CDs can be manually modified to filter the results of the PageVKD algorithm.

Next, the process goes to step 770. The process stores the set of CDs on the server. In step 775, the user selects the content (or part of the content). The process searches for keywords in the selected content in step 780. Next, the process proceeds to step 782, where categories from the keywords are determined, and in step 784, the system determines additional objects based on the target parameters. The process selects the advertisement by matching the categories from the server in step 785. The selected advertisement is displayed to the user in step 790.

According to one illustrative embodiment, the PageVKD algorithm is used. Page metadata is selected and stop words are deleted. Page titles are selected and stop words are removed from them. Paragraphs with content that is close to metadata / headings and stop words are deleted. The most commonly used keywords / bigrams are selected from the combined text. Keywords / bigrams are converted to CD. Weights are assigned based on the location of the keywords / bigrams - the paragraph is assigned more weight, and the heading is assigned less weight.

In FIG. 11 shows screens of a mobile device. Relevant data is displayed according to user goals / desires selected based on the context of the content. As soon as the screen element of the relevant data block is activated, the data associated with the goals / wishes are displayed. For example, part of the text may be flipped over, while showing goals, and the user can select one of the goals. Once a target is selected, relevant data blocks are displayed.

In FIG. 12 shows the exact selection of advertising and a graphical representation of the targeted advertising to the user. The user can select a portion of the content 910. An object 940 is displayed above the content either in the browser menu or in the pull-down menu. The user can open an additional window 960 (for example, a new window in the browser or a window in the current browser). Advertising or organic content 985 and 990 can be displayed in a window 960 in a designated area 970 along with objects 980 (search, text, YouTube, etc.) that reflect the user's intentions. Displayed advertising or organic content 985 and 990 is selected based on: analysis of content, content category, language, selected content area and user preferences. It’s worth noting that user intent is clearly displayed along with free organic content.

In FIG. 13 is a flowchart of an illustrative PageVKD algorithm using a search of relevant pages. In step 1005, metadata (M) is collected using Unified Resource Indexes (keywords, content). It is worth noting that metadata may not be available. Metadata is provided by the publisher (e.g., <meta> car sales </meta> tags). In step 1010, metadata is cleared of stop words. In step 1015, data from the headers (for example, <head>) and from the selected words are collected and cleared of stop words in step 1020. In step 1025, data from paragraphs is collected and cleared of tags in step 1030.

In step 1035, the cleared metadata from step 1010 and header data, plus the selected words from step 1020, are vectorized: V = vectorize (M ′ + H ′). At step 1040, the process checks if V> 0. If V> 0, then the process goes to step 1041. If V <= 0 (for example, the set of metadata and headings is empty), the process goes to step 1077, where the set of all text elements (paragraphs, parts of sentences, parts of text, words, parts words, etc.) becomes equal to R. Next, the process proceeds to step 1050.

In step 1041, the set of paragraphs P '(or part of the content, for example, two or more paragraphs, recognized information in the selected part of the image, video or audio) is split into a piece of text Ti, and the process proceeds to step 1045.

In step 1045, the process checks if there is still text (content: part of the image, part of the text, etc.) Ti. If the text exists, the next text Ti is obtained in step 1046. In step 1047, the Pi element is vectorized: VTi = V (Ti), where VTi is the product of vectorization. In step 1048, the text element Ti is applied to the base vector V, which produces the coordinate vector, and the process checks if the sum of the coordinates of the vector is greater by the value of the vector than α, where α is the level above which the paragraph is taken. α is determined empirically and, as a rule, is in the range 0.14-0.16. If the condition is not satisfied in step 1048, the paragraph is not included in the array of possible paragraphs for display, and the process returns to step 1045. Otherwise, the current Ti is added to R, and the number i of the text element is added to I in step 1049.

If in step 1057 the number of the current paragraph differs from the previously recorded one by more than one, the process takes one of the unverified missing paragraphs in step 1058. Otherwise, the process returns to step 1045.

