DeXTER, the alpha version

Amazing turn up for the online demo of the alpha version of the DeXTER app!!! This app allows users to explore visually and interactively textual material enriched with layers of information such as relations of people, organisations and locations, mapping, sentiment, and much more!


What happens in the passage from digitised sources to digital data? How unsupervised can this process really be? This is my experience of working with a digital heritage collection

So you've digitised your sources, now you have thousands and thousands of digital material and... you're lost. You thought you'd finally be able to access everything but you are struggling to navigate the abundance of material which paradoxically seems to have obscured information rather than bring it to light. Where do you start? These are some of the questions that inspired DeXTER - DeepTextMiner. You can read more about the technical description of DeXTER in my previous post here


The presentation

More than 40 people attended the online seminar From digitised sources to digital data on 17 February 2021. In the presentation, I discuss most of the struggles researchers find themselves confronted with when dealing with digital sources. I also stress the importance of a critical engagement with the digital as each step has direct, potentially heavy consequences on the material (and therefore on the results). The presentation is based on the following paper.


A critical engagement with the digital is fundamental as each step has direct, potentially heavy consequences on the material (and therefore on the results).


An app, you said?

DeXTER is a generalisable workflow that uses cutting edge deep learning techniques to contextually enrich digital material. For example, it allows users to identify referential entities such as people, organisations, and locations in the material, geo-coding the locations, identify the positive, negative, and neutral emotions that link the entities over time and across the whole collection. But how can users explore such an enriched amount of information? Instead of static visualisations, we created an interactive app that allows users to navigate the information independently. This means detached from any pre-conceived research question but also allowing researchers to truly explore the layers of meaning from multiple perspectives.


DeXTER allows researchers to access many layers of information visually and interactively and form multiple perspectives


Current and forthcoming developments

Currently, we are adding the final touches to the app. DeXTER will soon be released in the form of an Open Access app, of a GitHub repository and a tutorial so that researchers will be able to replicate the methodology with their own material. Stay tuned! In the meanwhile, please enjoy the video of the demo below :)))