An Efficient and Explanatory Image and Text Clustering System with Multimodal Autoencoder Architecture

TL;DR

This paper introduces a novel multimodal autoencoder system that efficiently clusters and interprets video and text data, demonstrating its application in summarizing news topics with thematic clusters and descriptive phrases.

Contribution

The paper presents a new CRVAE model with parallel CNN and LSTM encodings, integrated with a system for clustering, alignment, and LLM-based interpretation of multimodal news data.

Findings

Effective clustering of news videos into thematic groups

Successful generation of descriptive phrases for each cluster

Application to COVID-19 and Winter Olympics news topics

Abstract

We demonstrate the efficiencies and explanatory abilities of extensions to the common tools of Autoencoders and LLM interpreters, in the novel context of comparing different cultural approaches to the same international news event. We develop a new Convolutional-Recurrent Variational Autoencoder (CRVAE) model that extends the modalities of previous CVAE models, by using fully-connected latent layers to embed in parallel the CNN encodings of video frames, together with the LSTM encodings of their related text derived from audio. We incorporate the model within a larger system that includes frame-caption alignment, latent space vector clustering, and a novel LLM-based cluster interpreter. We measure, tune, and apply this system to the task of summarizing a video into three to five thematic clusters, with each theme described by ten LLM-produced phrases. We apply this system to two news topics, COVID-19 and the Winter Olympics, and five other topics are in progress.