In step 1059, the process checks if the sum of the coordinate vector divided by the vector is greater than α / 2, where α is the level above which the paragraph is taken. If the condition is not met, the process returns to step 1045. Otherwise, if the condition is satisfied, the process proceeds to step 1060, where the missing paragraph is added to R and its number is added to I. Then, in step 1061, the process checks to see if all the missing paragraphs. If not all the missing paragraphs were checked, the process returns to step 1058. Otherwise, the process proceeds to step 1045.

If, in step 1045, there is no Ti left, the process proceeds to step 1050, where M 'and H' are added to the resulting paragraphs in step 1050. Otherwise, the process proceeds to step 1071.

Then, in step 1052, the element is vectorized: VR = V (R), where VR is the product of vectorization. VR is sorted by frequency in step 1053. In step 1054, features are selected (or selected) by quantile. The selected data is displayed to the user in step 1055.

In FIG. 14 shows a block diagram of an illustrative PageVKD algorithm. In step 1005, metadata (M) is collected using Unified Resource Indexes (keywords, content). It is worth noting that metadata may not be available. Metadata is provided by the publisher (e.g., <meta> car sales </meta> tags). In step 1010, metadata is cleared of stop words. In step 1015, data from the headers (for example, <head>) and from the selected words are collected and cleared of stop words in step 1020. In step 1025, data from paragraphs is collected and cleared of tags in step 1030.

In step 1035, the cleared metadata from step 1010 and header data, plus the selected words, are vectorized in step 1020: V = vectorize (M ′ + H ′). At step 1040, the process checks if V> 0. If V> 0, then the process goes to step 1041. If V <= 0 (for example, the set of metadata and headings is empty), the process goes to step 1077, where the set of all text elements (paragraphs, parts of sentences, parts of text, words, parts words, etc.) becomes equal to R. Next, the process proceeds to step 1050.

In step 1041, the set of paragraphs P '(or part of the content, for example, two or more paragraphs, recognized information in the selected part of the image, video or audio) is split into a piece of text Ti, and the process proceeds to step 1045.

In step 1045, the process checks if there is still text (content: part of the image, part of the text, etc.) Ti. If the text exists, the next text Ti is obtained in step 1046. In step 1047, the Pi element is vectorized: VTi = V (Ti), where VTi is the product of vectorization. In step 1048, the text element Ti is applied to the base vector V, which produces the coordinate vector, and the process checks if the sum of the coordinates of the vector is greater by the value of the vector than α, where α is the level above which the paragraph is taken. α is determined empirically and, as a rule, is in the range 0.14-0.16. If the condition is not satisfied in step 1048, the paragraph is not included in the array of possible paragraphs for display, and the process returns to step 1045. Otherwise, the current Ti is added to R, and the number i of the text element is added to I in step 1049, and the process returns to step 1045.

If Ti is not left in step 1045, the process takes the number of the first paragraph P1 in step 1065. In step 1067, the process takes the number of the next paragraph stored in I. It is worth noting that the paragraph verification process can be started from the last paragraph or from any recorded paragraph, if paragraphs are chosen randomly. In step 1069, the process checks whether the number of the current paragraph differs from the previously recorded paragraph by more than 1.

If the condition is met, the process takes one of the unverified missing paragraphs in step 1070. Otherwise, the process proceeds to step 1071, where the next paragraph number from step 1067 becomes the current number, and the process returns to step 1067. In step 1072, the process checks whether the sum of the missing coordinate vectors divided by the magnitude of the vector is greater than α / 2. If this condition is not met, the process returns to step 1071.

Otherwise, the process proceeds to step 1073, where the missing paragraph is added to R and its number is added to I. Then, in step 1074, the process checks to see if all the missing paragraphs between the two paragraphs are checked. If not all the missing paragraphs were checked, the process returns to step 1070. Otherwise, the process checks in step 1075 whether this is the last paragraph number. If the paragraph number is the last, M 'and H' are added to the resulting paragraphs in step 1050. Otherwise, the process proceeds to step 1071, where the number of the next paragraph becomes current. Next, the process proceeds to step 1067.

