User Interface

First, on the upper-right corner of the page, you can see the metadata associated with the current document. These include the title of the issue, its publication date, the newspaper it is coming from, and the total number of pages of the issue.

The main panel of this page is the document viewer. It offers a high-resolution view of the issue and the location of the articles that were automatically extracted. When an article is clicked on, it is highlighted by the viewer, and its textual content generated by the automatic text recognition process is displayed in a panel on the right side of the window.

If the document contains named entities, they are displayed on this page. When an article is selected, the list of named entities is updated to reflect the named entities present in this article. All entities detected are associated with a particular class (persons, locations, organisations or human productions. Some of these mentions are linked to an entry in the knowledge base when possible. The associated page is accessible by clicking on the information button. Clicking on an entity reveals the mentions that were extracted from the text of the article or the issue.

Compound articles

The article separation is difficult, and the results are not always satisfying depending on the user's needs. The ability to create compound articles has been set up to counteract this and give users more control over the data. We can identify at least two problems that can come up with article separation. The first is an under-segmentation case, where multiple articles were identified as one. This case is more difficult to process currently, as it would imply modifying the underlying available article segmentation. The second case is that of over-segmentation. In this case, a single article is segmented into multiple parts. The creation of compound articles has been set up to counteract this particular problem. You can activate the compound article creation mode on the document show page.

Then, once in this mode, you can select several article parts from the viewer that are added to a list. Once the selection is finished, you can validate the creation of the compound article by giving it a title. If an issue contains previously created compound articles, they can be seen on the top-right panel of the page. Clicking on it will highlight its article parts in the viewer, and a new button will appear in the "Datasets handling" panel.