Artificial intelligence-assisted automatic profiling method for recommending customised audiovisual content or displaying targeted advertising

Abstract

The present invention discloses a computer-implemented method adapted to run on a single computer or on a network of computers or any other programmable device that makes use of Automatic Speech Recognition (ASR) and Generative Artificial Intelligence (GAI) techniques for the automatic generation of transcripts of audiovisual content, to which it applies advanced Natural Language Processing (NLP) techniques to automatically and continuously create catalogue metadata and a detailed profile of the consumer, containing their individual tastes and preferences, with dynamic adjustment of the profile over time, thus enabling the recommendation of other contents or content segments, as well as the presentation of targeted advertising based on the characteristics of the profile. To this end, the method employs a set of algorithms that use the metadata generated and temporal information on content consumption in order to improve the accuracy of suggestions and customisation of advertisements, these being refined on the basis of the consumer's individual preferences, providing them with a more engaging and relevant viewing experience and, at the same time, boosting the assertiveness rate and effectiveness of advertising.

Description

[0001] DESCRIPTION

[0002] ARTIFICIAL INTELLIGENCE-ASSISTED AUTOMATIC PROFILING METHOD FOR RECOMMENDING PERSONALIZED AUDIOVISUAL CONTENT OR DISPLAYING TARGETED ADVERTISING

[0003] FIELD OF INVENTION

[0004] The present invention falls within the technical field of audiovisual content management. In particular, the present invention falls within the technical field of digital systems designed to profile consumers and recommend audiovisual content or contextualized advertising presentation.

[0005] BACKGROUND OF THE INVENTION

[0006] Technological developments in recent decades have profoundly transformed the way audiovisual content is consumed. The proliferation of streaming and video-on-demand (VOD) platforms has provided consumers with an unprecedented variety of entertainment options. However, this abundance of choices has brought with it several challenges for different players.

[0007] From the consumer's perspective, one of the main challenges has been the problem of discovering, selecting, and accessing content that is truly relevant to them, given the vast amount of content available in the catalog and the individual preferences and tastes of each consumer. For audiovisual entertainment service operators, it is increasingly important to monetize and deliver real value to consumers by providing features and User Experiences (UX) that make the discovery of content aligned with individual interests more efficient, enhancing consumers' perception of the value of the content provided while simultaneously distinguishing themselves from other market competitors.

[0008] On the other hand, advertisers and advertising service providers aim to improve the effectiveness and efficiency of converting advertisements into actual revenue for advertisers, thus increasing their profitability. To this end, it is crucial to identify with greater precision the target audience and the most appropriate time to display each advertisement.

[0009] The issue of discovering and choosing content suitable for each consumer in streaming and video-on-demand services has been addressed through traditional recommendation system solutions (using collaborative approaches, based on content, or hybrid approaches) and indexed content search, approaches that, in most cases, prove to be limited and extremely dependent on the quality and pre-existence of metadata for the content in the catalog.

[0010] Most traditional content recommendation algorithms use collaborative filtering, directly comparing past consumption patterns among consumers based on general content characteristics and metadata, but not on its content, nor on the content of segments or excerpts actually viewed, or on the frequency and number of views. Therefore, the results are often inaccurate or fail to capture the deeper essence of consumer preferences. Additionally, they struggle to address changes in individual consumer preferences over time.

[0011] On the other hand, the vast majority of search systems for audiovisual content catalogs are based on conventional approaches to indexing pre-existing metadata, using, among others, general information such as title, theme, genre, synopsis, cast and year of release. These methods, although they reasonably allow for finding content with some relevance, are not always able to meet individual consumer preferences, and therefore present considerable limitations.

[0012] One of the main limitations of this type of approach is the excessive reliance on the pre-existence and quality of content metadata, which may not be available or of sufficient quality, given that: 1) not all content providers offer complete or high-quality metadata with their content; 2) manually producing this metadata, or contracting third-party services to provide it, implies significant costs for the content provider or operator; 3) producing metadata for content in foreign languages ​​is sometimes difficult or even impossible and again implies costs; 4) live content broadcasting is not compatible with the manual pre-preparation of quality metadata, thus limiting the set of searchable content.Furthermore, it is common, in traditional research approaches, to present results of content in which the consumer is only partially interested, not allowing the identification of segments or excerpts of this content where a certain subject, topic or set of words are aligned with the filtering criteria or take into account the true tastes and preferences of the consumer, criteria that, from now on, we will call profile characteristics or simply characteristics.

[0013] Additionally, the most sophisticated current methods of content recommendation or targeted advertising leverage the existence of a consumer profile, their characteristics, and information on past consumption history and / or patterns. However, consumer profiling systems are limited to producing profile characteristics based on consumption events or pre-existing metadata, which, as mentioned above, are often incomplete or of poor quality. They neglect the importance of the actual content content and its segmentation according to its associated characteristics—segments that have been or are actually viewed by the consumer. Therefore, the quality of the profiling reflects these same problems and is only superficial.

