Search

CN-121980525-A - Interactive data processing method and system based on artificial intelligence

CN121980525ACN 121980525 ACN121980525 ACN 121980525ACN-121980525-A

Abstract

The application provides an interactive data processing method and system based on artificial intelligence, wherein the method comprises the steps of collecting various interactive events within a preset time range; based on the difference of different types of interaction events on a data structure, unified semantic feature modeling is carried out on the interaction events to form semantic feature vectors, unified time sequence feature representation is built by combining a unified time expression mechanism, a model input feature sequence is built based on the unified time sequence feature representation, the model input feature sequence is input into an interaction event time sequence analysis model to be subjected to time sequence modeling processing, the interaction event time sequence modeling feature vectors are obtained, behavior sub-features corresponding to the different types of interaction events are divided, interaction event association feature vectors are built by carrying out interaction processing on the different types of behavior sub-features, fusion feature vectors are obtained after feature fusion processing, and interaction behavior analysis is carried out based on the fusion feature vectors to obtain interaction behavior analysis results.

Inventors

  • GAO JIAN

Assignees

  • 上海云梯信息科技有限公司

Dates

Publication Date
20260505
Application Date
20260408

Claims (10)

  1. 1. An interactive data processing method based on artificial intelligence is characterized by comprising the following steps: the method comprises the steps of collecting interaction event data, and collecting various types of interaction events within a preset time range, wherein the types of the interaction events comprise praise events, comment events, bullet screen events and forwarding events; the method comprises the steps of carrying out unified construction processing on characteristics, carrying out unified semantic characteristic modeling on interaction events based on the difference of the interaction events of different types on a data structure so as to map the interaction events of different types into semantic characteristic vectors with consistent structures, and constructing unified time sequence characteristic representation of the cross-type interaction events by combining a unified time expression mechanism; feature modeling and fusion processing are carried out, an interactive event time sequence analysis model is constructed, a model input feature sequence is constructed based on the unified time sequence feature representation, the model input feature sequence is input into the interactive event time sequence analysis model for time sequence modeling processing, so that an interactive event time sequence modeling feature vector is obtained, behavior sub-features corresponding to different types of interactive events are divided based on the interactive event time sequence modeling feature vector, interactive event correlation feature vectors are constructed through interactive processing of the different types of behavior sub-features, and feature fusion processing is carried out on the interactive event time sequence modeling feature vectors and the interactive event correlation feature vectors, so that fusion feature vectors are obtained; And the interactive behavior analysis is carried out based on the fusion feature vector so as to obtain an interactive behavior analysis result, wherein the interactive behavior analysis comprises a change trend analysis, a behavior association mode analysis, a key behavior driving analysis and an abnormal interactive analysis.
  2. 2. The interactive data processing method based on artificial intelligence according to claim 1, wherein the feature unified construction process comprises: Semantic alignment processing, namely carrying out semantic mapping processing on the interaction events based on the data structure difference of each type of interaction event, and uniformly converting the interaction events of multiple types into semantic feature vectors corresponding to preset semantic feature structures; And (3) time alignment processing, namely performing alignment and aggregation processing on the semantic feature vectors of the interaction events according to a time dimension by constructing a unified time window so as to construct unified time sequence feature representation of the cross-type interaction events.
  3. 3. The interactive data processing method based on artificial intelligence according to claim 1, wherein the feature modeling and fusion process comprises: The model input construction is carried out, the input construction processing is carried out on the unified time sequence feature representation, and a model input feature sequence meeting the input requirement of the interactive event time sequence analysis model is formed through carrying out structural recombination and input format conversion on the unified time sequence feature representation; modeling and reasoning the model, constructing the time sequence analysis model of the interaction event, inputting the model input feature sequence into the time sequence analysis model of the interaction event for time sequence modeling processing, and extracting dynamic change features of the interaction event in a time dimension to obtain time sequence modeling feature vectors of the interaction event corresponding to each unified time window; Performing association modeling processing, namely dividing the corresponding behavior sub-feature vectors according to the types of the interaction events based on the time sequence modeling feature vectors of the interaction events, and performing interaction processing on the behavior sub-feature vectors of different types to construct interaction event association feature vectors representing association relations among the interaction events of different types; And carrying out feature fusion processing on the basis of the interactive event time sequence modeling feature vector and the interactive event association feature vector so as to simultaneously represent the time variation feature and the behavior association feature of the interactive event in a unified feature space and obtain a fusion feature vector.
  4. 4. The interactive data processing method based on artificial intelligence according to claim 2, wherein the preset semantic feature structure comprises: The participation intensity feature is used for representing the participation degree of the interaction event; The expression degree feature is used for representing the information expression intensity of the interaction event; a propagation affecting feature for characterizing the propagation or diffusion capabilities of the interaction event; each type of interaction event is converted into a semantic feature vector containing the participation intensity features, the expression level features and the propagation influence features through semantic mapping processing.
  5. 5. The interactive data processing method based on artificial intelligence according to claim 2, wherein the semantic mapping process performs feature mapping process for praise event, comment event, barrage event and forwarding event respectively to map different types of interactive events to a unified semantic feature structure, and the semantic mapping process includes praise event semantic mapping process, comment event semantic mapping process, barrage event semantic mapping process and forwarding event semantic mapping process.
  6. 6. The interactive data processing method according to claim 3, wherein the association modeling process comprises: performing feature division on the interaction event time sequence modeling feature vector according to the interaction event type to obtain a plurality of behavior sub-feature vectors; Performing interaction processing based on the behavior sub-feature vectors to construct interaction features corresponding to different interaction event combinations; and carrying out combination processing based on the interaction characteristics to construct the interaction event associated feature vector.
  7. 7. The artificial intelligence based interactive data processing method according to claim 3, wherein the interactive processing comprises: And combining the behavior sub-feature vectors corresponding to at least two interaction events to construct corresponding interaction features.
  8. 8. The artificial intelligence based interactive data processing method according to claim 6, wherein the feature classification of the interactive event time-series modeling feature vector comprises: Based on a preset feature dimension corresponding relation, the interactive event time sequence modeling feature vector is divided into a plurality of sub-vectors according to feature dimension intervals corresponding to different interactive event types, and the sub-vectors are respectively used as corresponding behavior sub-feature vectors.
  9. 9. The interactive data processing method based on artificial intelligence according to claim 1, wherein the feature fusion process comprises: And splicing the interactive event time sequence modeling feature vector and the interactive event association feature vector to construct a fusion feature vector.
  10. 10. An artificial intelligence based interactive data processing system, comprising: The interactive event data acquisition module acquires various types of interactive events within a preset time range, wherein the types of the interactive events comprise praise events, comment events, bullet screen events and forwarding events; The feature unified construction processing module is used for carrying out unified semantic feature modeling on the interaction events based on the difference of the interaction events of different types on a data structure so as to map the interaction events of different types into semantic feature vectors with consistent structures, and constructing unified time sequence feature representation of the cross-type interaction events by combining a unified time expression mechanism; The feature modeling and fusion processing module is used for constructing an interaction event time sequence analysis model, constructing a model input feature sequence based on the unified time sequence feature representation, inputting the model input feature sequence into the interaction event time sequence analysis model for time sequence modeling processing to obtain an interaction event time sequence modeling feature vector, dividing behavior sub-features corresponding to different types of interaction events based on the interaction event time sequence modeling feature vector, constructing an interaction event correlation feature vector by carrying out interaction processing on the different types of behavior sub-features, and carrying out feature fusion processing on the interaction event time sequence modeling feature vector and the interaction event correlation feature vector to obtain a fusion feature vector; And the interaction behavior analysis module is used for carrying out interaction behavior analysis based on the fusion feature vector so as to obtain an interaction behavior analysis result, wherein the interaction behavior analysis comprises change trend analysis, behavior association mode analysis, key behavior driving analysis and abnormal interaction analysis.

