To explain it on a very basic level, it saves sentences and expressions individually in a database so as to remember them later when we encounter them in a similar sentence. The program creates specific folders for each project, and the user has to place the documents that are going to be translated, the translation memories and the glossaries in them. When the project is opened, OmegaT will extract the text that has to be translated from the documents it recognizes. During the process of translating each segment, it will add the translated text to the translation memory.
Among some of the most important features of OmegaT we'll find the following: partial coincidences, searches by keywords, glossaries, translation memories or to be able to make use of previous translations in new projects.
If you're looking for a new translation tool to add to those you are already using, this application could be of great help to you.