Audience behavior analysis-oriented precise content pushing method for radio and television

Abstract

This invention discloses a method for precise content delivery in broadcasting based on audience behavior analysis, belonging to the field of broadcast television information processing technology. The method includes the following steps: collecting continuous viewing records generated by the same account during alternating use on multiple terminals, organizing them according to a unified chronological order, marking the locations where the viewing rhythm suddenly changes, forming a behavior change sequence; reviewing the time segments corresponding to the marked locations of sudden changes in viewing rhythm in the behavior change sequence, segmenting the continuous viewing records, extracting the segments where continuity is interrupted, and marking them as candidate segments for subject switching in the behavior change sequence. This invention achieves clearer expression of interests and improves the accuracy of content matching by separating the interleaved viewing behaviors across multiple terminals and constructing independent interest trajectories; simultaneously, it filters and pushes content based on interest-driven paths, ensuring consistent delivery direction, reducing interest interference, and enhancing the stability of content distribution and the continuity of the viewing experience.

Description

Technical Field

[0001] This invention relates to the field of broadcast and television information processing technology, specifically to a method for precise content delivery in broadcasting based on audience behavior analysis. Background Technology

[0002] Precise content delivery in broadcasting based on audience behavior analysis refers to the continuous collection and systematic organization of multi-dimensional behavioral information of viewers during the broadcasting, television, or converged media process. This includes information such as channel switching patterns, viewing time, replay records, interactive operations, and time period preferences. Further analysis of these behavioral changes is conducted to extract features that reflect viewers' interests and content consumption habits. Based on this, a correspondence between viewer preferences and program content is established. Program resources are then selectively selected and arranged according to this correspondence, and content that aligns with the current viewer interests is delivered at appropriate times, thus forming a dynamic content distribution mechanism based on viewer behavior.

[0003] The existing technology has the following shortcomings: In existing technologies, broadcast content delivery typically relies on continuous recording and analysis of viewer behavior based on account dimensions. When the same account is used alternately on different devices, the viewing behaviors generated on each device overlap and interleave in time. During processing, the system may assume that these behaviors all originate from the same user, failing to effectively distinguish the differences between actual users. This results in multiple, potentially independent behavioral trajectories being pieced together into a single interest path. In this situation, interest preferences formed under different usage scenarios can become mixed, leading to inconsistencies in direction judgment during subsequent content matching and delivery. This can cause the delivered content to be misaligned or even conflicting, affecting the overall accuracy of the delivery.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method for precise content delivery in broadcasting based on audience behavior analysis, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for precise content delivery in broadcasting based on audience behavior analysis, comprising the following steps: Collect continuous viewing records generated by the same account during the alternating use of multiple terminals, organize them in a unified time sequence, mark the locations where the viewing rhythm suddenly changes, and form a sequence of behavioral changes; By reviewing the time segments corresponding to the locations of sudden changes in viewing rhythm marked in the behavior change sequence, the continuous viewing records are segmented and organized to extract the segments where continuity is interrupted and mark them as candidate segments for subject switching in the behavior change sequence. Expand the behavioral change sequence of the candidate segment for subject switching, compare the viewing content change direction within each candidate segment segment by segment, identify the positional relationship of inconsistent content preferences before and after, and divide the behavioral change sequence into mutually separated independent interest trajectory segments. The behavior change sequence is reorganized based on independent interest trajectory segments. Different interest trajectory segments are grouped according to their respective continuous time ranges and their positions are re-corresponded to the original time sequence to form multiple interest advancement paths that do not interfere with each other. The system reads subsequent viewing records based on the re-corresponding interest-driven path, filters content on the interest-driven path corresponding to the current time position, and organizes the push order according to the content preferences corresponding to the path, so as to maintain the consistency of the push direction when multiple terminals are used alternately.

[0007] Preferably, the steps for forming the behavior change sequence are as follows: The viewing records generated by the same account on multiple terminals are centrally collected and expanded into a unified data format that includes playback start time, playback end time, pause time, playback resume time, content identifier, and terminal source identifier, and then mapped to a unified time base for time sorting. Based on time-ordered viewing records, viewing records with overlapping times are split to form a continuous viewing record sequence; The system analyzes the time intervals, content identifier changes, and terminal source changes between adjacent viewing records one by one in a continuous viewing record sequence, and generates rhythm change markers at the corresponding time positions. By combining rhythm change markers with timeline extensions to form rhythm change segments, these segments are embedded into a continuous viewing sequence and alternated with rhythm stable segments to construct a behavioral change sequence.

