DocViz: A Web Interactive Document Visualization Tool with Sentence-BERT
What is it?
DocViz is a web app for you to visualize and inspect your documents. You can input up to 4 different documents. The model will compute the vector representation of each sentence in your document, and plot them in a 3D space.
How should I use it?
It is easy! Follow the link at the bottom of this page to go to the application page. Input document names and document text into the corresponding boxes, and click on the PLOT button! It may take a while for the results to be shown, so be patient!
How does it work?
Check the How it works page for the details, including what models I am using and how to interpret the visualization
Why do you make it?
Traditional text mining tools mainly provides word-level, frequency based analyses, such as summary statistics of the document, word clouds, term frequency graphs, etc. I want to create a tool that utilizes state-of-the-art NLP models for text analysis. This visualization app is a step forward that focuses on sentence-level semantic information in input documents. It is designed for digital humanities researchers and people who are interested in NLP to play around.
The application site is embedded in this Github Page. The repo for the web app is here.
What should I use it for?
- Get a quick grasp of your text data from a sentence-level perspective.
- Compare the semantics of different documents from a higher level.
- Come up with questions and hypotheses of text data.
- Serve as a stepping stone for further close reading if you have large amounts of documents.
- If you are interested in NLP and digital humanities!