SITS: A Hierarchical Nonparametric Model using Speaker Identity for Topic Segmentation in Multiparty Conversations
Nguyen, Viet-An and Boyd-Graber, Jordan and Resnik, Philip

Article Structure

Abstract

One of the key tasks for analyzing conversational data is segmenting it into coherent topic segments.

Topic Segmentation as a Social Process

Conversation, interactive discussion between two or more people, is one of the most essential and common forms of communication.

Modeling Multiparty Discussions

Data Properties We are interested in turn-taking, multiparty discussion.

Inference

To find the latent variables that best explain observed data, we use Gibbs sampling, a widely used Markov chain Monte Carlo inference technique (Neal, 2000; Resnik and Hardisty, 2010).

Datasets

This section introduces the three corpora we use.

Topic Segmentation Experiments

In this section, we examine how well SITS can replicate annotations of when new topics are introduced.

Evaluating Topic Shift Tendency

In this section, we focus on the ability of SITS to capture speaker-level attributes.

Related and Future Work

In the realm of statistical models, a number of techniques incorporate social connections and identity to explain content in social networks (Chang and Blei,

Topics

segmentations

Appears in 11 sentences as: Segmentations (1) segmentations (11)
In SITS: A Hierarchical Nonparametric Model using Speaker Identity for Topic Segmentation in Multiparty Conversations
  1. For evaluation, we used a standard set of reference segmentations (Galley et al., 2003) of 25 meetings.
    Page 4, “Datasets”
  2. Segmentations are binary, i.e., each point of the document is either a segment boundary or not, and on average each meeting has 8 segment boundaries.
    Page 4, “Datasets”
  3. To get reference segmentations , we assign each turn a real value from 0 to 1 indicating how much a turn changes the topic.
    Page 5, “Datasets”
  4. This results in a set of non-binary reference segmentations .
    Page 5, “Datasets”
  5. For evaluation metrics that require binary segmentations , we create a binary segmentation by setting a turn as a segment boundary if the computed score is 1.
    Page 5, “Datasets”
  6. Unlike the previous two datasets, Crossfire does not have explicit topic segmentations , so we use it to explore speaker-specific characteristics (Section 6).
    Page 5, “Datasets”
  7. Evaluation Metrics To evaluate segmentations , we use Pk (Beeferman et al., 1999) and WindowDiff (WD) (Pevzner and Hearst, 2002).
    Page 5, “Topic Segmentation Experiments”
  8. First, they require both hypothesized and reference segmentations to be binary.
    Page 5, “Topic Segmentation Experiments”
  9. Many algorithms (e. g., probabilistic approaches) give non-binary segmentations where candidate boundaries have real-valued scores (e.g., probability or confidence).
    Page 5, “Topic Segmentation Experiments”
  10. addition, because EMD operates on distributions, we can compute the distance between non-binary hypothesized segmentations with binary or real-valued reference segmentations .
    Page 6, “Topic Segmentation Experiments”
  11. Experimental Methods We applied the following methods to discover topic segmentations in a document:
    Page 6, “Topic Segmentation Experiments”

See all papers in Proc. ACL 2012 that mention segmentations.

See all papers in Proc. ACL that mention segmentations.

Back to top.

latent variable

Appears in 6 sentences as: latent variable (3) latent variables (3)
In SITS: A Hierarchical Nonparametric Model using Speaker Identity for Topic Segmentation in Multiparty Conversations
  1. Instead, we endow each turn with a binary latent variable lat, called the topic shift.
    Page 2, “Modeling Multiparty Discussions”
  2. This latent variable signifies whether the speaker changed the topic of the conversation.
    Page 2, “Modeling Multiparty Discussions”
  3. To find the latent variables that best explain observed data, we use Gibbs sampling, a widely used Markov chain Monte Carlo inference technique (Neal, 2000; Resnik and Hardisty, 2010).
    Page 3, “Inference”
  4. The state space is latent variables for topic indices assigned to all tokens z = {ZQM} and topic shifts assigned to turns 1 2 {lat}.
    Page 3, “Inference”
  5. We marginalize over all other latent variables .
    Page 3, “Inference”
  6. as a distinct latent variable (Wang and McCallum, 2006; Eisenstein et a1., 2010).
    Page 8, “Related and Future Work”

See all papers in Proc. ACL 2012 that mention latent variable.

See all papers in Proc. ACL that mention latent variable.

Back to top.

topic models

Appears in 4 sentences as: topic modeling (1) Topic models (1) topic models (2)
In SITS: A Hierarchical Nonparametric Model using Speaker Identity for Topic Segmentation in Multiparty Conversations
  1. Topics—after the topic modeling literature (Blei and Lafferty, 2009)—are multinomial distributions over terms.
    Page 2, “Modeling Multiparty Discussions”
  2. However, topic models alone cannot model the dynamics of a conversation.
    Page 2, “Modeling Multiparty Discussions”
  3. Topic models typically do not model the temporal dynamics of individual documents, and those that do (Wang et al., 2008; Gerrish and Blei, 2010) are designed for larger documents and are not applicable here because they assume that most topics appear in every time slice.
    Page 2, “Modeling Multiparty Discussions”
  4. For example: models having sticky topics over n-grams (Johnson, 2010), sticky HDP-HMM (Fox et al., 2008); models that are an amalgam of sequential models and topic models (Griffiths et al., 2005; Wal-
    Page 8, “Related and Future Work”

See all papers in Proc. ACL 2012 that mention topic models.

See all papers in Proc. ACL that mention topic models.

Back to top.

generative model

Appears in 3 sentences as: generative model (2) generative models (1)
In SITS: A Hierarchical Nonparametric Model using Speaker Identity for Topic Segmentation in Multiparty Conversations
  1. Topic segmentation approaches range from simple heuristic methods based on lexical similarity (Morris and Hirst, 1991; Hearst, 1997) to more intricate generative models and supervised methods (Georgescul et al., 2006; Purver et al., 2006; Gruber et al., 2007; Eisenstein and Barzilay, 2008), which have been shown to outperform the established heuristics.
    Page 1, “Topic Segmentation as a Social Process”
  2. These topics are part of a generative model posited to have produced a corpus.
    Page 2, “Modeling Multiparty Discussions”
  3. In this section, we develop SITS, a generative model of multiparty discourse that jointly discovers topics and speaker-specific topic shifts from an unannotated corpus (Figure la).
    Page 2, “Modeling Multiparty Discussions”

See all papers in Proc. ACL 2012 that mention generative model.

See all papers in Proc. ACL that mention generative model.

Back to top.

hyperparameters

Appears in 3 sentences as: hyperparameter (1) hyperparameters (3)
In SITS: A Hierarchical Nonparametric Model using Speaker Identity for Topic Segmentation in Multiparty Conversations
  1. Marginal counts are represented with - and >x< represents all hyperparameters .
    Page 4, “Inference”
  2. Initial hyperparameter values are sampled from U (0, 1) to favor sparsity; statistics are collected after 500 burn-in iterations with a lag of 25 iterations over a total of 5000 iterations; and slice sampling (Neal, 2003) optimizes hyperparameters .
    Page 6, “Topic Segmentation Experiments”
  3. 2008 Elections To obtain a posterior estimate of 7r (Figure 3) we create 10 chains with hyperparameters sampled from the uniform distribution U (0, l) and averaged 7r over 10 chains (as described in Section 5).
    Page 6, “Evaluating Topic Shift Tendency”

See all papers in Proc. ACL 2012 that mention hyperparameters.

See all papers in Proc. ACL that mention hyperparameters.

Back to top.

PDF available from www.aclweb.org.
Send us an email