[0014] This innovation aims to overcome these limitations by employing advanced Automatic Speech Recognition (ASR) and Generative Artificial Intelligence (GAI) techniques to generate, continuously (or on-demand) and automatically, detailed textual transcriptions of the audio of audiovisual content, regardless of the language spoken, on which a semantic analysis is performed to extract profile characteristics and to segment the content according to theme, subject or characteristic, with a view to complementing the pre-existing metadata of the content and enabling automatic, rich and time-segmented indexing.It also incorporates the processing of consumption, including viewing time, segments viewed and the frequency or number of views, with a view to creating automatic and highly detailed consumer profiling, also powered by Generative Artificial Intelligence (GAI), and with the ability to dynamically adjust over time.

[0015] It represents a significant advance compared to conventional methods, allowing a deep and accurate understanding of the content, whether transmitted via live streaming or on-demand, and of the consumer's current viewing context, a detailed knowledge of the consumer's individual tastes and preferences, thanks to the introduction of an innovative, more comprehensive and effective approach to making recommendations and researching audiovisual content or content segments or in the presentation of highly targeted advertising.

[0016] These advantages also aim to provide consumers with a more personalized and relevant user experience, thus representing a significant evolution in the field of discovery and appreciation of a catalog of audiovisual content.

[0017] SUMMARY OF THE INVENTION The present invention describes a computer-implemented method adapted to run on one or a network of computers or any other programmable devices that uses automatic profiling assisted by Generative Artificial Intelligence (GAI) to offer personalized recommendations, content search mechanisms or audiovisual content segments and presentation of targeted advertising, responding to the challenges described above inherent to their discovery, choice and access, ensuring that they are effectively relevant to the consumer because they are aligned with their individual tastes and preferences, are presented at the most opportune time according to the viewing context and also guaranteeing high effectiveness of targeted advertising.

[0018] The method employs Automatic Speech Recognition (ASR) techniques on the audio content to produce textual transcriptions and uses Generative Artificial Intelligence models to perform semantic analysis of the transcriptions to:

[0019] — identify characteristics that reflect the true content of the content and that can be used to index and conduct research on the content and be cross-referenced with the characteristics of the individual consumer profile;

[0020] — produce metadata for segmenting audiovisual content according to each identified characteristic, subject or inferred themes, through an aggregated and continuous analysis of the characteristics of each content.

[0021] Additionally, the method records and uses information regarding current or past views to determine the set of content or content segments that were actually viewed by consumers, and uses Generative Artificial Intelligence to create and execute analysis models on the segments or content viewed and the inferred profile characteristics to automatically and continuously compute a new, extremely detailed and personalized consumer profile.

[0022] Based on the consumer's dynamic profile, the method provides means to generate recommendations for audiovisual content or content segments and / or targeted advertising that:

[0023] — are relevant to the consumer, taking into account their tastes and preferences;

[0024] — take into account the viewing context, i.e. the profile characteristics of the content or segment of content that the consumer has seen or is seeing.

[0025] To this end, in its preferred embodiment, the method of the present invention comprises:

[0026] — at least one Automatic Speech Recognition (ASR) module;

[0027] — at least one module for recording, indexing and searching textual transcriptions;

[0028] — at least one processing module, comprising Natural Language Processing (NLP) and Generative Artificial Intelligence means;

[0029] — at least one module for recording information relating to a catalogue of audiovisual content; — at least one module for recording information and consumption history, containing metadata relating to current and past consumption;

[0030] — at least one module responsible for generating and processing profile features, comprising NLP and Generative Artificial Intelligence processing means;

[0031] — at least one consumer profile registration module;

[0032] — at least one profile features update module;

[0033] — at least one audiovisual content recommendation module;

[0034] — at least one targeted advertising module;

[0035] — at least one module containing an ad catalog.

[0036] DESCRIPTION OF FIGURES

[0037] The following description is based on the attached drawings which, without any limiting character, represent schematically the preferred embodiment of the invention.

[0038] Figure 1 illustrates a general conceptual scheme of the method, where the various modules that comprise it are represented.

[0039] Figure 2 illustrates a conceptual diagram of the continuous flow of metadata enrichment from the audiovisual content catalog, which uses the automatic generation of textual transcripts and the segmentation and generation of features from the transcripts. Figure 3 contains a schematic view of the profile feature generation flow, which uses consumption information and metadata from the profile catalog to extract context from viewed segments and update consumer profile features.

[0040] Figure 4 illustrates the preferred implementation of the on-demand flow of audiovisual content recommendations, where the characteristics of the consumer profile are crossed with the viewing context extracted from consumption information and catalog metadata to identify new content or content segments.

[0041] Figure 5 contains a conceptual diagram of the targeted advertising flow, in which third-party targeted advertising providers register ads and associate them with specific viewing contexts, which ads are then delivered to the consumer whenever their viewing context aligns with those associated with each ad.

