Monitoring online activity to rank content in real time

Abstract

According to one embodiment of the present invention, a content item containing content is received. A value for the content item is determined based on values for one or more content items associated with the content item. Online activities related to the content item are monitored, and the value of the content item is updated in real-time according to the user activities. The value of the content item may be displayed as the value varies in real time. Embodiments of the invention may include one or more methods, computer program products, and systems for monitoring user activity and updating values of content items in real time. Embodiments of the invention may also include identifying value curves for one or more content items associated with the new content item, and combining the identified value curves to produce a value curve for the new content item.

Description

[0001] Cross-reference to related applications

[0002] This application claims priority to U.S. Provisional Patent Application No. 63 / 454,106, filed March 23, 2023, entitled “Monitoring Online Activity to Rank Content in Real Time,” the disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0003] Embodiments of this invention relate to content distribution and search systems, and more specifically, to using machine learning to monitor online activity in real time and adjust content rankings. Background Technology

[0004] The internet provides access to a wide variety of content, including news articles, blogs, tweets, images, chat, and more. This content is provided by online publishers such as magazines, newspapers, journals, databases, and other information services. Traditional search engines typically analyze the characteristics of potential search results without actually evaluating the content itself in order to provide and rank relevant results. For example, these search engines might calculate the number of links to a website, the number of clicks or visits to that website, and the characteristics / popularity of other websites linking to it. This can result in low-quality but popular content being found and prioritized over high-quality content. This prioritization can also remain the same even if the quality of the website changes. Furthermore, these search engines may take a significant amount of time to crawl the internet, or there may be long intervals between crawls, so the prioritization may become outdated and fail to reflect current content demand. Summary of the Invention

[0005] According to one embodiment of the invention, a content item containing content is received. The value of the content item is determined based on the values ​​of one or more content items associated with a content comparison result, and by examining the inherent characteristics of multiple aspects of the content, referred to as a content genome. Online activity related to the content item is monitored, and the value of the content item is updated in real time based on user activity. The value of the content item is displayed as the value changes in real time. Embodiments of the invention may include one or more methods, computer program products, and systems for monitoring user activity and updating the value of content items in real time in a substantially similar manner as described above. Attached Figure Description

[0006] Typically, the same reference numerals are used in various figures to denote the same parts.

[0007] Figure 1 This is a schematic diagram of an exemplary computing environment according to an embodiment of the present invention.

[0008] Figure 2This is a block diagram of an exemplary computing device according to an embodiment of the present invention.

[0009] Figure 3A According to an embodiment of the present invention Figure 1 A block diagram of the content modules.

[0010] Figure 3B This is a block diagram of the content genome according to an embodiment of the present invention.

[0011] Figure 4 This is a flowchart of a method for determining the value of a content item in real time based on online activity according to an embodiment of the present invention.

[0012] Figure 5A This is a flowchart of a method for updating the value curve of a content item according to an embodiment of the present invention.

[0013] Figure 5B This is a flowchart of an exemplary embodiment of the present invention that uses data from a web crawler to generate a content genome and value curves.

[0014] Figure 5C It is a flowchart of a procedure for integrating feedback loops from users into the system to retrain machine learning models.

[0015] Figure 5D It is a flowchart of a program that dynamically changes weight values ​​based on surges.

[0016] Figure 6A This is a graph illustrating the generation of a new content item's value curve based on the value curve of an associated content item, according to an embodiment of the present invention.

[0017] Figure 6B This is an update according to an embodiment of the present invention. Figure 6A The graph of the value curve of the new content item.

[0018] Figure 7A This is a schematic diagram of an exemplary graphical user interface that provides real-time updated content items and associated values ​​according to an embodiment of the present invention.

[0019] Figure 7B This is an embodiment of the invention that provides historical information of content items. Figure 7A A schematic diagram of an exemplary graphical user interface.

[0020] Figure 7C This invention provides historical information on different historical periods of content items according to embodiments of the present invention. Figure 7A A schematic diagram of an example graphical user interface. Detailed Implementation

[0021] This invention analyzes content items and network operations (or accesses) to those content items in order to discover, prioritize, and rank them to obtain results. Content items can be any type of digital or electronic item or object (e.g., document, webpage, file, data object, etc.) containing any type or combination of data (e.g., text, multimedia, video, audio, images, streaming data, etc.). For example, content items may include news or other articles, websites or webpages, papers, documents, program code or applications, audio recordings, videos, images, live or recorded podcasts, streaming media, streaming of live events, blogs, messages, chats, conversations or other topic strings, any combination thereof, etc. Content items may be associated with indicators or links that allow access to them. As content items and access to them change, the value or ranking of the content items similarly changes. The value or ranking of content items is determined in real time, so that these values ​​are dynamically updated in real time when content items or results are displayed to the user (e.g., similar to a continuously updated and changing stock quote auto-collector).

[0022] Incorporating real-time updates into the system is crucial because it considers not only online user activity but also information dissemination and price updates for similar content items. This is especially important where real-time online activity data may not be readily available. By leveraging information dissemination and price updates, the system ensures a comprehensive understanding of market dynamics and content interaction. This approach enables the system to dynamically adjust its pricing strategy in response to evolving trends and changing demands, ensuring timely and effective decision-making. By integrating online activity data and external updates, the system maintains responsive and accurate pricing, thereby enhancing its effectiveness in meeting user needs and maximizing revenue opportunities.

[0023] Embodiments of this invention are designed to be fair and transparent to both the users who create and receive content items. Embodiments of this invention use machine learning models and natural language processing (NLP) techniques to analyze content items and user behavior in order to determine the value of the content item.

[0024] For example, when a content creator uploads a new content item, embodiments of the present invention perform initial value prediction based on historical data and various factors such as brand popularity, audience demographics, and previous transaction data of similar content items. Embodiments of the present invention utilize real-time user behavior to adjust the value of new content items as interaction with them increases, based on different markets and demands.

[0025] In some cases, the value of a content item may change rapidly due to variations in demand or audience demographics. For example, a content item important to a small number of users may have a very high value in the first 1-2 minutes, then decrease significantly to accommodate a wider range of users. Embodiments of the present invention take these different scenarios into account and adjust the value accordingly to ensure that the value remains fair and transparent to both the user who creates the content item and the user who receives it.

[0026] The embodiments of this invention are fair because the values ​​are based on actual data, not arbitrary value determinations. These embodiments consider a wide range of historical and real-time factors to accurately predict demand for content items and optimize values ​​for each distinct audience population in a fair and transparent manner. This helps ensure a fair exchange of content items for both the users who create them and the users who acquire them.

[0027] Furthermore, embodiments of the invention are advantageous because the value is mutually beneficial for both the user creating the content item and the user receiving it. Content creators can offer their content items at a value that accurately reflects the item's value, without guessing or selling at a low price, while users can obtain the content items they want at a competitive value that reflects real-time demand. This means that both the user creating and receiving the content item receive a fair transaction and are able to participate in a transparent and efficient dynamic online marketplace.

[0028] When the value of a content item is low enough, users can choose to acquire it through advertising (such as pre-roll video ads or coupons). This allows users to access the content items they want even if they cannot afford the full value, while also providing content creators with an additional source of revenue through advertising. Overall, this feature creates a more inclusive and accessible online marketplace for both users who create and users who acquire content items.

[0029] Figure 1 An example environment for use with embodiments of the present invention is illustrated. Specifically, environment 100 includes one or more server systems 110 and one or more client systems or end-user systems 114. Server systems 110 and client systems 114 may be geographically isolated from each other and communicate via network 112. The network may be implemented using any number of suitable communication media (e.g., wide area network (WAN), local area network (LAN), Internet, intranet, etc.). Alternatively, server systems 110 and client systems 114 may be deployed locally to each other and communicate via any suitable local communication media (e.g., local area network (LAN), wired connection, wireless link, intranet, etc.).

[0030] Client system 114 enables users to interact with server system 110 to provide (or upload) and / or obtain content items. Content items can be any type of digital or electronic item or object (e.g., document, webpage, file, data object, etc.) containing any type or combination of data (e.g., text, multimedia, video, audio, images, streaming data, etc.). As described below, the server system includes a content module 116 for managing content items. Content module 116 enables users to upload content items and facilitates user search and transactions to identify and obtain desired content items. The content module determines the value of content items in real time based on monitoring online user activity associated with the content items. These values ​​can be updated in real time on the user's display as user activity associated with the content items changes (e.g., similar to continuously updated and changing stock quotes).

[0031] Database system 118 can store various information for analysis (e.g., activity measurements, value curves, content items, values, etc.). The database system can be implemented using any conventional or other database or storage unit, can be deployed locally or remotely with server system 110 and client system 114, and can communicate via any suitable communication medium (e.g., local area network (LAN), wide area network (WAN), Internet, wired connection, wireless link, intranet, etc.).