[0008] Preferably, the process for marking candidate segments for main body switching is as follows: The viewing rhythm change positions marked in the behavior change sequence are segmented and processed. The consecutive viewing records before and after the viewing rhythm change position are selected to form a time segment, and the playback start time, playback end time, content identifier and terminal source are retained to construct a local viewing sequence. Based on the local viewing sequence, the continuous viewing record is analyzed for continuity. The continuous segments, time interruption segments, and content switching segments are divided into sub-segments by the alignment relationship between the playback end time and the playback start time and the changes in content identifiers. By combining sub-segments, the temporal connection status, content identifier change direction, and terminal source change between adjacent sub-segments are compared to identify the location where continuity is interrupted and form a location segment. The location segments are marked in the behavior change sequence and written to the time axis position to form candidate segments for subject switching.

[0009] Preferably, the continuity break position is determined at the connection position where there is a time gap between sub-segments, the content identifier changes category and the terminal source changes. The position segment is formed by covering the end time range of the previous sub-segment and the start time range of the next sub-segment around the continuity break position. At the same time, the break time, the content identifier before the break and the content identifier after the break are recorded and written into the time axis position corresponding to the behavior change sequence.

[0010] The preferred process for segmenting independent interest trajectory segments is as follows: For the subject switching candidate segments marked in the behavior change sequence, the continuous viewing records within the corresponding time range are extracted and formed into a viewing sequence arranged in chronological order. At the same time, the content identifiers are converted into content category tags, program theme tags and presentation form tags to construct the content sub-segment sequence. The content category tags, program theme tags, and presentation format markers of adjacent content segments are compared segment by segment to record the direction of content change and form the content change path. By combining content change paths, we can identify changes in content category and program theme between adjacent content sub-segments to determine the locations of content preference inconsistencies and form content preference inconsistency segments. The behavioral change sequence is divided into segments based on inconsistent content preferences to form mutually separate independent interest trajectory segments.

[0011] Preferably, the steps for forming the interest-driven path are as follows: expand the independent interest trajectory segments and arrange the continuous viewing records in chronological order, while recording the playback start time, playback end time, content category tags, and program theme tags to form a time range identifier; Independent interest trajectory segments are sorted according to their start time based on time range identifiers and time ranges are marked on a unified time axis. At the same time, overlapping time ranges are split to form time segments corresponding to independent interest trajectory segments. By combining time segments, continuous viewing records in the sequence of behavioral changes are matched one by one and classified into corresponding independent interest trajectory segments; Based on the matching results, consecutive viewing records within the same independent interest trajectory segment are linked together to form an interest path; The positions of interest paths on a unified timeline are re-corresponded to form interest-driven paths that do not interfere with each other.

[0012] Preferably, the correspondence of the interest advancement path on a unified timeline is defined by limiting the mapping relationship between time segments and independent interest trajectory segments, so that each time segment corresponds to only one independent interest trajectory segment, and the content category tag and program theme tag are kept connected within the interest path to ensure the consistent expression of the interest path in the time dimension and content dimension.

[0013] Preferably, the following steps are taken: Read subsequent viewing records based on the re-corresponding interest-driven path, filter content on the interest-driven path corresponding to the current time position, and organize the push order according to the content preferences corresponding to the path: Locate the current time position along a unified timeline and match it with the time range of the interest advancement path to determine the interest advancement path corresponding to the current time position; The continuous viewing records within the interest-driven path are arranged and content category tags and program theme tags are extracted to form a content evolution path; By combining the content evolution path, the candidate content records after the current time position are read one by one and filtered according to the content category tags and program theme tags to form a candidate content set; The candidate content is arranged according to playback time and sorted sequentially based on the content evolution path to form the content push order; As the timeline advances, the current time position is updated, and interest-driven path matching, content filtering, and order reorganization are repeatedly performed to maintain consistency in the push direction. The technical effects and advantages provided by this invention in the above technical solution are as follows: This invention reconstructs the time sequence and identifies behavioral changes in continuous viewing records generated by the same account during alternating use on multiple terminals. It segments the originally overlapping viewing behaviors according to changes in viewing rhythm and further identifies candidate segments for subject switching and independent interest trajectory segments. This allows the viewing behaviors formed by different users to be expressed separately in the time sequence, thereby avoiding the overlap of different interests in the same behavior sequence. This makes the expression of content preferences clearer and more stable, and can be judged based on a single interest trajectory in the subsequent content matching process, effectively improving the accuracy and consistency of content push.

[0014] This invention constructs an interest-driven path and establishes a correlation between the current time position and the corresponding path, ensuring that the content filtering and push order always revolves around the corresponding interest-driven path. This allows subsequent viewing behavior to maintain a continuous relationship with the existing interest trajectory as time progresses, thereby avoiding cross-interference between different interest directions. It also enables content push to maintain a unified direction even in scenarios where multiple terminals are used alternately, enhancing the stability and continuity of the content distribution process and improving the overall coherence of the viewing experience. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of the broadcasting and television precise content push method based on audience behavior analysis according to the present invention. Detailed Implementation