[0042] Reference signs:

[0043] 1 - audiovisual content;

[0044] 2 - module for recording information relating to a catalogue of audiovisual content, responsible for storing, indexing and searching metadata and references to audiovisual content, as well as information on the segments that make up the content and associated metadata;

[0045] 3 - Automatic Speech Recognition (ASR) module responsible for processing the audio of audiovisual content and automatically generating the textual transcription and respective temporal information; 4 - module for recording, indexing and searching textual transcriptions of audiovisual content or segments;

[0046] 5 - processing module for identifying segments, enriching associated metadata and associating possible advertisements with segments or profile characteristics, using NLP and Generative Artificial Intelligence to perform a semantic analysis of textual transcriptions;

[0047] 6 - module for recording information on consumption history, containing metadata relating to current and past consumption of audiovisual content;

[0048] 7 - profile features generation and processing module, responsible for identifying and extracting unique features from the textual transcription of the content or segment of the audiovisual content that the consumer has watched or is watching and that materialize into features of the consumer profile;

[0049] 8 - consumer profile characteristics update module;

[0050] 9 - consumer profile registration module, responsible for storing, indexing and searching metadata containing profile characteristics;

[0051] 10 - audiovisual content recommendation module based on content consumption information and consumer profile characteristics;

[0052] 11 - targeted advertising module responsible for determining a viewing context based on the characteristics of the content and the consumer profile, for obtaining a set of advertisements according to the context, as well as for updating the advertisement catalog and subsequent updating of the content catalog with information on the advertisements available for each segment or characteristic, and for choosing advertisements to be presented to the consumer;

[0053] 12 - supplier of advertising, external to the digital system of the invention;

[0054] 13 - information recording module relating to an advertising catalogue;

[0055] 100 - general scheme of the invention of the invention;

[0056] 200 - consumer or third-party system that uses or interacts with the digital system of the invention.

[0057] DETAILED DESCRIPTION OF THE INVENTION

[0058] The most general advantageous embodiments of the present invention are described in the "Summary of the Invention" section. These embodiments are detailed below, in accordance with other advantageous and / or preferred modes of implementing the present invention.

[0059] The method of the present invention allows to circumvent the following limitations identified in the state of the art, namely:

[0060] — the difficulty of discovering, choosing and accessing audiovisual content or content segments that are truly relevant and of interest to the consumer, meeting their individual preferences and tastes;

[0061] — the need to enhance consumers' perception of the value of the content made available by the operator of an audiovisual entertainment service; — the ineffectiveness and inefficiency of converting advertisements into actual revenue for advertisers and advertising service providers;

[0062] — inaccuracy in identifying the target audience of advertisements;

[0063] — the difficulty of choosing the most appropriate time to present each advertisement, taking into account the viewing context of each consumer and the characteristics of the advertisement;

[0064] — the fact that current recommendation and targeted advertising systems depend on the quality and pre-existence of metadata for the content in the catalogue;

[0065] — the significant costs that may be involved in obtaining quality metadata;

[0066] — the difficulty in obtaining quality metadata for the various languages ​​in which the content is made available;

[0067] — and the limitations of current recommenders, as they are based primarily on general information (such as title, theme, genre, synopsis, cast, year of release, among others) of the content, to make comparisons of consumption patterns between consumers, without taking into account the real content of the content or content segments actually viewed and the frequency and number of views.

[0068] The system (100) enables the recommendation of new content or content segments and the presentation of targeted advertising contextualized over time, through automatic and continuous profiling of the consumer, using the analysis of the textual transcription of the content or content segments actually viewed by the consumer and their subsequent analysis with a view to constructing the consumer profile, as well as the analysis of the segments and extraction of the current viewing context for the presentation of targeted advertising.

[0069] Specifically, in the proposed system, illustrated in Figure 1 (100), there are four flows with distinct purposes that, together, contribute to the final objective of recommending personalized audiovisual content or content segments and delivering consumer-directed advertising.

[0070] The flow of automatic enrichment of metadata from the content catalog and generation of textual transcriptions, represented in Figure 2, is characterized by continuous analysis of both live and on-demand audiovisual content (1). In this flow, the Automatic Speech Recognition module (3) employs ASR and Generative Artificial Intelligence techniques to automatically generate a continuous flow containing the full textual transcription of the audio from the audiovisual content.

[0071] The transcription result is saved, together with its temporal information, in the transcription registration module (4) , to allow indexed searching of the transcription of any content or content segment in the catalogue (2) .

[0072] Subsequently, the module for processing and semantic analysis of textual transcriptions of audiovisual content (5) applies advanced Natural Language Processing (NLP) algorithms and techniques to the textual transcriptions and temporal metadata (4), with the aim of identifying and generating a flow of metadata containing the most relevant standardized characteristics (regardless of the language of the content under analysis) of the analyzed text, duly indexed to the time of the content, including, subject or theme portrayed, keywords, topics, product brands, objects referred to, personalities, feelings, preferences, beliefs or consumer orientations (200), including philosophical, artistic, food, cultural, technological, political, religious, sexual or sporting inclinations.