[0032] The client system may include an interface module or a browser module 120, which displays a graphical user interface (e.g., a GUI) or other interfaces (e.g., a command-line prompt, a menu screen, etc.). The graphical user interface solicits information from the user related to providing and / or obtaining content items, and may provide content search results with continuously updated values.

[0033] Server system 110 and client system 114 can be implemented by any conventional or other computer system, which is preferably equipped with a display or monitor, a dock, optional input devices (e.g., keyboard, mouse, or other input devices), and any commercially available and custom software (e.g., server / communication software, content module 116, interface module / browser module 120, etc.). The dock includes at least one hardware processor 115 (e.g., microprocessor, controller, central processing unit (CPU), etc.), one or more memories 135, and / or internal or external network interfaces or communication devices 125 (e.g., modem, network card, etc.).

[0034] Various modules in embodiments of the present invention (e.g., content module 116, interface module 120, etc.) may include one or more modules or units to perform the various functions of embodiments of the present invention described below. Various modules (e.g., content module 116, interface module 120, etc.) may be implemented by any combination of any number of software and / or hardware modules or units, and may be located in the memory 135 of the server system and / or client system for execution by the processor 115.

[0035] Figure 2 An example of a computing device 200 in environment 100 is shown (e.g., implementing server system 110 and / or client system 114). This exemplary computing device can perform the functions described herein. The computing device 200 can be implemented by any personal or other type of computer or processing system (e.g., desktop computer, laptop computer, handheld device, tablet computer, smartphone or other mobile device, etc.) and can be used in any computing environment (e.g., cloud computing, client-server, network computing, mainframe, standalone system, etc.).

[0036] The computing device 200 may include one or more processors 115 (e.g., microprocessors, controllers, central processing units (CPUs), etc.), a network interface 125, a memory 135, a bus 210, and an input / output interface 220. The bus 210 connects these components for communication and can be any type of bus architecture, including a memory bus or memory controller, a peripheral bus, and a processor or local bus using any of a variety of conventional or other bus architectures. The memory 135 is connected to the bus 210 and typically includes computer-readable media, including volatile media (e.g., random access memory (RAM), cache memory, etc.), non-volatile media, removable media, and / or non-removable media. For example, the memory 135 may include a storage device 250 containing non-removable, non-volatile magnetic or other media (e.g., hard disk drives, etc.). The computing device may also include disk drives and / or optical disk drives (not shown) (e.g., CD-ROMs, DVD-ROMs, or other optical media, etc.) connected to the bus 210 via one or more data interfaces.

[0037] Furthermore, the memory 135 includes a set of program modules 215 (e.g., corresponding to content module 116, interface module 120, etc.) configured to perform the functions of the embodiments of the invention described herein. The memory may also include an operating system, at least one application and / or other modules, and corresponding data. These can provide an implementation of a network environment.

[0038] Input / output interface 220 is connected to bus 210 and communicates with one or more peripheral devices or external devices 230 (e.g., keyboard, mouse or other pointing devices, monitor, etc.), at least one device that enables a user to interact with computing device 200, and / or any device that enables computing device 200 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Computing device 200 can communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), public network (e.g., Internet)) through network interface 125 connected to bus 210.

[0039] Regarding certain entities (e.g., client system 114, etc.), computing device 200 may further include: or be connected to a touchscreen or other display 225, a camera or image capture device 235, a microphone or other sound sensing device 240, a speaker 245 for transmitting sound, and / or a keypad or keyboard 255 for inputting information (e.g., alphanumeric information, etc.). These devices may be connected to bus 210 or input / output interface 220 to transmit data with other components of computing device 200.

[0040] Figure 3A A block diagram of content module 116 is shown. Specifically, content module 116 includes a value engine 300A and an adaptive engine 350A. The value engine determines the value of a content item based on its value curve (or value curve), while the adaptive engine continuously updates the value curve of the content item in real time (or near real time) based on online user activity related to the content item.

[0041] The value engine 300A includes a feature mapper 310A, a classifier 320A, a combiner 330A, and a value module 340A. The feature mapper 310A analyzes content items (e.g., content items uploaded by users through client system 114) to extract features and generate feature vectors. The feature vectors include multiple dimensions or elements, each representing a feature of the content item. Any conventional or other natural language processing (NLP) techniques (e.g., entity extraction, relation extraction, sentiment / emotion analysis, keyword extraction, part-of-speech (POS) tagging, etc.) can be used to extract features. In cases where the content item includes audio, the feature mapper can transcribe the audio into text via any conventional or other natural language processing (NLP) techniques and / or automatic speech recognition (ASR) techniques to extract features and generate feature vectors. Features can include any number of any type of features (e.g., keywords, topics, events, word count, term frequency, word embeddings, term frequency-inverse document frequency (TF-IDF), etc.).

[0042] Classifier 320A analyzes the feature vector and contextual information of a content item from feature mapper 310A and identifies one or more value curves (e.g., stored in database system 118) of other content items associated with that content item. The classifier may employ any conventional or other nearest neighbor technique (e.g., K-nearest neighbor algorithm, etc.) to identify the nearest neighbor value curves (or content items). In this case, feature mapper 310A may weight the features in the feature vector so that the classifier can identify the value curve of a content item based on features with higher weights (e.g., value curves may be identified based on the similarity of content item features with higher weights, etc.). The identified value curves represent the values ​​of associated content items over time, and the identified value curves are combined to produce the value curve of the content item, as described below. Contextual information may include any attributes that provide context for the content item (e.g., location, brand popularity, audience demographics, purchasing power, brand, described event, content type, etc.). A portion of the contextual information may be extracted by feature mapper 310A using natural language processing (NLP).

[0043] Classifier 320A can employ one or more machine learning models to identify one or more value curves of content items. The machine learning model can be implemented using any conventional or other machine learning model (e.g., mathematical / statistical, classifier, feedforward, deep learning, recurrent, large language model (LLM), convolutional or other neural networks, etc.). For example, classifier 320A may include a neural network.

[0044] For example, a neural network may include an input layer, one or more intermediate layers (e.g., including any hidden layers), and an output layer. Each layer includes one or more neurons, where input layer neurons receive input (e.g., feature vectors of content items, etc.) and may be associated with weight values. Neurons in the intermediate and output layers are connected to one or more neurons in the preceding layer and receive the outputs of the connected neurons in the preceding layer as inputs. Each connection is associated with a weight value, and each neuron produces an output based on a weighted combination of its inputs. For some types of neural networks (e.g., recurrent neural networks), the output of a neuron may also be based on a bias value.

[0045] The weight (and bias) values ​​can be adjusted based on various training techniques. For example, a neural network machine learning can be performed using a training set of feature vectors of content items as input and the corresponding classification (e.g., the value curve of the associated content item, etc.) as the known output, where the neural network attempts to produce the known output (or classification) and uses the error from the output (e.g., the difference between the produced output and the known output) to adjust the weight (and bias) values ​​(e.g., via backpropagation or other training techniques).

[0046] In one embodiment, machine learning can be performed using a training set of feature vectors of content items and background information as input and a known value curve of the associated content item as output, wherein the neural network attempts to produce the provided output (or the value curve of the associated content item).

[0047] The dynamic pricing system integrates a diverse set of machine learning models tailored to specific functions. These models include sentiment analysis models, such as neural networks or recurrent neural networks (RNNs), used to analyze user sentiment towards content. They are trained on labeled datasets for sentiment classification. Additionally, clustering algorithms (such as K-means, hierarchical clustering, or DBSCAN) are used to group similar articles based on content attributes. These algorithms extract features and apply clustering techniques for content classification. Furthermore, reinforcement learning algorithms form part of the system's dynamic pricing model. These models optimize pricing strategies by leveraging user interactions and market dynamics, and incorporate continuous learning mechanisms to maximize revenue.

[0048] The output layer of the neural network represents the classification of the input data (e.g., value curves, etc.). For example, the categories used for classification can include the category associated with each value curve of a content item. Output layer neurons can provide a classification (or specify a particular category) representing the corresponding value curve of the associated content item. For example, output layer neurons can be associated with different categories and represent the probability that the input data belongs to the corresponding category (e.g., the probability that the input data belongs to the category associated with the corresponding value curve of the associated content item, etc.). Preferably, the category associated with the highest probability is selected as the category of the input data (or the value curve of the associated content item). Classifier 320A can identify any number of value curves of content items associated with a content item to generate the value curve of that content item. Alternatively, the classifier can be trained based on the content of the content item (or a combination of content and feature vectors) to generate value curves.

