In this project, we introduce a user-friendly web application designed to analyse political news articles. The application is designed primarily for political scientists but can also benefit journalists and researchers interested in examining sentiment towards entities in text or tracking changes in the portrayal of topics over time. The analysis pipeline Exploratory Data Analysis, Named Entity Recognition with Entity Disambiguation, Sentiment Analysis, Community Detection, Topic Modelling, N-grams and Text Summarisation. The tool was developed using Python and the Shiny framework, incorporating primarily LLMs models e.g. BERT. Users can analyse predefined datasets related to the Russia-Ukraine War and the Israel-Palestine Conflict, or upload their own articles for examination. The application is available online within our Faculty network, providing an opportunity for broader academic collaboration. In addition, a blame and praise recognition model was developed but was not integrated into the application due to the limited availability of training data; however, preliminary results indicate promising potential for further research in this area.
Our application "Global Times: Articles Analysis" is divided into four modes: Single, Double, All and About.
The Single Mode allows users to analyse one article in detail. Users can select an article from the predefined datasets or upload their own. The right menu provides a dropdown to choose from available datasets, such as Gaza before conflict. Alternatively, users can upload a text file for analysis.
Once an article is selected, the analysis begins. A progress bar and loading animation indicate the status of the process.
Upon completion, the analysis displays the article's header, summary, and legends, along with highlighted entities in the text, using colours and icons to indicate sentiment.
Hovering over a highlighted entity reveals detailed information, including its type and sentiment.
Interactive plots, such as sentiment distribution and word clouds, further illustrate key insights.
The Double Mode allows users to compare two articles side by side, enabling direct analysis of similarities and differences.
Users can upload one or both articles through the right menu.
Both articles are analysed independently, with results displayed side by side. Headers, text highlights, and legends are provided for each.
Plots such as entity distribution comparison offer a visual representation of differences between the two articles
The All Mode provides an overview of entire datasets, offering powerful filters and visualisation options. An example plot showcases insights from a selected dataset.
Filters, such as selecting an entity type, dynamically update the plots.
Users can examine specific word contexts using the concordance table.
Filters can also refine the concordance table, narrowing results to specific criteria.
The application provides loading animations during analysis and dynamically updates with new plots after uploading a dataset.
The About Mode provides access to the user guide, helping users navigate the application effectively.
Before running the application, install Docker: Docker Desktop.
Open Terminal or Command Prompt:
-
Download the Docker image:
docker pull szuvarska/globaltimesanalysis:latest
-
Run the Docker container:
docker run -p 8000:8000 -it --rm szuvarska/globaltimesanalysis:latest
This will map port 8000 on your local machine to port 8000 in the container.
-
Access the application: Open your web browser and go to the following address:
http://127.0.0.1:8000/
To stop the container:
- Press
CTRL+Cin the terminal where the container is running. - If needed, stop manually:
docker ps # List running containers docker stop <container-id> # Replace with actual container ID
- Łukasz Grabarski (@LukaszGrabarski)
- Marta Szuwarska (@szuvarska)
- Dr Anna Wróblewska (@awroble)
- Prof. Agnieszka Kaliska
- Prof. Anna Rudakowska
- Dr Daniel Dan