[0073] The module for processing and semantic analysis of textual transcriptions of audiovisual content (5) uses the result of the previous analysis to also generate temporal metadata and segmentation information relating to possible segmentations of the audiovisual content (1) taking into account the different dimensions and characteristics identified, in order to automatically enrich the content metadata in the catalogue (2) in great detail, with these being reused by the remaining flows of the digital system (100) and common to all consumers (200).

[0074] The flow for generating consumer profile characteristics, represented in Figure 3, includes the set of steps that allow generating the consumer profile (9) based on the characteristics and content segments (1) actually viewed. For each content or content segment (1) consumed by the consumer (200), consumption information is recorded in the information and consumption history recording module (6).

[0075] This information concerns the period of time that the consumer (200) actually viewed audiovisual content (1), as well as the number and frequency of viewings thereof. In contrast to traditional profiling systems, the present invention represents an advantage, since it provides for the discarding of content segments that do not arouse interest in the consumer (200), valuing only those for which real interest has been expressed, which allows profiling to be much more precise and aligned with the tastes and preferences of the consumer (200).

[0076] For each new consumption recorded in the information and consumption history recording module (6), the characteristics processing and generation module (7) cross-references the indexed metadata of the catalogue (2) with the consumption information and history (6), using NLP and Generative Artificial Intelligence processing methods to generate the characteristics of the segments actually viewed, which are then communicated to the consumer profile characteristics update module (8).

[0077] Based on the characteristics received from the content or content segment that the consumer (200) is viewing or has already viewed and on the query of the current consumer profile information (200) stored in the profile registration module (9) , the consumer profile characteristics update module (8) employs NLP algorithms and models and Generative Artificial Intelligence to generate a new consumer profile.

[0078] The new consumer profile results from the joint analysis of the characteristics extracted from the periods viewed (which may cross one or more content segments), with the characteristics of the current consumer profile, taking into account several factors, such as the number of times and frequency with which the consumer watches the content or content segments, as well as the common characteristics or their hierarchical relationships.

[0079] From this fine-grained analysis, the consumer profile is generated, which is subsequently communicated and saved in the consumer profile registration module (9) for future reuse in other flows. The consumer profile (200) is constantly updated as more consumption information is recorded (6).

[0080] In the content or content segment recommendation flow, represented in Figure 4, the data generated in the automatic metadata enrichment flow (Figure 2) and in the profile characteristics generation flow (Figure 3) are reused to generate recommendations appropriately contextualized to the consumer profile (9), the time and context of viewing.

[0081] The flow can be triggered in two ways: 1) by the consumer's own initiative (200), through the request for recommendations; 2) by determination of the consumption information registration module (6) which notifies the recommendations module (10) that a new viewing context has appeared.

[0082] In either case, the content recommendations module (10) consults the consumer profile registered in the profile registration module (9) in order to obtain the current list of profile features that were generated in the profile features generation flow (Figure 3).

[0083] The content recommendation module (10) then performs a search in the catalog module (2) for audiovisual content or content segments (1) whose segmentation and characteristics resulting from the automatic metadata enrichment flow (Figure 2) match the characteristics of the profile.

[0084] In an alternative implementation of the method, the search takes into account other criteria, whether or not explicitly indicated by the consumer, which limit the context of the recommendation (for example, to limit the results of the content recommendation to only the comedy genre in which the consumer is interested). The search result thus corresponds to recommendations aligned with the interests, tastes and preferences inferred for the consumer (200) and / or other criteria, and may contain a plurality of content recommendations or content segments.

[0085] Subsequently, the recommendations module (10) applies sorting and filtering criteria to the search results, taking into account the consumer's consumption history (200), with a view to prioritizing unseen content, avoiding repetitions and enhancing the content in the catalog (2).

[0086] Furthermore, there is a need to ensure that recommendations are contextualized in time, that is, based on profile analysis (9) the system records consumption habits in temporal terms and, with this information, emphasizes recommendations that suit not only the profile, but also the consumer's habits (200) and the moment of consumption.

[0087] To this end, the recommendations module (10) consults the consumption history information registration module (6) and excludes recommendations whose number of views, frequency or viewing height do not comply with the defined business rules. On the other hand, the same consumption history metadata can, in accordance with the defined business rules, be used to reorder the remaining recommendations, recommendations that are sent to the consumer (200).

[0088] The presentation of targeted advertising, the flow of which is represented in Figure 5, is triggered by consumer consumption or actions (200), allowing the method to determine the target audience, the best moment and which advertisement(s) are appropriate to the viewing context.

[0089] To make the method scalable, since it works continuously by performing a semantic analysis of the automatic transcription of all content and content segments available in the catalog (live streaming and on-demand), as well as to guarantee consumer confidentiality, the advertisements available on its platforms are recorded in the advertisement catalog module (13), by the advertising providers (12) or by a method adapted for this purpose, also indicating the most appropriate viewing context for each advertisement.

[0090] The targeted advertising presentation flow is divided into three phases:

[0091] — phase 1: ad provisioning;

[0092] — phase 2: associating features, content and content segments with ads;

[0093] — phase 3: request for advertisements or suggestion, initiated by the system, selection and presentation of advertisements.