[0049] Combiner 330A processes the value curves from classifier 320A to generate value curves for content items. The combiner can select identified value curves or any portion thereof to generate value curves for new content items. For example, the combiner can filter identified value curves based on various criteria, such as using a predetermined number of identified value curves (e.g., having the highest probability), a similarity measure between the content item and content items associated with the identified value curves, the most recent value curve, etc. The selected value curves are combined by weighting the values ​​of each selected value curve to generate the value curve for the content item. The value curve represents the value of the content item over time, and content module 116 assigns a value to the content item based on this value curve. Therefore, the value of the content item changes over time according to the value curve. The weight values ​​used to combine the selected value curves can be determined based on any desired criteria, such as the probability of the selected value curves, a similarity measure between the feature vector of the content item and the feature vector of associated content items (e.g., Euclidean distance or other distances, cosine similarity, etc.).

[0050] Value module 340A accesses the corresponding value curve of a content item to extract or determine the value of the content item over time (e.g., retrieving the value corresponding to the current time from the value curve). The value module can update the value of the content item at specific time intervals (e.g., seconds, minutes, etc.) and / or in response to any update to the value curve.

[0051] The adaptive engine 350A monitors activities associated with each content item. The adaptive engine receives the value curve for each content item and updates it at predetermined time intervals (e.g., seconds, minutes, etc.) based on the monitored activities. Activities can include any online or other activities associated with the content item (e.g., clicks to access / initiate transactions, cursor hover time, content item selection, ad viewing, etc.). The updated curve and any additional contextual information from the activity monitoring are provided to the value engine 300A. The value module 340A provides updated values ​​for the content items based on the updated value curves. Additionally, the updated value curves can be used as new curve instances to train the classifier 320A, enabling the updated curves to be selected and used to determine the value curves for subsequent content items. The updated curves can be associated with additional contextual information 360A (e.g., new location, audience demographics, purchasing power, etc.) derived from monitoring activities corresponding to the content item.

[0052] Natural Language Processing (NLP) techniques play a crucial role in extracting meaningful insights from textual content within dynamic pricing systems. These techniques encompass a variety of tasks and algorithms designed to enhance the system's understanding and analysis of textual data. First, text preprocessing tasks, such as word segmentation, stemming, and stop word removal, are performed to clean and prepare the textual data for further analysis. Second, Named Entity Recognition (NER) tasks are used to identify entities in the text, such as people and locations, thereby enhancing the understanding of the content context. Finally, the system leverages topic modeling algorithms, such as Latent Dirichlet Allocation (LDA) and Nonnegative Matrix Factorization (NMF), to identify general topics in the text, facilitating more nuanced analysis and decision-making processes.

[0053] Figure 3B This is a block diagram of an architecture for generating content vectors using a content genome. In embodiment 300B of the present invention, a digital content processor 301B is configured to dynamically generate content vectors 303B based on a content genome 302B. The content genome 302B includes multiple features 304B, such as quality 351B, originality 352B, rating 353B, seasonality 354B, user engagement 355B, relevance 356B, etc. The system employs sophisticated algorithms to analyze the inherent content genome 302B of the content. Utilizing advanced machine learning techniques, the digital content processor 301B autonomously identifies key patterns and relationships within the content, enabling the creation of content vectors 303B that encompass subtle information about the content. The dynamic vector generation process adapts to the constantly evolving nature of the content, ensuring adaptive and responsive representation. Table 1 shows the attribute vectors input to the content genome database or from the analysis.

[0054] Table 1

[0055]

[0056]

[0057] In one embodiment of the invention, the value of a content item is determined by: based on the value of one or more content items associated with the content comparison result of the content item, and by examining the inherent characteristics of multiple aspects of a user referred to as the user genome and by examining the inherent characteristics of multiple aspects of a publisher referred to as the publisher genome.

[0058] In another embodiment of the invention, the publisher genome comprises a multi-layered profile that captures the content publisher's identity, behavior, and historical performance metrics within a dynamic pricing framework. This comprehensive profile facilitates publisher segmentation to develop tailored pricing strategies, incorporating elements such as identity and reputation, content mix, engagement metrics, monetization history, user demographics, and compliance and ethical standards. By leveraging the publisher genome, the pricing system ensures publishers align with optimal pricing structures, thereby improving the transparency and efficiency of content valuation and distribution.

[0059] In another embodiment of the invention, the user genome serves as a dynamic profile encompassing user preferences, behaviors, and interactions to drive personalized and responsive content pricing. This evolving profile includes user preferences, interaction history, engagement metrics, dynamic pricing responsiveness, content attributes, trends and seasonality, and competitive content consumption patterns. By integrating the user genome with the publisher genome, the pricing system achieves a symbiotic relationship that enables dynamic pricing adjustments based on both content value and user preferences, thereby improving user satisfaction and optimizing publisher revenue.

[0060] 57 Figure 4 A method 400 for determining the value of a content item in real time (e.g., via content module 116, interface module 120, server system 110, and / or client system 114) according to an embodiment of the present invention is illustrated. Initially, reference value curves are generated at operation for an initial set of content items. The reference value curves may be generated based on experimental data (e.g., randomized controlled trials (RCTs) of user activity associated with the initial set of content items). The reference value curves represent predictions of how the values ​​of the initial set of content items change over time. The initial set of content items and the corresponding value curves are used to generate value curves for subsequent content items, as described below.

[0061] While processing new content items, the value curve and value of the content items are continuously updated. The update of the value curve essentially corrects value predictions based on user activity. Specifically, when a user (e.g., through client system 114) provides (e.g., uploads or registers) a new content item determined in operation 410, in operation 415, value engine 300A receives the new content item for processing. The new content item can be any type of digital or electronic item or object (e.g., document, webpage, file, data object, etc.) containing any type or combination of data (e.g., text, multimedia, video, audio, images, streaming data, etc.). For example, a new content item may include news or other articles, websites or pages, papers, documents, program code or applications, audio recordings, videos, images, live or recorded podcasts, streaming media, streaming of live events, blogs, messages, chats, conversations or other posts, any combination of the above, etc. Value module 340A can generate a new content ledger or other data structure to record future transactions or other activities involving the new content item. The ledger can be stored in database system 118.

[0062] The initial value curve for the new content item is determined based on the value curves of previous content items. Weight values ​​are used to determine the relative importance (or contribution) of the value curve of each previous content item when generating the value curve of the new content item. As described below, the value curve of the new content item is determined by a weighted sum of the values ​​from the value curves of previous content items.

[0063] The value curve of a content item represents the value of the content item over time and can be represented as a polynomial function of degree n (e.g., a regression polynomial). The degree of the polynomial can be any value, but is preferably greater than or equal to 2. The polynomial function of the value curve can be represented as:

[0064] V(X) = C0 + C1X + C2X 2 +…+C n X n Where V(X) represents a value function or value curve of the variable X (e.g., time), and C0 to C n C represents the coefficient, and C0 represents the Y-intercept.

[0065] Properties or attributes of a polynomial function (e.g., coefficients, degree, etc.) can be stored, and these properties or attributes can then be retrieved and applied based on the expression to produce data points along the value curve (e.g., the value of a content item at a given time). This avoids storing and calculating all data points of the value curve, thus improving processing speed and saving memory. The value curve of a content item can be any function, graph, graphical element, or other object (e.g., table, list, etc.) that expresses the value of the content item over time (e.g., a flat curve or straight line, any curve, any combination of curved and / or straight sections, a step function, or a graph, etc.).

[0066] To identify the value curve of a previous content item, in operation 420, feature mapper 310A analyzes the new content item and extracts features from it to generate a feature vector. The feature vector includes multiple dimensions or elements, each representing a corresponding feature of the new content item. Features can include any number of any type of feature (e.g., keywords, topics, events, word count, term frequency, word embeddings, term frequency-inverse document frequency (TF-IDF), etc.). Furthermore, the feature mapper can extract contextual information from the new content item. Contextual information can include any attributes that provide context for the new content item (e.g., location, brand, described event, content type, etc.). The feature mapper can employ any conventional or other natural language processing (NLP) techniques (e.g., entity extraction, relation extraction, sentiment / emotion analysis, keyword extraction, part-of-speech (POS) taggers, etc.) to identify and / or extract features and / or contextual information. In the case where the new content item includes audio, the feature mapper can transcribe the audio into text via any conventional or other natural language processing (NLP) techniques and / or automatic speech recognition (ASR) techniques to extract features and generate a feature vector.

[0067] In operation 425, classifier 320A processes the feature vector and contextual information of the new content item (e.g., through a machine learning model, etc.) and identifies one or more value curves (e.g., stored in database system 118, etc.) of other content items associated with the new content item. The value curves of previous content items represent the values ​​of previous content items over time. Contextual information may include any attributes that provide context for the new content item (e.g., location, brand popularity, audience demographics, purchasing power, brand, the event described, content type, etc.). Contextual information may be provided by the user and / or extracted by feature mapper 310A using natural language processing (NLP) techniques as described above.

[0068] In one embodiment, during operation 450, feature mapper 310A identifies adjacent content of the new content item. Furthermore, during operation 451, feature mapper 310A generates, for example... Figure 3BThe content genome is shown. Identification of adjacent content can be performed by classifier 320A and / or combiner 330A.

