Click on the links on the top left to navigate to different random elements of the debates. The left half of the screen shows an original document of a debate (or a page thereof), and on the right you can see the processed version of the same document, or an utterance within it. If you refresh the browser, a new random document will be presented.
Full session and Single page are self-explanatory. Utterance, President and Orador present a single utterance, but in Presidnt the speaker is always the President of the Assembly and in Orador the speaker is someone who was given the floor, started speaking, was interrupted, and is identified only as "Orador" in subsequent utterances, until the President grants the floor to another speaker.
Note: There are three utterances in all utterance views, to provide context. The utterance you are evaluating is the one in the middle, between the two lines of dashes ('----------').
Note: If you see the text "Utterance does not exist in this session", this means that particular type of utterance (e.g. Orador) is not present in that document. It's not an error, just hit refresh and move to the next document.
Evaluate the detection of page headers. Headers are the first line of the page, showing information like page number, date and session identification. They must be eliminated so they don't get included in speaker utterances.
Evaluate the detection of the end of session. The session ends on the last utterance, so this task consists of checking whether there are unidentified utterances in the end of a document (i.e. something like "O Sr. Presidente: - Blah blah blah", indicating that someone spoke)
Evaluate whether utterances were correctly tagged.
Evaluate whether utterances of the President of the Assembly are correctly attributed. Utterances from the president have the attribute role, with value "president".
Evaluate whether Orador instances are correctly attributed. This involves going back to the utterance where the President granted the floor and verifying that the speaker identified as Orador (attribute putative_speaker in the processed document) is indeed the Orador.