Next, in step 1052, the element is vectorized: VR = V (R), where VR is the product of vectorization. VR is sorted by frequency in step 1053. In step 1054, features are selected (or selected) by quantile. The selected data is displayed to the user in step 1055.

The above algorithm for the analysis of text documents of web pages PageVKD is a representative implementation of the algorithm PageVKD:

VKD conversion parameters are defined as VKD = {ImageVKD, MusicVKD, VideoVKD, PartImagesVKD ...} of some non-textual content descriptors. In the case of music, speech is selected and recorded in the CD. For images, the algorithm recognizes images, describes them and saves them in CD. For video content, the algorithm breaks the content into audio stream and images and processes them separately. The algorithm determines the extraction of the basic elements of E - the algorithm can receive metadata and a header. The hood takes into account the page geometry and the current view and determines the q relevance quantile.

The algorithm performs the following steps:

1. Determine the textual content of T and non-textual content of P;

2. For P apply the transformation:

Make the set of the transformed element P 'empty;

For each p of P:

Determine type y = Toure (p)

Apply the VCD conversion (y, p) for this type and get the descriptors, and add the descriptors as a separate set to P ';

3. For T, obtain a set of transformed elements of T and save t from T to CD;

4. Get T 'U P'. determine the basic elements of B = E (V U P '). Other elements are A = (T'UP ') \ B;

5. Collect all CD elements from B into one set of V unique elements. Define the order and generate a vector function of the CV vector argument, which calculates the number of intersections of an element from V with an arbitrary set M. In addition, a weight vector can be created by applying B to V. Next, CV multiplies the number of intersections by weight.

6. the set of relevant R is empty;

For each a from A:

If the sum (CV (a)) / lengths (CV)> = q, add a to R;

Check the missing elements, if found, reduce the quantile by half and add a to R;

7. Select CD for RUB elements by quantile;

8. Give the resulting CD.

Claims (64)