[0069] In operation 430, combiner 330A processes the identified value curves from classifier 320A to generate a value curve for a new content item. The combiner can select either the identified value curves or any portion thereof for use in generating the value curve for the new content item. For example, the combiner can filter the identified value curves based on various criteria, such as using a predetermined number of identified value curves (e.g., those with the highest probability), a similarity measure between the content item and content items associated with the identified value curves, the most recent value curve, etc. For instance, the combiner can select a value curve with the highest probability that is associated with content items whose similarity to the new content item exceeds a similarity threshold, and / or is within a specific time interval.

[0070] In one embodiment of the invention, in operation 452, combiner 330A feeds the combined value curve from operation 430 and the generated content genome from operation 451 to the machine learning model.

[0071] Alternatively, classifier 320A can be trained using additional features for filtering and produce a filtered set of value curves (e.g., generally similar to the filtered value curves of combiner 330A described above) as the value curves identified by combiner 330A. For example, the classifier can employ any conventional or other nearest neighbor technique (e.g., K-nearest neighbor, etc.) to identify the nearest value curves as described above.

[0072] Value curve filtering (e.g., via classifier 320A and / or combiner 330A) can be used in various scenarios. For example, a new content item may refer to the same event (or include the same content) as a previous content item. In this case, the similarity measure between the feature vectors of the content items can be high, indicating very close similarity. The similarity measure can use weighted features that emphasize features that can identify the same event or content (e.g., topic, keywords, time features, location of the event, nature of the event, brand, proximity to the event, purchasing power, common location in space and time, user type, etc.). In these cases, value curves can be filtered based on time requirements (e.g., within a specific time interval, etc.).

[0073] Furthermore, a new content item can refer to a new occurrence of a previous type of event within a previous content item. Similarity measures can use weighted features that emphasize characteristics that identify events of the same type (e.g., topic, audience demographics, event location, brand, purchasing power, user type, event nature, etc.). In these cases, value curves can be filtered based on weighted features to select value curves for content items associated with the same type of event.

[0074] The selected value curves of the previous content item are combined based on a weighted sum of the values ​​of each value curve to produce the value curve of the new content item. The formula can be expressed as follows:

[0075] N0(value,t)=∑(W i *N i (value,t))

[0076] Where N0 represents the value curve of the new content item, i represents the index of several previous content items (or value curves), t represents time, "value" represents the value of the value curve at time t, and W i Let N represent the weight value of the i-th content item, and N... i This represents the value curve of the i-th content item.

[0077] Summation is performed on the selected value curves of the already analyzed content items. The value curve of the new content item at time t is obtained by multiplying each weight value (W) by the corresponding value N of the associated content item's value curve (at time t). i The result is calculated by summing the results of previous content items (value,t). The resulting value represents the value curve of the new content item being analyzed.

[0078] Weights used to combine selected value curves can be determined based on any desired criteria. For example, traditional or other Natural Language Processing (NLP) techniques can be used to determine weights, taking into account similarity between new and previous content items and / or other factors such as brand popularity, audience demographics, purchasing power, brand, described event, content type, word count, keywords, location / situation, reader type, language model, etc. Some factors (e.g., brand, purchasing power, reader type, factors without quantifiable measures, etc.) can be associated with predetermined weights that can be used to determine weight values. For example, well-known brands can be assigned higher weights compared to lesser-known brands. These predetermined weights can be summed or otherwise combined, and / or used in conjunction with similarity or other metrics to determine the weight values ​​of the value curves. Alternatively, neural networks can be used in Natural Language Processing (NLP) to analyze content items and provide insights into weight values ​​(e.g., representing relevant factors, etc.).

[0079] For example, the probability of the identified value curve from classifier 320A can be used as a weight value (or combined with weight values ​​from other factors). Furthermore, the similarity measure between the feature vectors of new content items and the feature vectors of previous content items can be used as a weight value (or combined with weight values ​​from other factors). For instance, the similarity measure between feature vectors can include any conventional or other distance or similarity measure (e.g., Euclidean distance or other distances, cosine similarity, etc.).

[0080] Weight values ​​can change dynamically over time based on monitored user activity. For example, changes in user activity that meets a threshold (e.g., access, transactions, etc.) (e.g., an increase or decrease in an activity exceeding a threshold amount) may trigger changes in feature weight values. For instance, feature weight values ​​can be adjusted by a set amount or by an amount corresponding to a percentage change in that activity (e.g., a 10% increase (or decrease) results in a 10% increase (or decrease) in the weight value of the corresponding feature).

[0081] Furthermore, portions of the value curves of associated content items can be used to generate corresponding portions of the value curves of new content items. For example, a portion of the value curve of a new content item (corresponding to the desired time interval) can be generated from the corresponding portions of the value curves of associated content items that have certain similar characteristics (e.g., user type, purchasing power, etc.). The values ​​of these value curves for the desired time interval are weighted and summed in a manner substantially similar to that described above to generate the corresponding portion of the value curve of the new content item. Similarly, unique users or outliers can have their value curve portions (having certain similar characteristics and corresponding to the unique user's time) combined in a manner substantially similar to that described above to generate the corresponding portion of the value curve of the new content item.

[0082] Once the value curve for a new content item is determined, this value curve represents a prediction of how the value of the new content item will change over time. Based on user activity, this curve is updated along with the value curves of other content items (e.g., to continuously correct values ​​or value predictions). In operation 452, the adaptive engine 350A continuously monitors user activity, including user interactions, content consumption patterns, and engagement metrics. Specifically, in operation 435, the adaptive engine 350A updates the value curves of content items based on user activity (e.g., ledgers of content items recorded in database system 118). When updating value curves, in operation 440, the adaptive engine provides the updated value curves to the value module 340A to update the values ​​of the corresponding content items based on these value curves. The value curves are continuously tracked over time, with the values ​​of content items retrieved from the corresponding value curves at the current time. This allows the values ​​of content items to be continuously updated in real time. In operations 435 and 440, the value curves and values ​​of content items are continuously updated (e.g., to continuously correct values ​​or value predictions), and this is performed during the processing of new content items as described above.

[0083] The Adaptive Engine 350A preferably utilizes multiple parallel hardware or virtual processors to process and / or update value curves and values ​​of content items in parallel, thereby improving computational performance. The Adaptive Engine can allocate processing (e.g., updating value curves, updating the values ​​of content items, etc.) to processors in various ways. For example, the Adaptive Engine can allocate processing to the next processor or the next available processor in a round-robin or sequential manner. Furthermore, the Adaptive Engine can employ any conventional or other load balancing technique to allocate processing based on metrics representing the processing load on the processor (e.g., queued processes or jobs, throughput, processing speed, hardware or memory usage, etc.). In this case, processing can be allocated to the processor that provides the fastest completion time (e.g., with the lowest processing load, fastest throughput, or processing speed, etc.). Load balancing optimizes processing speed and performance.

[0084] The value curve is updated and new content items are processed as described above until a termination condition (e.g., power failure, interruption, etc.) is determined in operation 445.

[0085] Figure 5 illustrates a method for updating value curves according to an embodiment of the present invention (e.g., via adaptive engine 350A and server system 110 and / or client system 114). This can correspond to Figure 4 Operation 435. Initially, the adaptive engine 350A improves the value curve of a content item based on real-time user behavior to generate an updated value curve (e.g., reflecting corrected values ​​or value predictions based on user behavior). As the user interacts with the content item, the updated value curve is calculated in real time using adaptive techniques. The updated value curve reflects the current demand for the content item and is used to further improve the value curve in subsequent iterations of the adaptive techniques.

[0086] External factors influencing the updated value curve can also be considered, such as changes in the overall market, shifts in consumer behavior, and other relevant events. Neural networks or other natural language processing (NLP) techniques can be used in adaptive techniques to analyze real-time data and provide insights into the factors.

[0087] As more data is collected over time, the Adaptive Engine 350A continues to improve its predictions and adjusts the value curves accordingly. The initial and updated value curves become increasingly similar, reflecting the ability to accurately predict the demand for content items.

[0088] The Adaptive Engine 350A maximizes revenue generated from transactions of content items. This is achieved by continuously refining the initial and updated value curves to accurately reflect demand and optimize values ​​in real time. Neural networks or other Natural Language Processing (NLP) techniques can be used to analyze data, provide insights, and improve predictions to optimize values ​​for maximum revenue.

[0089] In operation 505A, the adaptive engine 350A monitors user activity or other activities related to content items. This activity can be captured by the user interface on client system 114 (e.g., via interface module 120, etc.) and provided to the adaptive engine 350A on server system 110 (e.g., directly or by being stored in database system 118). This activity can include any online activity or other activity related to the content item (e.g., clicks to access / initiate transactions, cursor hover time, content selection, ad viewing, etc.), and information about this activity can be stored in the ledger of the corresponding content item, as described above.

