System for the automatic generation of esports highlights using contextual machine learning models

Abstract

A system (100) for automatically generating esports highlights using contextual machine learning models, the system comprising: a) a data ingestion module configured to capture and synchronise multiple data streams including game telemetry, video footage, audio signals and external viewer metadata from live or recorded esports matches; (b) an event detection module configured to detect significant game events by applying rule-based algorithms and machine learning classifiers to the ingested data; (c) a contextual analysis module configured to analyze the significance of the detected events by assessing the match context, emotional impact, player status, team dynamics and external audience reactions using contextual machine learning models; (d) a highlight scoring module configured to assign points to events based on their contextual significance and predicted audience engagement value; (e) an editing and compilation module configured to select, enhance, and compile the highest-scoring events into a coherent highlight reel using predefined or dynamically generated templates; and f) optionally, a user personalisation module configured to adapt highlight selection and stitching based on user preferences, including preferred players, teams or playing styles.

Description

The present invention relates generally to automated generation of video content, and more particularly to a system for automatically generating esport highlights using contextual machine learning models.The rapid ascent of the report has led to an increasing demand for highlight reels showing spanning moments from competition games. Traditionally, the process of creating such highlights is manual, labor intensive, and subjective, as the editors must see through extensive gameplay material to select the best moments.Existing automated systems for creating highlights often rely on simple event triggers (such as kills or sieges) without understanding the entire context of the match, such as the player's status, strategic importance, viewers' responses, or the match's dynamics. As a result, many highlight reels leave critical, shaded moments that the human audience estimates.There is therefore a need for a system that enables context-aware, automated highlight generation that captures not only event-based actions, but also strategic and emotional importance of moments in export matches.An object of the present disclosure is to automate the entire process of generating export highlights, thereby reducing manual labor.Another object of the present disclosure is to improve quality by using contextual models to capture real-sensing moments.Another object of the present disclosure is to operate in real-time and enable live highlights to be generated during running games.Another object of the present disclosure is to support personalization so that the highlights can be tailored to the user's preferences.Another object of the present disclosure is to process multimodal data combining gaming history, audio and viewer responses for more comprehensive analysis.Another object of the present disclosure is to adapt to various games that can be easily configured for various estport titles and genres.Another object of the present disclosure is to increase viewer's engagement by producing emotionally appealing and strategic highlights.Another object of the present disclosure is to provide continuous improvement through self-learning based on user feedback and engagement metrics.The present invention relates to a system for automatically generating Excessives in the Esport using contextual machine learning models. The system captures and processes multiple data streams including live or recorded game telemetry, video material, audio signals, and external metadata such as viewer responses. By a combination of rule-based methods and machine learning classifiers, it automatically recognizes important events in the game. These events are then analyzed by contextual models to evaluate their strategic and emotional significance within the game. Based on this analysis, a highlight scoring engine assigns a probability score to each event so that the system can select the most spanning moments and compile them into a captive highlight reel. The system evaluates the highlight points with optional filming effects and provides the user with personalization functions by which he can tune the output to specific players, teams or games styles. The invention has been developed for both live and post-processing applications and significantly improves the efficiency, quality and relevance of the production of esport highlights compared to conventional manual or purely event based methods.The present invention relates to a system for automatically generating easternly-via-the-sea using contextual models of machine learning, with which the most exciting and contextually significant moments of esport matches are automatically recognized, analyzed and compiled to be safened highlight reels. The system overcomes the limitations of traditional highlight generation, which relies heavily on manual processing or simple event triggers by introducing advanced contextual analysis using machine learning.The components of the system (100) are as followsThe heart of the system is the data-augmentation module, which is responsible for acquiring various types of input data in real time or from recorded sources. These streams include telemetry data from the game (e.g., kills, death, assistants, targets reached, and resource usage), video material from the game event, audio signals (including game sounds, comments, and audience sounds), and external metadata (e.g., chat responses or moods in social media). By collecting various multimodal data, the system provides a comprehensive understanding of the running gaming environment.After data capture, the event detection module identifies important gaming events that could potentially serve as highlight moments. This module uses a hybrid approach that combines rule-based algorithms (for basic triggers such as multikills or final beats) with machine learning classifiers trained to recognize more complex events such as clutch plays, turnarounds, or strategic maneuvers. The module ensures that a critical event is not missed, nor those that may not be immediately recognizable to conventional systems.Detected events are then fed into the context analysis module, which evaluates the strategic and emotional significance of each event. Unlike conventional systems which handle all events equally, this module uses contextual models such as deep neural networks and transformer-based architectures to account for factors such as current score, game history, player health, resource usage, historical player / team performances, and external viewer responses. From the analysis, the system can determine whether an event was only routine or whether it was truly game changing, exciting or emotionally contributing.The results of context analysis are processed by the highlight scoring module, which assigns a highlight likelihood score to each detected event and associated time frame. The scoring model was trained on a large dataset of annotated estport film material to learn from real examples which moments in the past were perceived as peaks. Events are ranked based on their score, and dynamic thresholds provide only the most severe moments to be selected, taking into account the length and style requirements of the final highlight reels.Selected high score segments are passed to the edit and compile module where they are cropped, enhanced, and compiled into a filed video. The module may apply optional movie improvements such as slow motion repetitions, critical moment magnification effects, audio emphasis, and seamless transitions. Prefabricated templates can be used to produce various types of highlight reels, such as "Top 5 Plains", "Epic Combuabs", or "Player of the Match.".An optional user personalization engine further refines the system by allowing the users to specify preferences regarding favourites, teams, game styles, or event types. By adjusting highlight selection to these personalized settings, the system can create customized movie sequences for particular target groups or individuals, thus increasing viewer satisfaction and engagement.The system supports two modes of operation: a live mode that generates highlights in near real time during running games, and a post-processing mode that analyzes completed images for highlight generation. The system is designed to be adaptable to multiple export titles by reconfiguring the event detectors and retraining the models for game specific contexts.The invention is explained again below with reference to the figure. Illustrated here: FIG. 1 shows a system ( 100) for the automated generation of highlights in the esport using context-related machine learning model.FIG. 1 shows a system ( 100) for automatic emphasis generation in the esport with context-dependent machine learning models. The system operates by first capturing multiple data streams, including live or recorded game telemetry, video material, and audio signals, via the data congestion module. The event detection module continuously monitors the game data to identify important events in the game, such as kills, targets, or important strategic trains, using a combination of rule-based algorithms and machine learning classifiers. The detected events are then passed to the context analysis module, which evaluates the general meaning of each event by analyzing the game status, player conditions, team dynamics, viewers' responses, and other situational factors using contextual models. These findings are introduced into the highlight scoring module, which assigns each event a numerical score based on the predicted excitation, meaning, and relevance to the audience. Segments with a score above a dynamic threshold are selected and passed to the processing and assembly module where they are tailored, optionally enhanced with filming effects, and assembled into a sophisticated highlight reel. In addition, a user personalization module can adapt the selection of the highlight points to the preferences of the viewers, e.g., to the preference for a particular player or style of game, and thus enable the generation of individual highlight points. The system can be used in real time during live transmissions or as a re-processing tool for recorded material.