1. A method for displaying relevant information based on content on a computer device, comprising: connecting a computer device to a media server; receiving content from a media server on a computer device; displaying the content to the user of the computer device, wherein any part of the content is selectable by the user and is used to show potential goals to the user for their selection by the user; uploading content to an ad server; establishing a connection between a computer device and an advertising server; generating relevant data blocks based on categories, content part and target parameters, where the target parameters include user information containing the user's social profile data in social networks; analysis of the content by the advertising server; generating attributes associated with content elements based on content analysis; recognition of a part of the text of the content selected by the user of the computer device; Removing stop words from content identification of attributes associated with content elements associated with content and a user-selected portion of the content using Natural Language Processing technology, where at least several attributes are pre-calculated; determination of potential goals of the user of the computer device using attributes associated with the user-selected piece of content; display to the user of the computer device of the potential goals of the user for selection by the user; determining the goal selected by the user of the computer device; applying an associative matrix with a predetermined associative model to the goal selected by the user of the computer device; selecting at least one relevant data unit corresponding to the purpose of the user of the computer device; downloading the identifier of the relevant data block to the computer device from the advertising server; and generating a relevant data block using a predefined template and displaying the relevant data block on a computer device screen. 2. The method according to claim 1, further comprising displaying relevant organic content to the user of the computer device if no relevant advertisement is found. 3. The method of claim 1, further comprising determining potential goals of the user of the computer device based on attributes associated with content elements associated with a user selected portion of the content and attributes associated with the content. 4. The method of claim 1, further comprising converting the keywords and context of a portion of the content into contextual descriptors (CD); generating a set of keywords from the content selected by the user of the computer device by the algorithm for converting Pages to Contextual Descriptors (PageVKD), which converts the contents of the page into a set of CDs, and the CDs are used as coordinates in the context space to measure similarity or correspondence between them; and determining the category of the relevant data block displayed based on the keywords defined by the advertiser, target parameters, categories, and displaying the advertisement if the content is associated with certain keywords. 5. The method according to claim 1, further comprising determining the categories of data blocks by voting. 6. The method according to claim 1, further comprising generating a set of keywords from the context of a portion of the content selected by the user of the computer device, and further comprising analyzing a portion of the speech of the content to identify keywords. 7. The method of claim 1, further comprising converting the element or set of page elements to the active link (s). 8. The method of claim 1, further comprising defining attributes associated with a portion of the content, keywords, metadata of web pages, tags, metadata associated with the title of the web page associated with the portion of the content selected by the user of the computer device; and selecting a relevant data block based on attributes associated with a user selected content part, content attributes, and context attributes of the content part. 9. The method of claim 1, further comprising applying a heuristic algorithm to find keywords in the content specific to the content page and for a category from all categories. 10. The method of claim 1, further comprising applying constraints to relevant data blocks based on target parameters, where the constraints are any of the following: geographic location of the user of the computer device; user language of the computer device; age of the user of the computer device; history of viewing documents; history of browsing the Internet by a user of a computer device; statistics of media sites; part of the content of the website; Page metadata social profile data of a computer device user; tags associated with the current user of the computer device provided by third-party services (DMP, DSP); type of computer device; and type of operating system of the computer device. 11. The method of claim 1, further comprising receiving user feedback of the computer device to the displayed relevant data unit and sending user feedback of the computer device to the advertising server. 12. The method according to claim 1, further comprising determining the location of the data block on the computer device, determining the number of displayed relevant data blocks, and organizing advertising based on the relevance of the data blocks. 13. The method according to p. 1, characterized in that the removal of stop words from the content is carried out before converting the page content into a set of contextual descriptors, and when converting the CD, the cosine of the associative matrix is used. 14. The method according to claim 4, further comprising using a context to determine the received content for display to the user, the data block areas being created being active for performing user actions of a computer device, where the user actions of the computer device are any of the following: display of detailed data; image display; display of three-dimensional models; playing audio and video; and display of additional text, moreover, the areas of the data blocks are made active for performing user actions, where the areas of the data blocks are cyclically scrolled. 15. A system for displaying content-based advertising on a computer device, including: an ad server associated with a computer device receiving a page from a computer device and configured to process the contents of the page, store and select an advertisement; ad libraries connected to an ad server, ad libraries configured to integrate an ad server with a computer device; and a media server storing pages and connected to a computer device and configured to store content and provide content to a computer device, wherein any part of the provided content is user selectable and used to show the user potential goals for user selection; where the media server is configured for content analysis; advertising libraries are configured for: determining metadata associated with the content selected by the user of the computer device by generating a set of keywords from the content using the Page To Contextual Descriptors (PageVKD) conversion algorithm that converts the page content to a set of contextual descriptors (CD), and the CDs are used as coordinates in a context space for measuring similarities or correspondences between them; using the Natural Language Processing technology to determine the potential goals of the user of a computer device based on metadata associated with a user-selected part of the content, categories, part of the content and target parameters, where the target parameters include user information containing user social profile data in social networks, and recognition of part of the text of the content, removal of stop words from the content; showing potential goals of the user of the computer device to the user of the computer device; providing organic content to a computer device; determining the content selected by the user of the computer device and the context of a portion of the content for analyzing the metadata of the content; determining a goal selected by the user of the computer device and applying an associative matrix with a predetermined associative model to the goal selected by the user of the computer device; and generating relevant ads using a predefined template and selecting at least one advertisement relevant to the selected goal by the user of the computer device, and downloading the advertisement to the computer device from the advertising server, moreover, the system is configured to receive user feedback on the computer device on the relevant data block shown and to send user feedback to the advertising server, and generating relevant data blocks based on categories, content part and target parameters, where the target parameters include user information containing data social user profile on social networks.

Images