Databases Reference
In-Depth Information
Research on detecting opinions, sentiment and subjectivity is concerned with identifying what
people think or feel, in terms of opinions and emotions expressed.
A dialogue/speech act represents the illocutionary meaning of an utterance, or the action
performed by the utterance. Dialogue act modeling is the task of labeling each conversation
turn with the dialogue act(s) it is intended to perform.
Decision detection is the task of identifying sentences related to a decision process. This may
involve identifying sub-types of decision sentences.
Action item detection is the task of identifying sentences that relate to the assignment of
responsibility for completing tasks. This may also involve identifying sub-types of action item
sentences.
Decision and action item detection can be thought of as focused summarization.
Synchronous conversations, especially written ones like chats, often require disentanglement
to determine reply-to relationships between turns.
Extracting a fragment quotation graph from a conversation can reveal a finer level conversa-
tional structure which can be beneficial to other mining tasks.
3.7
FURTHER READING
For subjectivity research, the best general reference is by
Pang and Lee
[
2008
]. They include nu-
merous case studies including research on blog data. The topic is available online
10
and the topic
site contains a useful searchable bibliography.
Jurafsky and Martin
[
2008
] have a chapter on dialogue and conversation, including coverage
of dialogue acts and adjacency pairs, as well as a discussion of the relation between dialog acts and
the theory of speech acts [
Austin
,
1962
], [
Searle
,
1975
]. This topic also includes discussions on many
mining tasks not covered here, such as named entity recognition, relation extraction and rhetorical
parsing.
On topic segmentation, Jurafsky and Martin briefly describe unsupervised and supervised
approaches as well as segmentation evaluation. A very recent, comprehensive review on topic seg-
mentation is provided by
Purver
[
2011
].
On topic modeling, David Mimno has maintained a bibliography of relevant pa-
pers and available software
11
. A review paper has also been posted by Blei and Lafferty
12
,
and
Steyvers and Griffiths
[
2006
] give a very gentle introduction to probabilistic topic models such
as LDA.
Search WWH ::
Custom Search