[0017] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0018] This invention provides, for example Figure 1 The broadcasting content delivery method based on audience behavior analysis, as shown, includes the following steps: Collect continuous viewing records generated by the same account during the alternating use of multiple terminals, organize them in a unified time sequence, mark the locations where the viewing rhythm suddenly changes, and form a sequence of behavioral changes; In scenarios involving the alternating use of multiple devices, to uniformly organize continuous viewing behavior and identify changes in viewing rhythm, the continuous viewing records are processed in detail. The specific process is as follows: Viewing records generated by the same account across multiple terminals are centrally aggregated. Each viewing record is expanded according to a unified data format, ensuring that each record includes playback start time, playback end time, pause time, resume time, channel or content identifier, and terminal source identifier. All viewing records are mapped to the same time base. By uniformly converting the time fields in the records from each terminal, all time information is represented from the same timing starting point. Based on this, the viewing records generated by each terminal are sorted in chronological order. For viewing records with overlapping time periods, the overlapping time periods are split, dividing the continuous playback process into multiple time segments. The corresponding terminal source and content identifier are retained for each time segment. This allows all viewing behaviors to form a continuous sequence of viewing records on a unified timeline, maintaining a clear and unambiguous temporal connection between each record within this sequence.

[0019] On the established continuous viewing record sequence, the temporal and content relationships between adjacent viewing records are read one by one. By parsing the time interval between the current viewing record and the next viewing record, and combining the changes in content identifiers and terminal source identifiers corresponding to the two viewing records, the rhythmic state of the viewing behavior in the time progression is characterized. In the specific implementation, the time interval between adjacent viewing records is divided into three types: continuous playback interval, short interruption interval, and cross-segment switching interval. At the same time, the change relationship between content identifiers is divided into three cases: continuation of the same type of content, switching between different types of content, and cross-channel jump. The change of terminal source is divided into two states: the terminal remains consistent and the terminal changes. When any of the time interval type change, content type change, or terminal source change occurs simultaneously between adjacent viewing records, a rhythm change mark is generated at the corresponding time position and attached to the corresponding viewing record node, thereby gradually forming a behavior change sequence structure with rhythm change marks in the continuous viewing record sequence.

[0020] In a sequence of behavioral changes marked with rhythm changes, several consecutive viewing records are selected forward and backward from the position of each rhythm change mark, forming a time segment around the rhythm change mark. Within this time segment, the continuous playback relationship between the viewing records is analyzed in detail. By comparing the start and end connections between adjacent viewing records within the segment, playback interruption points, re-entry points, and continuous playback segments are identified. Combined with the change trajectory of content markers within the segment, the evolution of the content theme within this time segment is organized, establishing a one-to-one correspondence between the rhythm change mark and the corresponding time segment. In this process, the rhythm change mark, originally located at a single time point, is expanded into a rhythm change segment covering multiple viewing records. This rhythm change segment is then embedded into the behavioral change sequence, allowing the behavioral change sequence to express the specific time range of the rhythm change and the corresponding viewing behavior structure while maintaining the chronological order.

[0021] The rhythm change segments in the entire behavioral change sequence are organized as a whole, arranged in chronological order, and the time intervals between adjacent rhythm change segments are analyzed. By identifying the parts of the continuous viewing record that are not covered by rhythm change segments, these parts are defined as rhythm stable segments and are arranged alternately with rhythm change segments to form a complete sequence structure. In this structure, each rhythm stable segment corresponds to a continuous viewing behavior with coherent content evolution, while each rhythm change segment corresponds to the position where the viewing behavior switches. By unifying the continuous viewing record, rhythm change markers, and rhythm change segments, a behavioral change sequence that can be continuously expressed in the time dimension and clearly identified in the structure of rhythm change positions is finally formed, thus providing a complete and orderly data foundation for subsequent subject switching recognition and interest trajectory division based on rhythm changes.

[0022] The formation process of the behavior change sequence is as follows: A one-to-one correspondence is established between consecutive viewing records arranged in a uniform chronological order and rhythm change markers at corresponding time positions. Rhythm change segments are then formed by extending forward and backward from the positions of the rhythm change markers. These rhythm change segments are then embedded into the original time sequence, forming a structured time sequence together with consecutive viewing segments that do not exhibit rhythm changes. This results in a behavior change sequence that maintains temporal continuity while containing rhythm switching information. For example, the following consecutive viewing records are formed under the same account: 10:00-10:20 watching news channel A (terminal 1), 10:21-10:40 watching financial program B (terminal 1), 10:41-10:45 no operation interval, 10:46-11:10 watching animation program C (terminal 2), 11:11... At 11:30, children's program D (terminal 2) is watched. As time progresses, at 10:41, due to a change in time interval and a switch from terminal 1 to terminal 2, the content changes from finance to children's content. A rhythm change marker is generated at this time position, and a rhythm change segment, such as 10:38-10:48, is formed around this marker. After embedding this segment into the original time sequence, a structured expression can be obtained: stable rhythm segment (10:00-10:40, news + finance continuous viewing) - rhythm change segment (10:38-10:48, terminal switch + content switch) - stable rhythm segment (10:46-11:30, animation + children's continuous viewing). In this way, a behavioral change sequence containing time sequence information, rhythm change markers, and rhythm change segments can be formed.

