Semantic-Aware Representation of Multi-Modal Data for Data Ingress: A Literature Review

TL;DR

This literature review addresses the challenge of efficiently ingesting and semantically representing large, multi-modal data in data lakes, with particular attention to temporal dynamics in time-series data. It surveys state-of-the-art embedding and fusion techniques post-2020, highlighting the predominance of contrastive learning for mono-modal representations and the ubiquity of early/late fusion in multi-modal setups, while also noting emerging dynamic fusion methods. The study identifies temporal awareness as a key gap and emphasizes practical IR implications for data lakes, proposing dynamic, time-aware approaches as fruitful directions for future work. Overall, the paper provides a roadmap for designing semantically-rich, storage- and retrieval-efficient representations to support scalable data ingress and retrieval in evolving, multi-modal data ecosystems.

Abstract

Machine Learning (ML) is continuously permeating a growing amount of application domains. Generative AI such as Large Language Models (LLMs) also sees broad adoption to process multi-modal data such as text, images, audio, and video. While the trend is to use ever-larger datasets for training, managing this data efficiently has become a significant practical challenge in the industry-double as much data is certainly not double as good. Rather the opposite is important since getting an understanding of the inherent quality and diversity of the underlying data lakes is a growing challenge for application-specific ML as well as for fine-tuning foundation models. Furthermore, information retrieval (IR) from expanding data lakes is complicated by the temporal dimension inherent in time-series data which must be considered to determine its semantic value. This study focuses on the different semantic-aware techniques to extract embeddings from mono-modal, multi-modal, and cross-modal data to enhance IR capabilities in a growing data lake. Articles were collected to summarize information about the state-of-the-art techniques focusing on applications of embedding for three different categories of data modalities.