A System for Retrieval and Exploration of Figures from Collections of Research Articles. A demo of the system can be found here. For more elaborate information please refer to: FigExplorer: A System for Retrieval and Exploration of Figures from Collections of Research Articles. Saar Kuzi, ChengXiang Zhai, Yin Tian, and Haichuan Tang. In Proceedings of SIGIR 2020. paper
The goal of this system is to facilitate the exploration of research article collections using figures and includes the following main functionalities:
- Searching for figures using keyword queries.
- Searching for figures using different textual representations of a figure and different retrieval models.
- Collecting user feedback in order to build test collections.
- Exploring related figures for a seed figure using figure embeddings.
In order to set up the system, the following steps should be followed:
- Building a collection of figures from a set of pdf files.
- Learning figure representations (embeddings) and inferring semantic relations between figures (optional).
- Building a figure index.
- Setting up the web-application.
To facilitate to deployment of the system, we provide a toy dataset for which the system is configured to work smoothly.