4. Text Segmentation

The text segmentation interface allows you split a newly imported plain-text file into the segments you wish to code. The same interface allows you to change segment boundaries even after you have started coding. Initially, an imported text is treated as a single segment. You can then click at the places you want segment boundaries to be made. The "Sentences" button will automatically segment at sentence boundaries.

Figure 2 shows the text Segmentation Interface. The text is shown in the main text window, and on the left are a number of buttons (the toolbar). These buttons are either mode buttons (which change how you interact with the text window) or action buttons (which actually do something directly).

4.1 Modes

The segmentation interface allows two modes of interaction:

• Segment Mode: each click on the text window will produce a new segment divider at the point where you click. If there is already a segment marker at that point, it is deleted. All key presses are ignored.
• Edit Mode: in this mode, you can edit the text as in a normal text edit window. Note that you cannot delete segment markers using the delete/backspace keys.

4.2 Actions

There are two action buttons:

• Paragraphs: produces a segment boundary at each double paragraph mark in the text.
• Sentences: produces a segment boundary after each end of sentence. Note that the algorithm to spot ends of sentences is not perfect, and may go wrong if a sentence doesn't end in usual punctuation (e.g., . ? !). It can also go wrong where these symbols appear in other contexts, e.g., as the second char in "1. Introduction". It does however handle phenomena such as "e.g.", "i.e.", "etc.", "...". When the system produces wrong segmentation, insert/delete segmentation by clicking at the appropriate place while in Segment mode.

4.3 Note on Segmentation

The typical process after importing a plain text file is to initially hit the Sentences button, then add in segment marks where you see fit, and delete bogus marks.

Note that if you get it wrong now, you can return from coding and change the borders.

A problem occurs with embedded elements -- cases where a rhetorically dependent stretch of text occurs within another node. For instance, we might wish to treat the embedded clause in the following as dependent on the main clause: John, -- I think you know him -- is here for two weeks. At present, the interface does not handle such cases. A simple solution is for the user to move the embedded text outside of the enclosing text.

When you are ready to start coding, hit the "Codings" button on the Interface selection bar at the top.

4.4 Summary of Functionalities