[0090] The Adaptive Engine 350A can update the value curve of content items at any desired time interval (e.g., seconds, minutes, hours, etc.) to enable the measurement and accumulation of activity data. The time interval can be the same or different for different content items. Additionally, the Adaptive Engine 350A can update the value curve of content items before the time interval expires in response to events, such as certain behaviors in user activity data. These behaviors can include any outliers, anomalies, or activity shifts. For example, a sharp increase (or surge) or decrease in transactions or other activities (e.g., exceeding a threshold, exceeding a range, etc.) might trigger an update of the corresponding content item's value curve.

[0091] When operation 510A determines that the value curve of a content item needs to be updated (e.g., time interval expires, event occurs, etc.), the adaptive engine 350A updates the value curve in operation 515A. The adaptive engine can use various techniques to update the value curve.

[0092] For example, the Adaptive Engine 350A can update the value curve based on its sensitivity or resilience to changes in value. In this case, the activity measurement (e.g., purchases or other transactions) is compared relative to changes in value.

[0093] For example, the adaptive engine 350A can monitor and obtain a series of activity measurements or observations Y(t) (e.g., transaction volume, etc.) during the time interval τ prior to the update, Y(t) = Y(0), Y(1), ..., Y(τ). The measurements can be stored in the corresponding ledger of the content items in database system 118 and can be retrieved from the corresponding ledger of the content items in database system 118. The sensitivity or resilience d(t) at a certain point in time can be determined based on the corresponding difference between the change in the activity measurement at that point in time and the value of the content item within that time interval. The sensitivity or resilience at a certain point in time can be expressed as:

[0094]

[0095] Where Y(t) is the activity measurement value at time t, Y(t-1) is the activity measurement value at time t-1, V0(t) is the value of the content item at time t, and V0(t-1) is the value of the content item at time t-1.

[0096] The aggregation (e.g., average) sensitivity or elasticity E(τ) over the time interval is determined based on the sensitivity or elasticity at each time point within that time interval, and can be expressed as:

[0097] E(τ)= 1 / τ ∑d(t), t=1 to τ.

[0098] Aggregate sensitivity or elasticity can be used to update value curves in various ways. For example, aggregate elasticity can be used to determine a scaling factor applied to the current value curve. The scaling factor α can be determined based on the relative position of the overall elasticity within the range of elasticity values. For instance, the range of elasticity values ​​can be determined by the minimum value e of the elasticity. min And the maximum value of elasticity e max These minimum and maximum values ​​can be determined based on the range of values ​​of the active measurements applied to the overall elasticity expression above.

[0099] The scaling factor α can be expressed as:

[0100] Where 0≤α≤1

[0101] A scaling factor can be applied to the current value curve (e.g., multiplying the scaling factor by the value in the current value curve) to produce an updated value curve. This can be represented as:

[0102] V1(t)=α*V0(t)

[0103] Where V1(t) represents the updated value curve, α represents the scaling factor, and V0(t) represents the previous value curve.

[0104] Since the value curve can be represented as a polynomial function with coefficients as described above, a scaling factor can be simply applied to these coefficients to update the value curve. This avoids storing and computing all data points, thus improving processing speed and saving memory.

[0105] Alternatively, a scaling factor α can be applied to the Y-intercept of the value curve to update the value curve (e.g., by translating the value curve by the scaling factor, etc.). In this case, the scaling factor can be applied (e.g., by multiplying, etc.) to a constant value that represents the Y-intercept in the polynomial equation of the value curve. This avoids storing and calculating all data points, thus improving processing speed and saving memory.

[0106] In this context, the aggregate elasticity can be used as the gradient of the loss function of the parameters of the current value curve or function (e.g., activity, value, etc.) and represents the direction of the update, while the adjustment value from the adjustment curve can be used as the step size (or learning rate). The adjustment value is applied to the aggregate elasticity to form an adjustment to the current value curve. The adjustment curve can be a flat curve (e.g., applying a constant step size to the aggregate elasticity to achieve a set increase / decrease over time), or some other function or shape (e.g., reflecting user activity measurements over time).

[0107] The update of the value curve using learning techniques (e.g., gradient descent) can be expressed as:

[0108] V1(t)=V0(t)+E(τ)*V ADJ (t)

[0109] Where V1(t) represents the updated value curve, V0(t) represents the current value curve, E(τ) represents the aggregate elasticity (e.g., the gradient of the loss function used as a parameter of the current value curve), and V ADJ (t) represents the adjustment curve (e.g., step size or learning rate).

[0110] Since the value curve can be represented as a polynomial function with coefficients as described above, adjustments can be simply applied to the parameters of the polynomial function to update the value curve. This avoids storing and calculating all data points, thus improving processing speed and saving memory.

[0111] Furthermore, various traditional or other techniques can be used to update value curves based on user activity (e.g., stochastic gradient descent (SGD), adaptive gradient algorithm (AdaGrad), root mean square propagation (RMSProp), adaptive moment estimation (Adam), etc.). For example, stochastic gradient descent (SGD) is similar to gradient descent but updates parameters based on a randomly selected subset of training data. Adaptive gradient algorithm (AdaGrad) adjusts the learning rate of each parameter based on historical gradient information (e.g., parameters with sparse gradients have higher learning rates, while parameters with dense gradients have lower learning rates). Root mean square propagation uses the moving average of the squared gradient to scale the learning rate of each parameter. Adaptive moment estimation (Adam) combines gradient descent, SGD, AdaGrad, and RMSProp, and adjusts the learning rate of each parameter based on historical gradient information and the historical moving average of the gradient.

[0112] The value curve is updated as described above until operation 520A determines that a termination condition has occurred (e.g., power failure, interruption, etc.).

[0113] Figure 5B A block diagram 500B, illustrating an exemplary embodiment of the present disclosure, shows a method of generating values ​​for digital content using data from a crawler and employing a message broker and a database. A crawler 501B collects Uniform Resource Locators (URLs) from multiple sources, crawls their raw Hypertext Markup Language (HTML), and sends them to a processor 502B via a message broker (e.g., RabbitMQ). The processor 502B takes the raw HTML as input and parses it to extract article content and metadata. A vectorizer 503B and an indexer 504B index the digital content information and vectors for easy querying in a manner similar to retrieving similar articles, and then use the message broker to signal that the digital content is ready to be rendered by a value curve generator 505B. A content database (DB) 506B is kept updated by the indexer 504B, and a vector database 507B is maintained by the indexer 504B for use by the value curve generator 505B. The value curve generator 505B uses a content database (DB) 506B and a vector database 507B to generate the output of digital content. Furthermore, the curve database 508B is maintained by the value curve generator 505B.

[0114] In one embodiment of the invention, the crawler 501B collects Uniform Resource Locators (URLs) from multiple sources, crawls their raw Hypertext Markup Language (HTML), and sends them to a processor via a message broker (e.g., RabbitMQ, Apache Kafka, ZeroMQ, etc.). The processor takes the raw HTML as input and parses it to extract the content and metadata of the digital content. Furthermore, a reinforcement learning engine indexes the digital content information and vectors to facilitate easy querying in a manner similar to retrieving digital content, and then uses the message broker to signal that the digital content is ready to be presented along with its associated information.

[0115] Figure 5C This diagram illustrates the process flowchart of a feedback loop from users, which is integrated into the system to retrain the machine learning model. The dynamic pricing system incorporates feedback loops to improve the accuracy and effectiveness of its machine learning model. These feedback loops are designed to continuously collect user feedback and integrate it into the model training process, thereby improving its performance over time. The feedback mechanism provides users with a dedicated interface, such as ratings or reviews, through which they can express their opinions on content relevance, quality, or pricing satisfaction. This feedback is then systematically incorporated into the training process, ensuring the model adapts to evolving user preferences and market dynamics. Additionally, the system implements a traditional model update procedure, where these models are retrained at predefined intervals based on the collected feedback. Through this process, the algorithm is tuned to reduce bias and improve predictions, ultimately leading to a more accurate and responsive pricing strategy. Continuous learning mechanisms are also employed to ensure the model evolves in real time, enabling it to effectively adapt to evolving user preferences and market trends.

[0116] As per this instruction manual Figure 5B The described feedback loop mechanism outlines a multi-step process for incorporating user feedback into a dynamic pricing system. First, in step 501C, users are provided with a designated feedback channel, allowing them to provide ratings or reviews indicating content relevance, quality, or pricing satisfaction. As illustrated in step 502C, this feedback is then collected and stored for subsequent processing within the system. Following collection, as detailed in step 503C, the feedback data is systematically integrated into the training model of the dynamic pricing system. After integration, in step 504C, the system adjusts its algorithmic framework based on the absorbed feedback to mitigate bias and improve predictions. Furthermore, step 505C emphasizes the iterative nature of this process, where the algorithm continuously adapts to ensure an optimal pricing strategy. Finally, step 506C sets up a decision box designed to determine the appropriate timing for initiating the next training interval, thereby facilitating timely updates and enhancements to the pricing model in response to evolving user preferences and market dynamics.