[0023] By reviewing the time segments corresponding to the locations of sudden changes in viewing rhythm marked in the behavior change sequence, the continuous viewing records are segmented and organized to extract the segments where continuity is interrupted and mark them as candidate segments for subject switching in the behavior change sequence. The behavioral change sequence is unfolded and structurally organized layer by layer around the time segments corresponding to the changes in viewing rhythm. Continuous viewing records are segmented to extract the segments where continuity is interrupted and form candidate segments for subject switching. The specific process is as follows: Around the marked locations of sudden changes in viewing rhythm in the behavioral change sequence, the time points corresponding to each rhythm change mark are segmented. On the timeline, with this time point as the center, the first few records in the continuous viewing records are selected sequentially forward, and the last few records in the continuous viewing records are selected sequentially backward, so that the selected viewing records form a complete and continuous time segment. Within this time segment, the playback start time, playback end time, content identifier, terminal source, and operation trigger time information of each viewing record are retained and arranged in chronological order, so that a complete and non-overlapping local viewing sequence is formed within this time segment. At the same time, the original position of the rhythm change mark in this local viewing sequence is maintained, so that the rhythm change mark becomes a positioning reference point within this time segment, thus providing a clear time range and structural basis for subsequent segmentation and organization of continuous viewing records within this time segment.

[0024] Within the formed time segment, the connection status between consecutive viewing records is analyzed one by one. By aligning and comparing the playback end time of each viewing record with the playback start time of the next viewing record, the time breakpoints in the playback process are identified. At the same time, combined with the changes in content identifiers, the continuity of the viewing content in the time progression is divided. In the specific processing, viewing records where the playback end time and the playback start time of the next playback are completely connected and the content identifiers are consistent are grouped into the same continuous segment. Cases where there is a time gap between the playback end time and the playback start time of the next playback are divided into time interruption segments. Cases where the content identifier changes from news to children's or from sports to film and television are divided into content switching segments. Time interruption segments and content switching segments are recorded independently, so that the consecutive viewing records within the time segment are divided into multiple sub-segments with clear boundaries. Each sub-segment contains a continuous time range and a corresponding content change trajectory, thus forming a segmented structure based on playback connection relationships and content change relationships.

[0025] Based on the completion of sub-segment division, the connection positions between each sub-segment are analyzed segment by segment. By comparing the time connection status, content identifier change direction, and terminal source change between adjacent sub-segments, the locations where continuity is interrupted are identified. In the specific processing, when there is a time gap between the end time of a sub-segment and the start time of the next sub-segment, and at the same time the corresponding content identifier undergoes a category change and the terminal source switches from a living room terminal to a mobile terminal or vice versa, this connection position is identified as a continuity break position, and the preceding sub-segment is covered from this position. The end time range of a sub-segment covers the beginning time range of the next sub-segment. These two time ranges are combined to form a complete location segment. In this location segment, the specific time point of the break, the content markers before and after the break, and the changes in the terminal source are recorded. This location segment can fully reflect the entire process of the viewing behavior transitioning from one continuous state to another. This location segment is also mapped one-to-one with the rhythm change markers in the behavior change sequence, ensuring that the rhythm change markers and the continuous break positions are consistent on the time axis, thereby guaranteeing the uniformity of the behavior change sequence structure.

[0026] In the behavioral change sequence, the identified locations where continuity is interrupted are marked. The start and end times of each location segment are written into the corresponding timeline position of the behavioral change sequence. Subject switching candidate segment identifiers are added within the corresponding segment range. This allows the behavioral change sequence to further superimpose subject switching candidate segment information on the original rhythm change markers and rhythm change segments. In this process, the subject switching candidate segments are uniformly organized with the corresponding continuous viewing records, time segment division results, and rhythm change markers. This ensures that each subject switching candidate segment has a clear time range, complete contextual viewing records, and a clear expression of the break position, thus forming a multi-layered behavioral change sequence that includes continuous viewing records, rhythm change segments, and subject switching candidate segments. This provides accurate and traceable segmentation basis for subsequent division and reconstruction of different interest trajectories.

