Automatic MetaData extraction is a key technology needed by modern production systems. The AME Group is working on different aspects of this domain;
- tool's capabilities and performances,
- knowledge sharing and facilitating collaboration on AME methods, algorithms and projects,
- metadata schemas and best practices
- Sharing good practices, state of the art and catalogue of AME tools (2021)
- Development ML/DL models for target audience identification (Q1 2021)
- Study on action detection and identification (Q1 2020)
- Study on fake news detection (Q1 2020)
- Development of ML/DL models for authorship identification (Q3 2020)
- Development of ML/DL models for fake news identification (Q3 2020)
- Development of a POC for authorship identification (Q4 2020)
- Development of a POC integrating fake news and authorship identification (Q4 2020)
Automatic Metadata Extraction (AME) tools can be used to produce more information (including structured metadata) that is needed by modern production systems, at a lower cost. Artificial intelligence techniques such as machine learning and deep learning are behind most AME tools.
The EBU AME group is part of the EBU "Metadata and Artificial Intelligence" activities. Its main goal is to help Members adopt AI-based automatic information extraction tools, such as:
- face recognition and identification
- speaker identification
- location, event and object detection and identification
- natural language processing (NLP)
- action detection and identification
To so, the group will share best practices on AME tools, performances, state of the arts and research trends. The EBU is developing an AI tool to generate high-level tags from written contents. This tool uses state of the art Natural Language Processing and Machine Learning algorithms. It will be demonstrated and available for EBU members as a POC.
Related EBU work
The EBU AI Benchmarking project which develops tools to evaluate speech-to-text transcription.
The EBU "AI Data Pool" pre-study, which proposes a framework to share resources for training and assessing AI tools.