[0117] Figure 5D This diagram illustrates a process flowchart for dynamically adjusting weight values ​​based on surges. The dynamic pricing system incorporates mechanisms for dynamically adjusting weight values ​​to ensure adaptability and responsiveness to changing user preferences and market dynamics. These mechanisms utilize user activity monitoring, including real-time tracking of user interactions, content consumption patterns, and engagement metrics, to identify trends and patterns. By continuously monitoring user behavior, the system can detect sudden surges in interactions with articles on specific topics, indicating changing user preferences or emerging trends. Once identified, these trigger signals prompt the system to adjust the weight values ​​associated with relevant content attributes. For example, if interest in articles on a particular topic increases, the system can adjust the weights associated with that topic to reflect its higher relevance or value to users. These adjustments have a direct impact on the overall pricing model, affecting the prices allocated to relevant content. For instance, an increased weight for trending topics may lead to higher prices for articles on those topics, reflecting their perceived value and demand among users. By dynamically adjusting weight values ​​in response to evolving user behavior and market trends, the system ensures its pricing strategy remains relevant and competitive, ultimately improving user satisfaction and maximizing revenue.

[0118] Patent specification Figure 5B The process outlined herein requires monitoring for sudden surges in interaction with the article, achieved through a multi-step approach. Initially, in step 501D, the system continuously monitors various factors to identify sudden increases in user engagement with the content. Following detection, as shown in step 502D, the decision box evaluates whether a surge or drift has occurred in the interaction to determine subsequent course of action. If a surge is detected, step 503D requires automatic adjustment of the algorithm weights through retraining to optimize the dynamic pricing model in response to evolving user behavior. Subsequently, step 504D emphasizes the crucial role of these weight adjustments in shaping the overall pricing model, ensuring its responsiveness to changing market dynamics and user preferences.

[0119] exist Figure 6A and 6B An example of generating and updating value curves for new content items is shown. Initially, reference value curves A0, B0, and C0 are generated for an initial set of content items 620, 630, and 640. The reference value curves indicate the value of the corresponding content item over time and can be generated based on experimental data (e.g., randomized controlled trials (RCTs) of user activity related to the content item), as described above. The reference value curves are updated based on user activity in a substantially similar manner to that described above to produce an updated reference value curve A1 (e.g., as shown above). Figure 6A As shown, by F A1 (x) represents), B1 (for example, such as Figure 6A As shown, by F B1 (x) represents) and C1 (for example, such as Figure 6A As shown, by F C1 (x) represents that it can be stored as a new value curve in database system 118.

[0120] A new content item 610 is provided, and its value function N0 is determined based on the update value curves of previous content items 620, 630, and 640. Specifically, feature mapper 310A analyzes the new content item and extracts features from it, thereby generating a feature vector in a manner largely similar to that described above. Classifier 320A processes the feature vector and contextual information of the new content item in a manner largely similar to that described above (e.g., through a machine learning model, etc.), and identifies the update value curves of content items 620, 630, and 640 as associated with the new content item 610.

[0121] Combiner 330A processes the identified value curves from classifier 320A in largely the same manner as described above to produce the value curve N0 of the new content item 610 (e.g., as shown above). Figure 6A As shown, by F N0 (x) represents the value curve. The value curves are combined by a weighted sum of the values ​​of each updated value curve for content items 620, 630, and 640 to produce the value curve for the new content item 610. This can be represented as follows:

[0122] F N0 (X)=(W1*F A1 (X))+(W2*F B1 (X))+(W3*F C1 (X))

[0123] Among them, F N0 (X) represents the value curve of the new content item, F A1 (X) represents the updated value curve for content item 620, F B1 (X) represents the updated value curve for content item 630, F C1 (X) represents the updated value curve of content item 640, and W1, W2 and W3 are weight values.

[0124] Once the value curve for the new content item 610 is determined, the adaptive engine 350A updates the curve based on user activity in largely the same manner as described above, to produce an updated value curve F1 (e.g., as...). Figure 6B As shown, by F N1 (X) indicates). The value curves for other content items 620, 630, and 640 are also updated by the Adaptive Engine 350A based on user activity in a largely similar manner to the above.

[0125] After the initial value curves for a content item are generated, new content items and various value curve updates can be performed. In this case, the content item can be resubmitted to classifier 320A to identify a new set of value curves for the associated content item. This new set of value curves may include updated value curves and / or value curves for subsequently added content items. Additionally, the context of the content item may have changed, which may also generate a new set of value curves from classifier 320A. This set of value curves can be combined in a generally similar manner to produce updated value curves for the content item. Content items can be resubmitted to the classifier at any desired time interval or condition (e.g., daily, weekly, in response to a certain amount of updates, in response to a certain amount of new content items, etc.). This time interval is preferably longer than the time interval used to update the value curves. This allows the value of the content item to vary based on user activity related to the content item (e.g., in addition to user activity for the content item itself).

[0126] Additionally, content items can include content associated with a live event (e.g., podcasts, streaming of the event, etc.). For example, the audio of the event can be transcribed into text and analyzed via any conventional or other Natural Language Processing (NLP) and / or Automatic Speech Recognition (ASR) techniques. An initial topic can be determined, and value curves and corresponding values ​​for content items can be generated in a manner largely consistent with the above. However, when a topic change (e.g., during a podcast, interview, etc.) is detected (e.g., via any conventional or other Natural Language Processing (NLP) techniques, the content item is resubmitted to the value engine 300A to determine a new value curve based on the new topic in a manner largely consistent with the above. In this case, the new topic can be associated with new features extracted by the feature mapper 310A, which enable the classifier 320A to identify different groups of value curves for content items associated with the new topic in a manner largely consistent with the above. These identified value curves can be combined to generate a new value curve for the content item, which values ​​the content item based on the new topic. The new value curve of the content item is updated as described above, where the value curve of the content item (and therefore the value) can change dynamically as the theme changes in the live event.

[0127] exist Figure 7A and Figure 7BAn exemplary graphical user interface 700 providing real-time changing content items and values ​​is illustrated. Initially, a value module 340A can generate the graphical user interface 700 for presentation on a client system 114. The value module can receive search parameters input by the user from the client system and perform a search for desired content items that satisfy the search parameters. The user interface 700 includes a display area 705 that presents content items 710 (e.g., from a search, etc.). Each content item 710 is provided with corresponding information (e.g., title, author, source, etc.), a value area 715 representing the value of the content item, and a trigger 720 for obtaining the content item (e.g., a link, a transaction trigger for obtaining the content item, etc.). Additionally, the value area 715 includes an indicator 725 (e.g., an arrow, etc.) that indicates the trend or direction of the value (e.g., increasing or decreasing, etc.). The indicator may also be color-coded to indicate the trend or direction (e.g., green for increasing, red for decreasing, yellow for stable, etc.). For example, the content item corresponds to news or other articles, and the value corresponds to a purchase price. However, any type of content and value can be used.

[0128] The value of content item 710 is determined by value module 340A based on a corresponding value curve and is continuously updated over time. Therefore, the value provided by the graphical user interface 700 can be continuously updated or changed (e.g., similar to a continuously updating and changing stock quote auto-collector). Additionally, content items can be sorted based on their values ​​(e.g., ascending or descending order). In this case, the order of content items can dynamically change according to the updated or changed values.

[0129] refer to Figure 7B Users can select or trigger presented content items to display further information. For example, the graphical user interface 700 may present content item 730 in display area 705. The content item is provided with corresponding information (e.g., title, author, source, link, etc.), a value area 715 representing the value of the content item, and a trigger 720 for retrieving the content item (e.g., a link, a transaction trigger for retrieving the content item, etc.). Additionally, the value area 715 includes an indicator 725 (e.g., an arrow, etc.) indicating the trend or direction of the value (e.g., increase or decrease, etc.). The indicator may also be color-coded to indicate the trend or direction (e.g., green for increase, red for decrease, yellow for stability, etc.).

[0130] Users can also select or trigger content item 730 by manipulating an input device (e.g., a mouse) to move the cursor near or over content item 730. Users can trigger input mechanisms (e.g., selection operations) or allow the cursor to hover over content item 730 for a sufficient period of time. In response to selection or hovering operations, value module 340A can provide a data area 750 that provides additional information about content item 730. Data area 750 includes a graphical representation of the content item's value within a selectable time period (e.g., hours, days, months, etc.).

[0131] In addition, refer to Figure 7C The data range 750 can represent the demand level 755 (e.g., a series of bars on a range scale, demand values, or percentages) and the accuracy or confidence level 760. The data range can be updated in real time as information is collected and values ​​are updated.

