Discourse Coherence

25.1 Coherence relations

1. What are a nucleus and a satellite in Rhetorical Structure Theory (RST)?
   The two spans of a relation: the nucleus is more central to the writer’s purpose and interpretable independently; the satellite is less central and interpretable only with respect to the nucleus. Some symmetric relations hold between two nuclei.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 571
2. Define the RST relations Reason, Elaboration and Evidence with examples.
   Reason: the satellite gives the reason for an animate agent’s action (“[Jane took a train…] [She had to attend a conference.]”). Elaboration: the satellite adds detail (“[Dorothy was from Kansas.] [She lived in the midst of the great Kansas prairies.]”). Evidence: the satellite adds information meant to convince the reader of the nucleus (“[Kevin must be here.] [His car is parked outside.]”).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 571
3. Define the RST relations Attribution and List.
   Attribution: the satellite gives the source of reported speech (“[SAT Analysts estimated] [NUC that sales declined]”). List: a multinuclear series of nuclei without contrast (“[Billy Bones was the mate;] [Long John, he was quartermaster]”).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 572
4. How are RST relations drawn graphically?
   An arrow from the satellite to the nucleus, labeled with the relation:

   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 572
5. What are EDUs?
   Elementary discourse units (discourse segments): the leaves of an RST tree, spans of a sentence, clause or phrase. Finding their boundaries is an important task; they are analogous to constituents in syntax.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, pp. 572–573
6. What is the RST Discourse TreeBank?
   The largest discourse corpus: English Penn Treebank documents with full RST parses using 78 relations in 16 classes. RST treebanks also exist for Spanish, German, Basque, Dutch and Brazilian Portuguese.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 573
7. How can RST structure help summarization?
   Nuclei presumably express more important information than satellites, which can be dropped.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 573
8. How is PDTB annotation lexically grounded?
   Annotators start from discourse connectives (because, although, when, since, as a result) and mark the connective and its two spans, Arg1 and Arg2, then label the sense.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 573
9. How does PDTB annotate implicit relations?
   For adjacent sentences with no explicit connective, annotators choose the connective that could have signaled the relation (e.g. implicit = as a result) and then label its sense.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 573
10. How does PDTB differ from RST in the structure it annotates?
    PDTB annotates only pairwise relations between spans, with no commitment to higher-level structure; RST builds a global tree over the whole discourse.
    • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 573
11. What are the four top-level classes of the PDTB sense hierarchy?

    class	example types
    TEMPORAL	asynchronous (precedence, succession), synchronous
    CONTINGENCY	cause (reason, result), condition, pragmatic cause
    COMPARISON	contrast, concession
    EXPANSION	conjunction, instantiation, restatement, alternative, exception, list

    • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 574
12. Why was the Chinese Discourse TreeBank annotated differently from PDTB?
    Chinese marks far fewer relations with explicit connectives than English, so annotators mapped sentence pairs directly to 11 sense tags without starting from a connective.
    • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 25, §25.1, p. 574