[0094] In a first phase, which always occurs if it is intended to promote, change or delete an advertisement in the system (100), a second method adapted for this purpose or an advertising supplier (12) begins by registering, updating or removing the advertisement by sending an operational command to the targeted advertising module (11) containing metadata about the advertisement, including the external reference of the advertisement and the presentation context indicating a set of characteristics, which may or may not be hierarchical.

[0095] The advertising module (11) , in turn, normalizes these characteristics into internal profile characteristics used by the system (100) and proceeds to update, provision or de-provision the advertisement by sending an operational command to the advertisement catalog module (13) .

[0096] Furthermore, the advertising module (11) queries and updates all content, content segments and / or features indexed in the content catalog module (2) to ensure synchronization of the information signaling the existence of potential advertisements for each of these indexing dimensions.

[0097] The correspondence between the content catalogue (2) and the advertisement catalogue (13) is therefore completely autonomous from the consumer and can occur asynchronously, being triggered whenever new advertisements are added or removed to the advertisement catalogue or whenever new content is added to the content catalogue.

[0098] The responsibility for determining the appropriate time, the correct viewing context and the target audience with the appropriate profile for displaying the advertisements lies with the digital system of the invention (100). However, the selection of the advertisement can be shared between the system and the advertising provider (12), as explained later in the third phase of the targeted advertising flow.

[0099] In a second phase, and continuously, the system reuses the metadata of the content or content segment generated in the automatic metadata enrichment flow (Figure 2), in particular the characteristics, temporal information and content segmentation, to frame advertisements with the potential to be presented during the viewing of the content or content segment, which may or may not be presented according to the consumer profile (9) and the decision of the advertising provider (12).

[0100] This process is executed by the content processing module for segment identification (5) during the automatic metadata enrichment flow (Figure 2) by querying, based on each of the standardized characteristics of the content or segment under analysis, the targeted advertising module (11) which, in turn, queries the advertisement catalog module (13) to obtain metadata from previously registered advertisements, whose presentation context coincides with the characteristic or with any hierarchically superior characteristic.

[0101] If at least one advertisement has been found for a given characteristic, segment or content, the content processing module for segment identification (5) sends an operational command to the content catalog module (2) to signal and record that there is at least one advertisement with the potential to be displayed for the characteristic or segment viewed by the consumer (200). In a third phase, in which advertisements are presented to the consumer (200), the process can be triggered in two ways: — by an explicit request from the consumer (200) for advertisements;

[0102] — by automatic suggestion from the advertising module (11), taking into account the viewing context to determine the optimal time for displaying the advertisement.

[0103] In the first case, the consumer (100) sends an operational command to the advertising module (11) which consults the viewing context through the profile characteristics generation and processing module (7) which, among other metadata, indicates the reference of the content or content segment and the respective characteristics.

[0104] Based on the content or content segment reference, the advertising module (11) consults the content catalog module (2) to obtain information regarding the signaling of the existence of advertisements for the content, content segment or any of its characteristics. If the signaling of the existence of advertisements does not occur, no advertisement will be delivered to the consumer (200) by the targeted advertising module (11) •

[0105] If a signal is found indicating the existence of a potential advertisement, the targeted advertising module (11) consults the advertisement catalog (13) to obtain additional information about the advertisements, such as external references belonging to the advertisement provider itself or to the advertisers (12), as well as the standardized characteristics corresponding to the context in which the advertisement should be displayed.

[0106] Next, the targeted advertising module (11) queries the profile module (9) to obtain the consumer's profile characteristics and then compares them with the characteristics of the ad viewing context in order to find a match between them, also taking into account the hierarchical relationships of the characteristics. This match indicates that the consumer (200) has the appropriate profile for the ad display and is part of the target audience of the advertising campaign. Furthermore, since the consumer's current viewing context has already been validated, it is guaranteed that the timing is opportune.

[0107] Once a match has been found, the targeted advertising module (11) sends, via an operating command, the display context of the advertisement or advertisements found in the catalogue, as well as a pseudonymous identifier of the consumer and optionally an external reference of the advertisement to the advertising provider (12).

[0108] The advertising provider (12) processes the operational command to find, internally or through other advertisers with which it may have a relationship, suitable advertisements, as well as to implement capping mechanisms, among others, which are standard in an advertising system. Finally, the advertising provider (12) sends the advertisements and control metadata (such as impression tracking, progress, error, among others) to the targeted advertising module (11). In turn, the targeted advertising module (11) delivers the advertisement or advertisements to the consumer (200) and ensures that the control mechanisms required by the advertising provider (12) are met.Alternatively, that is, when the advertisement does not come from a stimulus from the consumer (200), it is the result of a suggestion from the digital system (100), with the flow being very similar to the first case, except that the process that triggers it is executed by the consumption history information recording module (6) as a result of consumption made by the consumer (200).

[0109] That is, the consumer (200), during his / her viewing experience, when starting to play a new content or content segment (1), results in an operational command being sent to the consumption history information recording module (6). This, in turn, notifies the profile characteristics generation and processing module (7) to start analyzing and extracting the current viewing context.