[0132] Embodiments of the present invention can provide various technical and other advantages. In one embodiment, a machine learning model can be continuously updated (or trained) based on feedback related to online user activity. For example, a classifier can initially identify relevant value curves. Once feedback related to content items is provided (e.g., continuous monitoring of user activity, additional background information for monitoring, etc.), the machine learning model can be updated (or trained) based on the feedback. For example, user activity and / or background information can be used to update or train the machine learning model (e.g., to update or train the machine learning model to adjust the probability of the value curve of the content item (or change the classification), etc.). Therefore, as user activity is continuously monitored, the machine learning model can continuously evolve (or be trained) to learn more attributes related to the value curves.

[0133] This invention utilizes past content, value performance, and user data to train value models for various content and formats. It employs a vast amount of past content, its past value performance, and user behavior and transaction data to train value models across different content types (such as sports, politics, medicine, business, finance, biology, and science) and various formats (such as articles, videos, news articles, blogs, etc.). Furthermore, deep learning algorithms can be used to analyze real-time user behavior and past performance to generate appropriate value performance. Deep learning algorithms can be used at any given moment to analyze and understand the potential transactional behavior of all users for each new content item based on real-time user behavior, similar content, and past content value performance and past user transaction performance to generate appropriate values.

[0134] Embodiments of this invention can continuously improve their value models and predictions through unsupervised learning to identify economic patterns and relationships. This allows them to continuously learn and improve their value models through unsupervised learning, enabling them to identify economic patterns and relationships in value performance without explicit guidance. Furthermore, embodiments of this invention utilize Natural Language Processing (NLP), adaptive algorithms, and neural networks to provide accurate and equitable value over time for both users who create and users who receive content. This invention leverages Natural Language Processing (NLP) and its adaptive algorithms and neural networks to become more accurate and efficient in its value decisions over time, ensuring that both users who create and users who receive content receive equitable value.

[0135] In one embodiment of the invention, a method for generating quantitative market value (QMV) forecasts with uncertainty is disclosed. Given a set of contents, at time t, the forecast (now including uncertainty) for a specific content C is represented by the following equation:

[0136]

[0137] Among them, Forecast c (t) is the predicted value of content C at time t, including modeling uncertainty. f is a function that combines the content genome of content C with the weighted historical QMV of similar content. It is a content genome, a numerical vector representing the characteristics of content C. SC i Let represent the i-th content similar to content C. k is the number of similar content items considered for prediction, determined by the similarity score. i It is the weight (normalized similarity score) assigned to the i-th similar content. Y(t) represents the QMV of the i-th similar content at historical time t'. Y(t) represents the variability of time t, modeled as a stochastic process, adding uncertainty to the prediction. Y(t) is a normally distributed random variable with a mean μ and a standard deviation σ(t) over time. The distinction between t and t' is crucial: t refers to the point in time where the prediction is made, either now or in the future. t represents the moment when we want to use the prediction function to predict the outcome. When we mention t+1, t+2, ..., we are considering predicting future points in time, extending from the present into the future. t': represents each similar content SC i The historical time point. t' is the time when similar content was observed in the past QMV. Historical data at time t' is used to inform predictions for the current or future time t. t' reflects the past performance or behavior of content similar to content C, and it should be understood that each SC... iIt has its own historical timeline t'+1, t'+2, ... marking subsequent points in history.

[0138] In another embodiment of the invention, a method for generating quantitative market value (QMV) is disclosed. At an initial time t = 0, an initial point estimate of the prediction for a specific content C is defined as the quantitative market value (QMV). QMV represents a baseline value from which future predictions are derived.

[0139]

[0140] Among them, QMV c This is the quantitative market value of content C at the initial prediction point t=0. o It is a function that combines the content genome of content C with the weighted sum of historical values ​​of similar content. It is a content genome, a numerical vector representing the characteristics of content C. SC i Let represent the i-th content similar to content C. k is the number of similar content items considered for prediction, determined by the similarity score. i It is the weight (normalized similarity score) assigned to the i-th similar content. Let be the QMV of the i-th similar content at historical time t'. At t=0, the QMV of content C is used as the base value for prediction, reflecting the inherent attributes of the content and the aggregated weighted historical performance of similar content. This initial estimate is crucial for setting a baseline from which future predictions are projected, providing a starting point that encompasses the characteristics of the content and empirical data on the market performance of related content.

[0141] In an embodiment of the present invention, a method for generating an adaptive quantitative market value (QMV) is disclosed, wherein the generated adaptive QMV can be used by partners. The adaptive QMV model of content C is represented as QMV at time t. c (t), integrating several key components to dynamically adjust QMV based on current and recent trend data: the equation representing QMV adaptation is given by the following:

[0142]

[0143] in, It is a numerical vector of the inherent attributes of the captured content, reflecting its fundamental characteristics. User Engagement Factor (UE) C (t) measures the level of user interaction with content C at time t, providing insights into content appeal and audience engagement. Entity and event relevance factor EER C(t) represents the timeliness and importance of entities and events related to content C at time t, and the QMV is adjusted based on external influences that may affect content performance. The weighted QMV of background similar content incorporates the data from similar content (SC) at the previous time step (t-1). i The aggregated weighted QMV value leverages recent performance trends. These components collectively enhance the model's ability to provide nuanced and dynamic predictions. Incorporating user engagement and relevance to relevant entities and events, along with inherent content attributes and recent performance trend data, ensures comprehensive and real-time relevance of the QMV.

[0144] In another embodiment of the invention, a method for generating an adaptive Quantitative Market Value (QMV) value is disclosed, wherein this value can be shared with non-partners. At time t, in the absence of user engagement (UE) data, the adaptive QMV model for content C simplifies the equation by removing the UE term. This modification ensures the applicability of the model when non-partner publishers / individual creators cannot provide specific engagement metrics. The revised equation is given by:

[0145]

[0146] While no direct user engagement metric exists, the improved version of the QMV adaptive model employs a multi-layered approach by integrating content genome, entity and event importance, and recent performance metrics from similar content. A key element here is the price propagation mechanism. This mechanism ensures that the quantitative market value (QMV) remains responsive and accurate, even for content lacking direct user engagement metrics. It does this by leveraging insights from similar content with direct user engagement data, effectively "propagating" these insights to adjust the QMV of the target content.

[0147] In one embodiment of the invention, a method for refreshing quantitative market value (QMV) forecasts is disclosed. The refresh function ensures that the forecast incorporates the latest data, reflecting any recent changes in the characteristics of the market or similar content. When the forecast refresh function is executed, the model is re-executed to account for any changes in a set of similar content or updates.

[0148]

[0149] in, It is the refresh prediction value of content C at time t, which is combined with the refresh function.

[0150] It should be understood that the embodiments shown above and in the accompanying drawings represent only a few of many ways to implement embodiments for monitoring online activity to sort content in real time.

[0151] The environment of this invention embodiment may include any number of computers or other processing systems (e.g., client or end-user systems, server systems, etc.) and databases or other repositories arranged in any desired manner, wherein this invention embodiment can be applied to any desired type of computing environment (e.g., cloud computing, client-server, network computing, mainframe, standalone systems, etc.). The computers or other processing systems used in this invention embodiment can be implemented by any number of any personal or other type of computers or processing systems (e.g., desktops, laptops, PDAs, mobile devices, etc.), and may include any commercially available operating system and any combination of commercially available software and custom software (e.g., communication software, server software, content module 116, interface module or browser module 120, etc.). These systems may include any type of monitor and input device (e.g., keyboard, mouse, voice recognition, etc.) for inputting and / or viewing information.

[0152] It should be understood that the software of embodiments of the present invention (e.g., content module 116, interface or browser module 120, etc.) can be implemented in any desired computer language and can be developed by those skilled in the art of computer science based on the functional descriptions included in the specification and the flowcharts illustrated in the accompanying drawings. Furthermore, any reference herein to software performing various functions generally refers to the computer system or processor performing those functions under software control. The computer system of embodiments of the present invention can alternatively be implemented by any type of hardware and / or other processing circuitry.

[0153] The various functions of a computer or other processing system can be distributed in any manner across any number of software and / or hardware modules or units, processing systems or computer systems and / or circuits, wherein the computers or processing systems can be deployed locally or remotely to each other and communicate via any suitable communication medium (e.g., LAN, WAN, intranet, Internet, wired connection, modem connection, wireless, etc.). For example, the functions of embodiments of the present invention can be distributed in any manner across various end-user / client and server systems and / or any other intermediate processing devices. The software and / or algorithms described above and illustrated in the flowcharts can be modified in any way to implement the functions described herein. Furthermore, the functions in the flowcharts or specifications can be executed in any order to achieve the desired operations.

