Research On Semantic-based Cross-media Content Quality Assessment

The purpose of cross-media content quality assessment is to use computer methods to measure people's subjective perception of omni-media content.The current methods for analyzing the quality of all media are divided into subjective evaluation and objective evaluation.The objective evaluation is mostly based on the resolution and frame rate to evaluate the video quality,the intensity and pitch to measure the sound quality,and the writing standard to judge the text quality.However,the new demand is to start from the deep content of the media,think about the values and connotations it contains,and be able to refer to the public's response.The establishment of a cross-media content quality assessment system will help to enhance the exploration of deep understanding and analysis of cross-media content.This paper proposes a cross-media content quality evaluation method.First,build a multi-modal positivity classification model,and then use Bayesian average method to comprehensively evaluate indicators to obtain quality scores.This paper uses core technologies such as self-attention,cross-attention and gating mechanism to construct a multi-modal positivity classification model.Self-attention strengthens the capture of long-term dependencies,cross-attention focuses on the interaction between bimodal information,and the gating mechanism regulates the information flow in the interaction process.This multi-modal positivity classification model proves that the cross-interaction based on the gating mechanism plays an important role in the ablation experiment.And compared with the current more advanced multi-modal sentiment classification algorithm on the self-constructed multi-modal data set,it verifies that the multi-modal positivity classification model proposed in this paper has outstanding effects.Finally,using the Bayesian averaging method of mathematical statistics,referring to the two evaluation dimensions of content positivity and being discussed,a quality score that conforms to people's subjective feelings is calculated.In general,in view of the lack of a unified technical framework for media content quality analysis,we will study cross-modal content quality analysis technologies,construct a quality intelligent evaluation model that conforms to mainstream values,and provide social creators with media content quality intelligent analysis and analysis.Evaluation has important research significance and practical value.