The relations used in ILEX

ILEX doesn't use quite the same set of relations as RST. This is partly because it operates primarily in a descriptive domain, and hence draws mainly on RST's PRESENTATIONAL relations, and partly due to theoretical differences between the motivations underlying RST's relations and those in ILEX.

Below is an inventory of the relations currently used in the system.

• Mononuclear relations:
  • RST-SIMILARITY: Like John, Bill is tall.
  • RST-CONTRAST: Unlike Bill, Bob is hungry.
  • RST-CONCESSION: Bob was tired, but he stayed awake.
  • RST-EXEMPLIFICATION: Most dogs have a name. For instance, my dog's name is Jip.
  • RST-AMPLIFICATION: My dog likes walks. Indeed, most dogs like walks.
  • RST-RESTATEMENT: My dog likes long walks. In other words, he's energetic.
  • RST-IN-THAT: The problem is hard, in that it needs lateral thinking.
  • RST-SPECIFICATION: John likes some classical composers. Specifically, he likes Mozart and Brahms.
  • RST-RATHER: John wasn't happy. Rather, he was angry.
• Multinuclear relations:
  • RST-LIST: Bob has three sisters: Tess, Jess and Bess.
  • RST-JOINT: John is 21 years old. He studies engineering.
• Binuclear relations:
  • RST-WHEREAS: Whereas Bill is tall, Jim is small.

As regards the differences between RST's relations and those used in ILEX, the biggest difference is that ILEX does not make use of ELABORATION relations. (For a rationale for this, see Section .) On top of this, there are two other important differences. Whereas in RST, relations are given abstract definitions, and not associated with particular linguistic signals, in ILEX, relations are associated explicitly with lexical signals (or groups of such signals), following Knott [].