Claims

A system (100) for automatically generating eastern eastern eastern using contextual machine learning models, the system comprising: a) a data acquisition module configured to acquire and synchronize multiple data streams including game telemetry, video material, audio signals, and external viewer metadata from live or recorded eastern matches; b) an event detection module configured to detect important game events by applying rule-based algorithms and machine learning classifiers to the acquired data; c) a contextual analysis module configured to analyze the meaning of the detected events by evaluating the game context, emotional impact, player status, team dynamics, and external viewer responses using contextual machine learning models; d) a highlight scoring module configured to assign points to events based on their contextual meaning and predicted audience binding value; e) a editing and compilation module configured to select, enhance, and merge the events with the highest scores into a coherent highlight role using predefined or dynamically generated templates; and f) optionally a user personalization module configured to adapt highlight selection and merging based on user preferences, including preferred players, teams, or game styles.The system (100) of claim 1, wherein the data input module synchronizes the game telemetry to corresponding video and audio streams in real time.The system (100) of claim 1, wherein the event detection module combines game-specific rule sets and supervised machine learning models trained on marked game data.The system (100) of claim 1, wherein the contextual analysis module uses transformer-based deep learning models to determine strategic and emotional weight of gaming events.The system (100) of claim 1, wherein the highlight scoring module dynamically adjusts the event scoring thresholds based on the game history such that events occurring during critical phases are given a weighted meaning.The system (100) of claim 1, wherein the editing and compilation module applies video effects including slow motion, zooming, and repetitive loops to enhance the presentation of highlight segments.The system (100) of claim 1, wherein the user personalization module modifies the composition of the highlight reels based on real-time user interactions or historical viewing preferences.The system (100) of claim 1, wherein the system operates in a real-time live mode to generate and deliver highlights during ongoing estos matches.

Images