[0110] Based on the characteristics of the content or content segment, the profile characteristics generation and processing module (7) queries the content catalog module (2) to obtain information regarding the signaling of advertisements for any of the identified characteristics or their hierarchical relationships and, if a signaling of the existence of advertisements for any of the characteristics is found, the profile characteristics generation and processing module (7) sends an operational command to the targeted advertising module (11) so that it begins the search for matches with the consumer profile (200). The rest of the process is, in all respects, identical to the first case.

[0111] The advertisements that are eventually displayed are the result of the execution of two processes carried out alternatively, one being executed in an on-demand logic and the second in a continuous and active manner.

[0112] Thus, in the preferred implementation of the method (100) of automatic profiling assisted by artificial intelligence for recommending personalized audiovisual content or displaying targeted advertising, this comprises:

[0113] — at least one module containing a catalogue of audiovisual content available for viewing and its metadata (2);

[0114] — at least one module employing Automatic Speech Recognition (ASR) algorithms to automatically generate a continuous stream containing the textual transcription of the audio of the audiovisual content and the respective temporal information indexed to the content (3);

[0115] — at least one module for recording, indexing and searching textual transcriptions (4);

[0116] — at least one processing module that employs semantic analysis algorithms of transcripts aided by Artificial Intelligence to generate standardized characteristics, topics of interest and content segmentation metadata according to these dimensions, with a view to enriching and / or complementing the metadata in the catalog, and also for interaction with the targeted advertising module to signal content or content segments for which there is potential for advertisements to be displayed (5); — a module for recording historical consumption information, both current and past, with which the consumer (or any digital device or system that the consumer may use) interacts to record views of content or content segments (6);

[0117] — at least one module responsible for generating and processing profile features of the audiovisual content or segments of the audiovisual content that are actually viewed by the consumer, comprising NLP and Generative Artificial Intelligence processing means (7);

[0118] — at least one consumer profile recording module, comprising means for storing, indexing and searching any metadata relating to the individual consumer's tastes and preferences, with the ability to dynamically update profiles (9);

[0119] — at least one profile features update module comprising means for processing a plurality of pre-existing profile features and those relating to the content or content segment to compute a new plurality of consumer profile features (8);

[0120] — at least one audiovisual content recommendation module, which employs a method for processing and recommending audiovisual content or content segments from the catalogue in accordance with the consumer profile, the consumer's consumption information and history and other filtering and context criteria indicated by the consumer (10);

[0121] — at least one targeted advertising module employing means for processing the characteristics of the content or segment of audiovisual content that the consumer is watching, the indication of the existence of advertisements and the characteristics of the consumer profile to determine a viewing context and send them to the system of the external advertising provider to obtain at least one contextual advertising advertisement to be delivered to the consumer (11);

[0122] — at least one module containing an advertisement catalogue, responsible for recording, indexing and searching for associations between advertisement references and profile characteristics (13);

[0123] In an alternative implementation of the proposed system, it comprises means for interconnection to an external recommendation system (10), to which it provides, among other metadata, a pseudo-anonymous identifier of the consumer and / or consumption session, as well as their profile characteristics and viewing context.

[0124] In an alternative embodiment of the proposed system, the advertising provider system (12), with which the targeted advertising module interacts (11), ceases to be an external system and is fully or partially integrated into the digital system of the present invention.

[0125] In an alternative implementation of the proposed system, the consumer (200) uses a human-machine interface, through which he interacts with the system to consume audiovisual content or content segments and where the results of the recommendations and targeted advertising are presented.

[0126] In an alternative implementation of the proposed system, the Automatic Speech Recognition (ASR) module for generating transcriptions (3), the processing module that employs semantic analysis algorithms for transcriptions (5) and the module responsible for generating and processing profile characteristics of the displayed audiovisual content or content segments (7), have means for the ASR and NLP functions assisted by Generative Artificial Intelligence to be executed externally, or be cloud-based.

[0127] In an alternative implementation of the proposed system, the textual transcription registration module (4) is not present and where the metadata resulting from the ASR module is sent through a continuous flow of information directly to the semantic analysis and feature generation module.

[0128] As will be clear to a person skilled in the art, the present invention should not be limited to the embodiments described in this document, and various modifications are possible that remain within the scope of the present invention.

[0129] Evidently, the preferential modes presented above can be combined in different ways, thus avoiding the repetition of all these combinations.