Description

Interactive data processing method and system based on artificial intelligence Technical Field The application relates to the technical field of data processing, in particular to an interactive data processing method and system based on artificial intelligence. Background In the self-media platform, a plurality of types of interaction events such as praise events, comment events, bullet screen events, forwarding events and the like can continuously generate a large amount of interaction data, and the data are not only used for interaction display, but also can reflect the attention degree and feedback condition of a user on the content. With the continuous expansion of the interaction scale, the prior art generally only simply records or performs statistical processing on the interaction events, or analyzes the interaction events of a single type, so that it is difficult to perform unified processing and comprehensive analysis on the interaction events of multiple types. Because different types of interaction events have differences in the data structure, for example, praise events are represented in a number form, comment events contain text information, barrage events are presented as high-frequency continuous event streams, and forwarding events relate to propagation relations, the prior art is difficult to perform unified semantic feature modeling and time sequence modeling processing on the multiple types of interaction events. Meanwhile, the prior art lacks modeling means for association relations among different types of interaction events in the analysis process, and is difficult to describe dynamic change characteristics and behavior association modes of the interaction events in the time dimension, so that the interaction change trend and key driving factors cannot be accurately identified, and the accuracy and stability of the interaction behavior analysis result are affected. Disclosure of Invention The application provides an interactive data processing method and system based on artificial intelligence, which are used for solving the problems in the prior art. The application provides an interactive data processing method based on artificial intelligence, which comprises the following steps: the method comprises the steps of collecting interaction event data, and collecting various types of interaction events within a preset time range, wherein the types of the interaction events comprise praise events, comment events, bullet screen events and forwarding events; the method comprises the steps of carrying out unified construction processing on characteristics, carrying out unified semantic characteristic modeling on interaction events based on the difference of the interaction events of different types on a data structure so as to map the interaction events of different types into semantic characteristic vectors with consistent structures, and constructing unified time sequence characteristic representation of the cross-type interaction events by combining a unified time expression mechanism; feature modeling and fusion processing are carried out, an interactive event time sequence analysis model is constructed, a model input feature sequence is constructed based on the unified time sequence feature representation, the model input feature sequence is input into the interactive event time sequence analysis model for time sequence modeling processing, so that an interactive event time sequence modeling feature vector is obtained, behavior sub-features corresponding to different types of interactive events are divided based on the interactive event time sequence modeling feature vector, interactive event correlation feature vectors are constructed through interactive processing of the different types of behavior sub-features, and feature fusion processing is carried out on the interactive event time sequence modeling feature vectors and the interactive event correlation feature vectors, so that fusion feature vectors are obtained; And the interactive behavior analysis is carried out based on the fusion feature vector so as to obtain an interactive behavior analysis result, wherein the interactive behavior analysis comprises a change trend analysis, a behavior association mode analysis, a key behavior driving analysis and an abnormal interactive analysis. In one possible design, the feature unified building process includes: Semantic alignment processing, namely carrying out semantic mapping processing on the interaction events based on the data structure difference of each type of interaction event, and uniformly converting the interaction events of multiple types into semantic feature vectors corresponding to preset semantic feature structures; And (3) time alignment processing, namely performing alignment and aggregation processing on the semantic feature vectors of the interaction events according to a time dimension by constructing a unified time window so as to construct unified time sequence