[0154] The software of this invention (e.g., content module 116, interface module or browser module 120, etc.) can be obtained on a non-transitory computer-usable medium (e.g., magnetic or optical medium, magneto-optical medium, floppy disk, CD-ROM, DVD, memory device, etc.) for use with a standalone system or a system connected via a network or other communication medium.

[0155] The communication network can be implemented by any number and type of communication network (e.g., LAN, WAN, Internet, intranet, VPN, etc.). The computer system or other processing system of this embodiment may include any conventional or other communication device to communicate over the network via any conventional or other protocol. The computer or other processing system can access the network using any type of connection (e.g., wired, wireless, etc.). The local communication medium can be implemented by any suitable communication medium (e.g., local area network (LAN), wired connection, wireless link, intranet, etc.).

[0156] The system can use any number of any conventional or other databases, data stores, or storage structures (e.g., files, databases, data structures, data, or other repositories) to store information. The database system can be implemented using any number of any conventional or other databases, data stores, or storage structures (e.g., files, databases, data structures, data, or other repositories) to store information. The database system can be included within or connected to server and / or client systems. The database system and / or storage structures can be deployed locally or remotely with computer systems or other processing systems and can store any required data.

[0157] Embodiments of the present invention can employ any number and type of user interface (e.g., graphical user interface (GUI), command line, prompt, etc.) to obtain or provide information, wherein the interface may include any information arranged in any manner. The interface may include any number and type of input or triggering mechanisms (e.g., buttons, icons, fields, boxes, links, etc.) located in any position to input / display information and initiate desired operations via any suitable input device (e.g., mouse, keyboard, etc.). The interface screen may include any suitable triggers (e.g., links, tabs, etc.) that navigate between screens in any manner.

[0158] Reports can include any information arranged in any way and can be configured based on rules or other criteria to provide users with the information they need (e.g., values, content items, historical data, etc.).

[0159] The embodiments of the present invention are not limited to the specific tasks or algorithms described above, but can be used to monitor user activities in real time and determine the value of any item or object.

[0160] Content items can be any type of digital or electronic item or object (e.g., document, webpage, file, data object, etc.) that contains data of any type or combination of types (e.g., text, multimedia, video, audio, images, streaming data, etc.). For example, content items can include news or other articles, websites or webpages, papers, documents, program code or applications, audio recordings, videos, images, live or recorded podcasts, streaming media, streaming of live events, blogs, messages, chats, conversations or other threads of topics, any combination thereof, etc.

[0161] The value of a content item can be any value within any desired numerical range or other range, and can represent any attribute of the content item (e.g., value, price, ranking, importance, relevance, etc.). The value curve of a content item can be any function, graph, graphical element, or other object (e.g., table, list, etc.) that expresses the value of the content item over time (e.g., a flat curve or straight line, any curve, any combination of curved and / or straight sections, a step function, or graph, etc.). Reference value curves can be generated in any way (e.g., randomized controlled trials (RCTs) or other experiments, randomly, pre-selected data, or default data, etc.). The value curve can be continuously updated at any desired time interval (e.g., seconds, minutes, etc.), and the values ​​can be continuously (in real-time) updated based on the value curve. The value curve can span any desired time interval (e.g., minutes, hours, days, weeks, months, years, etc.).

[0162] The weighting of value curves and features can use any weight within any desired value range and can be assigned based on any desired attribute or condition (e.g., user activity, user-assigned, etc.), and the weights can be determined by the algorithm and change as the model / algorithm is updated.

[0163] Feature vectors can include any number of dimensions or elements, each representing a feature of the content item. Features can include any number of features of any type (e.g., keywords, topics, events, word count, term frequency, word embeddings, term frequency-inverse document frequency (TF-IDF), etc.). Contextual information can include any attributes that provide context for the content item (e.g., location, brand popularity, audience demographics, purchasing power, brand, the event being described, content type, etc.).

[0164] Classification can be performed using any traditional or other machine learning model (e.g., mathematical / statistical; classifier; feedforward, deep learning, recurrent, convolutional, or other neural networks; unsupervised, supervised, or semi-supervised; etc.). Machine learning models can use unsupervised or supervised learning. Unsupervised machine learning uses data that is not labeled, classified, or categorized. For example, an unsupervised machine learning model (e.g., a neural network, etc.) can be trained using a training set of unlabeled data, where the neural network attempts to generate the provided data and adjusts the weight (and bias) values ​​using the error from the output (e.g., the difference between the input and the output). A supervised machine learning model (e.g., a neural network, etc.) can be trained using a training set that includes the input and known outputs, where the neural network attempts to generate the provided outputs and adjusts the weight (and bias) values ​​(e.g., via backpropagation or other training techniques) using the error from the output (e.g., the difference between the generated output and the known output).

[0165] The value curve can be updated using any conventional or other technique based on any metric representing the sensitivity of the activity to changes in value (e.g., scaling factor, gradient descent, stochastic gradient descent (SGD), adaptive gradient algorithm (AdaGrad), root mean square propagation (RMSProp), adaptive moment estimation (Adam), etc.).

[0166] Activities can include any online or other activities conducted by any entity in relation to a content item (e.g., clicks to access / initiate a transaction, cursor hover time, content item selection, ad views, etc.). Measurements or observations of the activities can include any information required (e.g., number of clicks to access / initiate a transaction, number of cursor hover times, number of content item selections, number of ad views, number of purchases, or number of other transactions, etc.).

[0167] Preferred embodiments of new and improved systems, methods, and computer program products for monitoring online activity and sorting content in real time have been described. Other modifications, variations, and alterations will likely occur to those skilled in the art in light of the teachings set forth herein. Therefore, it should be understood that all such variations, modifications, and alterations are considered to fall within the scope of the embodiments of the present invention.

Claims

1. A method comprising: receiving, by at least one processor, a content item containing content; determining, by the at least one processor, a value of the content item based on values of one or more content items associated with the content item; monitoring, by the at least one processor, online activity related to the content item; updating, by the at least one processor, the value of the content item in real-time as a function of user activity; and displaying, by the at least one processor, the value of the content item as a function of the real-time changes in the value.

2. The method of claim 1, further comprising: determining that the content item is a new content item; and extracting features from the content item upon determining that the content item is a new content item.

3. The method of claim 2, further comprising: identifying value curves of one or more content items associated with the new content item; and combining the identified value curves to produce a value curve for the new content item.

4. The method of claim 3, wherein the value curves are identified using one or more machine learning models. each value curve of a content item is classified by an associated class through an output layer neuron. a set of reference value curves is generated for an initial set of content items in operation, wherein the set of reference value curves are represented as a set of polynomial functions. the value curve of the new content item and the value of the new content item are continuously updated, wherein the value curve is represented as a polynomial function.

5. The method of claim 4, further comprising:

8. A method comprising:

6. The method of claim 5, further comprising: receiving, by at least one processor, a content item containing content; 7. The method of claim 6, further comprising: determining, by the at least one processor, a value of the content item based on values of one or more content items associated with the content item; monitoring, by the at least one processor, online activity related to the content item; updating, by the at least one processor, the value of the content item in real-time as a function of user activity; and displaying, by the at least one processor, the value of the content item as a function of the real-time changes in the value.

9. The method of claim 8, further comprising: determining that the content item is a new content item; and extracting features from the content item upon determining that the content item is a new content item.

10. The method of claim 9, further comprising: identifying value curves of one or more content items associated with the new content item; and combining the identified value curves to produce a value curve for the new content item.

11. The method of claim 10, wherein the value curves are identified using one or more machine learning models. each value curve of a content item is classified by an associated class through an output layer neuron. a set of reference value curves is generated for an initial set of content items in operation, wherein the set of reference value curves are represented as a set of polynomial functions. the value curve of the new content item and the value of the new content item are continuously updated, wherein the value curve is represented as a polynomial function.

15. A method comprising: receiving, by at least one processor, a content item containing content; ​ 12. The method of claim 11, further comprising: ​ 13. The method of claim 12, further comprising: ​ 14. The method of claim 13, further comprising: ​ ​ ​ determining, by the at least one processor, a value of the content item based on values of one or more content items associated with the content item; monitoring, by the at least one processor, online activity related to the content item; updating, by the at least one processor, the value of the content item in real-time based on user activity; and displaying, by the at least one processor, the value of the content item as it changes in real-time.

16. The method of claim 15, further comprising: determining that the content item is a new content item; and extracting features from the content item upon determining that the content item is a new content item.

17. The method of claim 16, further comprising: identifying value curves of one or more content items associated with the new content item; and combining the identified value curves to produce a value curve for the new content item.

18. The method of claim 17, wherein the value curves are identified using one or more machine learning models.

19. The method of claim 18, further comprising: associating each value curve of a content item with a class through an output layer neuron, each value curve of a content item being classified by the associated class.

20. The method of claim 19, further comprising: generating a set of reference value curves for an initial set of content items in operation, wherein the set of reference value curves are represented as a set of polynomial functions.

Images