Claims

CLAIMS 1. Method of automatic consumer profiling and automatic targeting of addressed advertising characterized by comprising: — recording of user consumption — textual transcription of audiovisual content and recording of the respective temporal information; — semantic segmentation of textual and temporal transcription using NLP algorithms — extraction of features related to each textual and temporal segment using IAG — crossing the technical characteristics of the audiovisual content with the consumption history in order to identify user profiles — comparison of the characteristics of a profile with a pu catalogue 2. Automatic consumer profiling system and automatic targeting of addressed advertising characterized by comprising: — at least one Automatic Speech Recognition (ASR) module, adapted for processing the audio of audiovisual content and automatically generating the textual transcription and respective temporal information; — at least one module for recording, indexing and searching textual transcriptions generated by the Automatic Speech Recognition module, their temporal information and reference to audiovisual content; — at least one processing module, comprising Natural Language Processing (NLP) and Generative Artificial Intelligence means for carrying out semantic analysis of textual transcriptions of audiovisual content, with a view to generating relevant metadata and duly indexed to the time of the content that allows enriching and / or complementing pre-existing metadata, resulting in the identification of characteristics of the transcriptions and a temporal segmentation of these characteristics; — at least one module for recording information relating to a catalogue of audiovisual content, with: o the capacity to store, index and search metadata and references to audiovisual content or segments of content; o the capacity to store, index and search metadata containing references to advertisements and their relationship with the audiovisual content or any of its segments; — at least one module for recording information and consumption history, containing metadata relating to current and past consumption of audiovisual content by the consumer, including temporal information on segments viewed and the number of views made; — at least one module responsible for generating and processing profile features, comprising NLP and Generative Artificial Intelligence processing means for: o query or react to a stimulus from the consumption history recording module to obtain references and temporal information of the contents or segments of the audiovisual contents that are actually viewed by the consumer; o query, based on the references and temporal information of the contents or any of their segments, the information recording module relating to a catalogue of audiovisual contents to obtain all the characteristics of the textual transcriptions for the intervals and contents viewed; o identify and generate unique characteristics of the textual transcription that materialize in characteristics of the consumer profile and send them to the consumer profile update module; o coordinate communication between the modules for recording catalogue information, recording consumption history and updating the characteristics of the consumer profile; — at least one consumer profile registration module, comprising means for storing, indexing and searching any metadata relating to the individual tastes and preferences of the consumer, with the capacity for dynamically updating the profiles; — at least one profile feature update module comprising means for: querying or reacting to a stimulus from the profile feature generation and processing module to obtain a plurality of characteristics relating to the content or segment of audiovisual content; query the consumer profile registration module to obtain a plurality of the current characteristics of the consumer profile; process the plurality of pre-existing profile characteristics relating to the content or segment of content to compute a new plurality of characteristics of the consumer profile, in order to reflect the tendency and inclination of their tastes and preferences; communicate to the profile characteristics registration module the new plurality of computed characteristics.

3. System, according to claim 1, characterized in that the Natural Language Processing (NLP) and Generative Artificial Intelligence module also comprises means for inquiring or reacting to a stimulus from a targeted advertising module to obtain information regarding the existence of advertisements, relevant and appropriate to the consumption context, which share at least one of the characteristics identified in the transcription of the content or in the audiovisual segment, or any hierarchically superior characteristic, as well as for sending operational commands to the catalog module to record the classification of the segments or contents regarding the existence of advertisements.

4. System according to claims 1 or 2, for processing, generating textual transcripts, their analysis and generating profile features can be applied to both previously recorded audiovisual content or live broadcast content.

5. System, according to claims 1, 2 or 3, characterized in that the processing, generation of textual transcripts, their analysis and generation of profile characteristics can be carried out on segments or excerpts of previously recorded or live broadcast audiovisual content.

6. System according to any one of the preceding claims, characterized in that it comprises at least one audiovisual content recommendation module, which employs a method for: — Processing and recommendation of audiovisual content or segments of audiovisual content according to: o the consumer’s personal tastes and preferences, stored in their profile and which correspond to profile characteristics inferred by NLP and Generative Artificial Intelligence processing on the textual transcriptions of the segments actually viewed; o the audiovisual content in the catalogue, previously indexed according to the characteristics identified by the processing and semantic analysis module of the textual transcriptions of the audio of the audiovisual content; o the information and history of consumption already made or in progress by the consumer; o other filtering and context criteria indicated by the consumer. — Coordination of communication between the catalogue information registration, consumption information and history registration and consumer profile registration modules; — Sending an operational command to a third-party system, containing a plurality of recommendations for audiovisual content or content segments, the third-party system may or may not contain a human-machine interface for presenting the results to the consumer.

7. System, according to claim 5, characterized by recommending audiovisual content or segments of content in a catalog, whose metadata includes characteristics identical or similar to those found in the consumer's profile and / or those of the audiovisual content that the consumer is watching.

8. System, according to claims 4, 5 or 6, characterized by giving emphasis to the specific moment or segment of the content that the consumer is watching to recommend new segments or excerpts of audiovisual content whose metadata includes characteristics identical or similar to those found in the consumer's profile and / or to those of the content or segment of the audiovisual content that the consumer is watching.