[0027] Expand the behavioral change sequence of the candidate segment for subject switching, compare the viewing content change direction within each candidate segment segment by segment, identify the positional relationship of inconsistent content preferences before and after, and divide the behavioral change sequence into mutually separated independent interest trajectory segments. For viewing behavior within candidate segments of subject switching, the direction of content change is meticulously depicted. Combined with differences in content characteristics before and after, the behavioral change sequence is hierarchically expanded and structurally adjusted to divide it into separate, independent interest trajectory segments. The specific implementation steps are as follows: For the candidate segments of subject switching already marked in the behavior change sequence, all continuous viewing records within the time range corresponding to each candidate segment are extracted one by one and arranged in chronological order to form a continuous viewing sequence within the candidate segment. In this viewing sequence, the playback start time, playback end time, content identifier, and terminal source are recorded for each record. At the same time, the content identifier in each viewing record is further decomposed and converted into clear content category tags, such as news, sports, children's, film and television, variety shows, and education, and further refined into program theme tags, such as financial news, international news, football matches, animation, TV series, and interview programs. At the same time, each viewing record is attached with a presentation format tag, such as live program, on-demand program, or replay program, so that the viewing sequence within the candidate segment forms a multi-layered expression structure in the time and content dimensions. On this basis, the viewing sequence is divided according to the time progression direction, and several viewing records with continuous consistency between content category tags and program theme tags are merged into a content sub-segment, thus forming an internal structure of candidate segments composed of multiple content sub-segments arranged sequentially.

[0028] Within the established sequence of content segments, the connection between adjacent segments is analyzed segment by segment. By sequentially reading the changes in content category tags, program theme tags, and presentation format markers between the preceding and following content segments, the direction of content change is clearly recorded. Specifically, when two adjacent content segments change from financial news to animation, from football matches to TV dramas, or from variety shows to educational programs, this change is marked as a shift in content direction, and the shift relationship is recorded at the corresponding time position in the behavioral change sequence. At the same time, the end time of the content segment before the shift and the start time of the content segment after the shift are retained, so that the direction of content change has a clear start and end position expression on the timeline, thereby forming a content change path arranged in chronological order within the candidate segment.

[0029] Based on the completion of content change path recording, the content preference relationship between adjacent content segments is identified segment by segment. By comparing content category tags, program theme tags, and presentation format markers, when adjacent content segments undergo cross-category changes in content category tags, and there is no continuous correlation between program theme tags, and the presentation format marker changes, this position is identified as the position where content preference changes. In specific processing, positions where the content category changes twice consecutively are combined, and multiple content segments within the range of continuous changes are grouped into the same change segment. The start and end times of this change segment are recorded so that the change segment can fully reflect the process of viewing behavior shifting from one interest orientation to another. At the same time, the change segment is marked in the behavior change sequence, so that the positional relationship of inconsistent content preferences forms a clear segment expression on the timeline. In the specific implementation process, the content segments arranged chronologically within the candidate segment can be read one by one, and the position of each content direction switch can be recorded. When multiple consecutive changes in content category occur during continuous time progression, these continuously changing content segments are combined into a change sequence. The time point of the first content category change in this change sequence is determined as the starting change point, and the time point after the last content category change is determined as the ending change point, thus obtaining a time segment that completely covers the content transformation process. This time segment is then marked as a content preference inconsistency segment. For example, in a certain candidate segment, there are the following viewing records: watching financial news from 14:00 to 14:15, 14: From 14:25 to 14:25, the user watched international news; from 14:26 to 14:40, they watched animation programs; and from 14:41 to 14:55, they watched children's programs. This content segmentation forms a path of change: financial news → international news → animation programs → children's programs. The transition from international news to animation programs and from animation programs to children's programs constitutes two consecutive content category changes. Therefore, 14:25 is designated as the starting point of this change, and 14:55 as the ending point. The time interval from 14:25 to 14:55 is marked as a content preference inconsistency segment, and this segment is marked at the corresponding time position in the behavioral change sequence. This makes this segment clearly present on the timeline as a complete process of shifting interest from news-related to children's-related interests.

[0030] Based on the identified content preference inconsistencies, the behavioral change sequence is divided as a whole. Each content preference inconsistency segment serves as a boundary, dividing the behavioral change sequence into multiple independent interest trajectory segments. During the division process, continuous viewing records located between the same boundary segments and not covered by the content preference inconsistency segments are grouped into the same interest trajectory segment. The start time, end time, and corresponding content category evolution order are recorded for each interest trajectory segment, such as from financial news to international news and then to interview programs, forming a complete interest evolution path. At the same time, each interest trajectory segment is re-embedded into the behavioral change sequence according to its original chronological order, so that the behavioral change sequence maintains continuous expression in the time dimension and is structurally presented as multiple mutually separate interest trajectory segments with consistent internal content. This achieves a clear division of different interest tendencies and provides accurate data support for subsequent content filtering and push based on interest trajectories.

[0031] The behavior change sequence is reorganized based on independent interest trajectory segments. Different interest trajectory segments are grouped according to their respective continuous time ranges and their positions are re-corresponded to the original time sequence to form multiple interest advancement paths that do not interfere with each other. Based on the existing independent interest trajectory segments, the behavioral change sequences are reorganized and re-corresponded to the original time sequence, so that different interest trajectory segments form separate but continuously advancing interest paths on a unified time axis. The specific process is as follows: Each of the already defined independent interest trajectory segments is expanded one by one. The continuous viewing records contained within each independent interest trajectory segment are arranged in chronological order, and the start time, end time, content category tag, and program theme tag are retained for each viewing record. At the same time, a complete time range identifier is formed in each independent interest trajectory segment. For example, an independent interest trajectory segment corresponds to the time range of 08:10 to 08:45, which includes financial news from 08:10 to 08:20, international news from 08:21 to 08:30, and interview programs from 08:31 to 08:45. This makes the independent interest trajectory segment present a continuous progression in the time dimension and a progressively expanding theme in the content dimension, so that each independent interest trajectory segment has a complete expressive ability.

[0032] Based on the unfolding of each independent interest trajectory segment, the time range of all independent interest trajectory segments is uniformly organized. The independent interest trajectory segments are sorted in ascending order of their start time, and the start and end times of each independent interest trajectory segment are marked on a unified timeline. For example, the first independent interest trajectory segment corresponds to 08:10 to 08:45, the second independent interest trajectory segment corresponds to 08:40 to 09:10, and the third independent interest trajectory segment corresponds to 09:15 to 09:50. The time ranges with overlaps on the timeline are split, and the overlapping time interval of 08:40 to 08:45 is divided into independent time segments, each corresponding to a different independent interest trajectory segment. At the same time, the time interval of 08:45 to 09:15 is retained, so that the timeline forms a structure in which consecutive time segments are arranged sequentially, and each time segment corresponds to a unique independent interest trajectory segment, thereby ensuring that there are no conflicts between the interest trajectory segments on the timeline.

[0033] Based on the independent interest trajectory segments that have already been processed in terms of time, the continuous viewing records in the behavior change sequence are matched one by one. Each viewing record is read in chronological order on the timeline. For example, when reading the financial news record at 08:12, this record is assigned to the independent interest trajectory segment from 08:10 to 08:45; when reading the animation program record at 08:43, this record is assigned to the independent interest trajectory segment from 08:40 to 09:10; when reading the sports event record at 09:20, this record is assigned to the independent interest trajectory segment from 09:15 to 09:50. Through this one-by-one matching method, each viewing record in the original behavior change sequence is mapped to a specific independent interest trajectory segment, thereby completing the mapping transformation from a single time series to a multi-interest trajectory structure, while keeping the position of each viewing record on the timeline unchanged.

[0034] After establishing the correspondence between viewing records and independent interest trajectory segments, the viewing records within each independent interest trajectory segment are continuously organized. Viewing records belonging to the same independent interest trajectory segment and with adjacent times are linked together. For example, within the independent interest trajectory segment from 08:10 to 08:45, financial news, international news, and interview programs are connected in chronological order to form a continuously advancing content path; within the independent interest trajectory segment from 08:40 to 09:10, animation programs and children's programs are connected in chronological order to form another continuously advancing content path. By organizing the viewing records within different independent interest trajectory segments separately, each interest trajectory segment forms a structurally independent continuous path, avoiding the cross-mixing between different interest trajectories, thereby achieving the separate expression of multiple interest paths.

[0035] Based on the formation of multiple independent interest trajectory paths, the positional relationship of each interest path on a unified timeline is re-corresponded. Each interest path is embedded into the unified timeline according to its corresponding time range, so that the time sequence of each interest path is consistent with the original behavioral change sequence. For example, the interest path of financial news corresponds to the time range of 08:10 to 08:45, the interest path of children's programs corresponds to the time range of 08:40 to 09:10, and the interest path of sports events corresponds to the time range of 09:15 to 09:50. The alternation relationship between each path is preserved on the timeline, so that the situation of different interest paths corresponding to different time periods is clearly presented on the timeline. This forms a behavioral change sequence structure composed of multiple non-interfering and continuously advancing interest paths, providing a stable and clear basis for subsequent content selection and push on the corresponding interest paths.

[0036] Read subsequent viewing records based on the re-corresponding interest-driven path, filter content on the interest-driven path corresponding to the current time position, and organize the push order according to the content preferences corresponding to the path, so as to maintain the consistency of the push direction when multiple terminals are used alternately; By combining the correspondence between the interest-driven path and the timeline position, subsequent viewing records are continuously read, and the pushed content is filtered and ordered according to the content characteristics of the interest-driven path, so that the content push maintains a consistent direction in scenarios where multiple devices are used alternately. The specific implementation steps are as follows: The system locates the current time position along a unified timeline and, in conjunction with the already re-corresponding interest-driven paths, performs segment-by-segment matching of the time interval containing the current time position. During the matching process, the current time position is compared one by one with the time range corresponding to each interest-driven path. When the current time position is found to fall within the time interval of a certain interest-driven path, the system further reads the viewing record content corresponding to that time position and, based on the content category tags and program theme tags of the viewing record, assigns the current time position to an interest-driven path consistent with its content characteristics. For example, on the timeline, there is an interest-driven path from 17:30 to 18:10 corresponding to "Sports Events - Basketball Events - Event Commentary," and another interest-driven path from 17:50 to 18:20 corresponding to "Children's Animation - Children's Stories - Early Education." If the current time position is 17:55 and the viewed content is a basketball event, then the time position is assigned to the interest-driven path of "Sports Events - Basketball Events - Event Commentary," thus forming a unique correspondence between the current time position and the interest-driven path.