9. System according to claims 1, 2, 3 or 4, characterized in that it comprises at least one targeted advertising module that employs means for: — inquire or react to a stimulus from the profile characteristics generation and processing module, to obtain metadata of the content or segment of audiovisual content that the consumer is watching, where includes temporal information, reference to the content in the catalogue, a plurality of characteristics of the textual transcription associated with it and the indication of the existence of advertisements by characteristic; — obtain the characteristics of the consumer profile by consulting the metadata stored in the consumer profile record; — process the metadata containing the characteristics of the consumer profile, the catalogue metadata containing the characteristics of the textual transcription of the audio of the content or segment of audiovisual content viewed, to determine a viewing context: one consisting of a plurality of common characteristics between profile and content or segment of content; one indicating the existence of at least one advertisement available. — send the viewing context to at least one system of the external advertising provider to obtain at least one contextual advertising advertisement to be delivered to the consumer; — coordinate communication between the modules for generating and processing profile characteristics, the profile registration module, the catalogue registration module and the external advertising provider's system.

10. System, according to claim 8, characterized by comprising means for: — allow the external advertising provider's system to indicate the existence of advertisements and the viewing context in which they should preferably be presented to the consumer; — allow the external advertising provider's system to change or delete a previous indication of the existence of advertisements; — map textual features present in the catalog with the viewing contexts indicated by the external advertising provider; — send an operational command to the content catalogue registration module to register the existence of at least one advertisement for at least one textual feature; — send an operational command to the content catalogue registration module to change or delete a previous record of the existence of advertisements for at least one textual feature.

11. System, according to claim 9, characterized by comprising, at least, one module containing an advertisement catalog, responsible for registering, indexing and searching for associations between advertisement references provided by the external advertising provider and viewing contexts in which the advertisements should preferably be displayed, wherein a viewing context is determined by a plurality of profile characteristics.

12. System, according to claim 9, characterized by comprising means for: — allow the reception of an operational command from the semantic analysis processing module of textual transcriptions so that, on a continuous basis, the existence of registered advertisements is assessed for each textual characteristic found in the content or content segment; — query the advertisement catalogue module to determine whether there are advertisements that should be displayed in a given viewing context or profile characteristic; — send an operational command to the content catalogue module to, across all content or content segments and by characteristic, record the existence or non-existence of at least one advertisement.

13. System according to claims 8, 9 or 10, characterized by: — understand means to query the registration and association module between advertisements and characteristics and thus obtain the list of advertisement references corresponding to the list of characteristics of the viewing context; — incorporate a confidentiality mechanism so as to avoid sending the viewing context to the external advertising provider's system, sending instead a plurality of external advertising references.

14. System according to claims 8, 9, 10, 11 or 12, wherein the module responsible for generating and processing profile characteristics, based on consumption information, comprises means for sending an operational command containing the identified characteristics to the targeted advertising module.

15. System according to claim 8, 9, 10, 11, 12 or 13, characterized in that the targeted advertising module emphasizes the specific moment or segment of content that the consumer is watching, as well as its characteristics, for better contextualization of advertisements based on the products or experiences that the consumer is observing.

16. System according to claims 5, 6, or 7 characterized by including means for interconnection to an external recommendation system, to which it provides: — pseudo-anonymous identifier of the consumer and / or consumption session; — the characteristics of the consumer profile; — the characteristics of the audiovisual content being viewed; — the characteristics of the segment or excerpt that the consumer is watching; — metadata of audiovisual content in the catalogue.

17. System, according to claims 5, 6, 7 or 15, characterized in that the recommendations module comprises means capable of filtering or reordering recommendations according to criteria extracted from the consumer's consumption history, being able to: — exclude or include audiovisual content already viewed by the consumer; — give greater or lesser prominence by reordering the results, moving content not yet viewed to the top or bottom of the list of recommendations.

18. System according to claims 8, 9, 10, 11, 12 or 13 characterized by fully or partially integrating the advertising supplier's system.

19. System according to any one of the preceding claims, characterized in that it can integrate a human-machine interface for visual, auditory or textual feedback.

20. System, according to any of the previous claims, characterized by having, simultaneously, an audiovisual content recommendation module and a targeted advertising module, both based on a consumer profiling model and descriptive metadata of the content inferred and generated from automatic textual transcriptions.

21. System according to any one of the preceding claims, characterized in that the profile registration module uses and allows the storage, search and indexing of profile information explicitly provided by the consumer, in addition to the inferred information.

22. System, according to any one of the previous claims, characterized by the Automatic Speech Recognition (ASR) module for generating the transcriptions, the processing module that employs algorithms for semantic analysis of the transcriptions, and the module responsible for generation and processing of profile characteristics of the contents or segments of the audiovisual contents viewed, have means so that the ASR functions, NLP assisted by Generative Artificial Intelligence can be executed externally, or be cloud-based.

23. System according to any one of the preceding claims, characterized in that the contents viewed may relate to broadcasts listed in an Electronic Program Guide (EPG), live or deferred, or to an on-demand catalog.

24. System according to any one of the preceding claims, wherein the textual transcription recording module is not present and wherein the metadata resulting from the ASR module is sent via a continuous flow of information directly to the semantic analysis and profile feature extraction module.

25. System, according to any of the previous claims, wherein the NLP and Generative Artificial Intelligence processing module responsible for the semantic analysis of the textual transcripts, extraction of features, and segmentation of the content, has means to, regardless of the spoken language of the content, perform an automatic translation and normalization of its features.

Images