[0037] Based on the established interest-driven path, the historical viewing records within that path are systematically organized. These records are arranged chronologically, and the content category and program theme tags for each record are extracted. These tags are then linked together chronologically to form a complete content evolution path. For example, in the interest-driven path of "Sports Events - Basketball Events - Event Commentary," there are general sports news from 17:30 to 17:45, a live basketball event from 17:46 to 18:00, and an event commentary program from 18:01 to 18:10. This content evolution path clearly indicates that the current time position is within the live basketball event phase, thus providing a clear content stage for the interest-driven path at the current time position and offering a concrete basis for subsequent content selection.

[0038] For the time interval following the current time position, each available program is read one by one, and each candidate content record is arranged in chronological order. Each candidate content record is also labeled with its playback time, content category tag, and program theme tag. For example, in the time interval after 18:10, there are basketball game replays from 18:10 to 18:25, live football matches from 18:20 to 18:40, variety shows from 18:25 to 18:45, and animation programs from 18:30 to 18:50. After reading these candidate content records, each one is compared with the content evolution path of the current interest progression path. Only candidate content whose content category and program theme are consistent with "sports events - basketball events - event commentary" is retained. For example, basketball game replays are retained, while live football matches, variety shows, and animation programs are removed from the candidate content set, thus forming a candidate content set consistent with the current interest progression path.

[0039] After obtaining the selected candidate content set, the records within the set are organized sequentially. Each candidate content is arranged from earliest to latest according to its playback time, and further refined in conjunction with the content stage sequence in the interest progression path. For example, if the selected results include a basketball game replay from 18:10 to 18:25 and a game commentary extension program from 18:25 to 18:40, the basketball game replay is arranged first, followed by the game commentary extension program, while maintaining the progression relationship from game replay to commentary analysis. This ensures that the organized content push sequence is consistent in both the time and content dimensions, thus obtaining a content push order that matches the current interest progression path.

[0040] As the timeline continues to advance, the process of time positioning, interest-driven path matching, content evolution path extraction, and candidate content filtering and sorting is repeated at new time positions. When viewing behavior switches between different terminals, the timeline position is repositioned to ensure that the new time position still corresponds to the correct interest-driven path. For example, when the time advances to 18:30 and the viewing content switches to an animation program, the current time position is reclassified into the interest-driven path of "children's animation - children's stories - early childhood education," and subsequent content filtering and push order are reorganized on this path. This allows the content push direction to adjust with the changes in the interest-driven path, thus ensuring that the content push always follows the corresponding interest direction continuously when multiple terminals are used alternately, avoiding inconsistencies in the content direction.

[0041] This invention reconstructs the time sequence and identifies behavioral changes in continuous viewing records generated by the same account during alternating use on multiple terminals. It segments the originally overlapping viewing behaviors according to changes in viewing rhythm and further identifies candidate segments for subject switching and independent interest trajectory segments. This allows the viewing behaviors formed by different users to be expressed separately in the time sequence, thereby avoiding the overlap of different interests in the same behavior sequence. This makes the expression of content preferences clearer and more stable, and can be judged based on a single interest trajectory in the subsequent content matching process, effectively improving the accuracy and consistency of content push.

[0042] This invention constructs an interest-driven path and establishes a correlation between the current time position and the corresponding path, ensuring that the content filtering and push order always revolves around the corresponding interest-driven path. This allows subsequent viewing behavior to maintain a continuous relationship with the existing interest trajectory as time progresses, thereby avoiding cross-interference between different interest directions. It also enables content push to maintain a unified direction even in scenarios where multiple terminals are used alternately, enhancing the stability and continuity of the content distribution process and improving the overall coherence of the viewing experience.

[0043] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for precise content delivery in broadcasting based on audience behavior analysis, characterized in that: Includes the following steps: Collect continuous viewing records generated by the same account during the alternating use of multiple terminals, organize them in a unified time sequence, mark the locations where the viewing rhythm suddenly changes, and form a sequence of behavioral changes; By reviewing the time segments corresponding to the locations of sudden changes in viewing rhythm marked in the behavior change sequence, the continuous viewing records are segmented and organized to extract the segments where continuity is interrupted and mark them as candidate segments for subject switching in the behavior change sequence. Expand the behavioral change sequence of the candidate segment for subject switching, compare the viewing content change direction within each candidate segment segment by segment, identify the positional relationship of inconsistent content preferences before and after, and divide the behavioral change sequence into mutually separated independent interest trajectory segments. The behavior change sequence is reorganized based on independent interest trajectory segments. Different interest trajectory segments are grouped according to their respective continuous time ranges and their positions are re-corresponded to the original time sequence to form multiple interest advancement paths that do not interfere with each other. Read subsequent viewing records based on the re-corresponding interest-driven path, filter content on the interest-driven path corresponding to the current time position, and organize the push order according to the content preferences corresponding to the path.

2. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 1, characterized in that, The steps for forming a behavioral change sequence are as follows: The viewing records generated by the same account on multiple terminals are centrally collected and expanded into a unified data format that includes playback start time, playback end time, pause time, playback resume time, content identifier, and terminal source identifier, and then mapped to a unified time base for time sorting. Based on time-ordered viewing records, viewing records with overlapping times are split to form a continuous viewing record sequence; The system analyzes the time intervals, content identifier changes, and terminal source changes between adjacent viewing records one by one in a continuous viewing record sequence, and generates rhythm change markers at the corresponding time positions. By combining rhythm change markers with timeline extensions to form rhythm change segments, these segments are embedded into a continuous viewing sequence and alternated with rhythm stable segments to construct a behavioral change sequence.

3. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 2, characterized in that, The process of marking candidate segments for main body switching is as follows: The viewing rhythm change positions marked in the behavior change sequence are segmented and processed. The consecutive viewing records before and after the viewing rhythm change position are selected to form a time segment, and the playback start time, playback end time, content identifier and terminal source are retained to construct a local viewing sequence. Based on the local viewing sequence, the continuous viewing record is analyzed for continuity. The continuous segments, time interruption segments, and content switching segments are divided into sub-segments by the alignment relationship between the playback end time and the playback start time and the changes in content identifiers. By combining sub-segments, the temporal connection status, content identifier change direction, and terminal source change between adjacent sub-segments are compared to identify the location where continuity is interrupted and form a location segment. The location segments are marked in the behavior change sequence and written to the time axis position to form candidate segments for subject switching.

4. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 3, characterized in that, The continuity break position is determined at the connection position where there is a time gap between sub-segments, the content identifier changes category, and the terminal source changes. The position segment is formed by covering the end time range of the previous sub-segment and the start time range of the next sub-segment around the continuity break position. At the same time, the break time, the content identifier before the break and the content identifier after the break are recorded and written into the time axis position corresponding to the behavior change sequence.

5. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 3, characterized in that, The process of segmenting independent interest trajectory fragments is as follows: For the subject switching candidate segments marked in the behavior change sequence, the continuous viewing records within the corresponding time range are extracted and formed into a viewing sequence arranged in chronological order. At the same time, the content identifiers are converted into content category tags, program theme tags and presentation form tags to construct the content sub-segment sequence. The content category tags, program theme tags, and presentation format markers of adjacent content segments are compared segment by segment to record the direction of content change and form the content change path. By combining content change paths, we can identify changes in content category and program theme between adjacent content sub-segments to determine the locations of content preference inconsistencies and form content preference inconsistency segments. The behavioral change sequence is divided into segments based on inconsistent content preferences to form mutually separate independent interest trajectory segments.

6. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 5, characterized in that, The steps for forming an interest-driven development path are as follows: The independent interest trajectory segments are expanded and the viewing records are arranged in chronological order. At the same time, the playback start time, playback end time, content category tags, and program theme tags are recorded to form a time range identifier. Independent interest trajectory segments are sorted according to their start time based on time range identifiers and time ranges are marked on a unified time axis. At the same time, overlapping time ranges are split to form time segments corresponding to independent interest trajectory segments. By combining time segments, continuous viewing records in the sequence of behavioral changes are matched one by one and classified into corresponding independent interest trajectory segments; Based on the matching results, consecutive viewing records within the same independent interest trajectory segment are linked together to form an interest path; The positions of interest paths on a unified timeline are re-corresponded to form interest-driven paths that do not interfere with each other.

7. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 6, characterized in that, The correspondence of interest-driven paths on a unified timeline is defined by limiting the mapping relationship between time segments and independent interest trajectory segments, so that each time segment corresponds to only one independent interest trajectory segment, and the content category tags and program theme tags are kept connected within the interest path.

8. The method for precise content delivery in broadcasting based on audience behavior analysis according to claim 7, characterized in that, Based on the re-corresponding interest-driven path, subsequent viewing records are read, content is filtered on the interest-driven path corresponding to the current time position, and the push order is organized according to the content preferences corresponding to the path. The steps are as follows: Locate the current time position along a unified timeline and match it with the time range of the interest advancement path to determine the interest advancement path corresponding to the current time position; The continuous viewing records within the interest-driven path are arranged and content category tags and program theme tags are extracted to form a content evolution path; By combining the content evolution path, the candidate content records after the current time position are read one by one and filtered according to the content category tags and program theme tags to form a candidate content set; The candidate content is arranged according to playback time and sorted sequentially based on the content evolution path to form the content push order; As the timeline progresses, the current time position is updated, and the interest-driven path matching, content filtering, and order sorting are repeatedly performed to maintain consistency in the direction of push notifications.

Images