Hearst, M. A. (1993). TextTiling: A Quantitative Approach to Discourse Segmentation. Technical Report UCB:S2K-93-24, Berkeley, CA.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....and because the automation of lexical chaining introduced errors. When building lexical chains manually, Morris and Hirst missed relations which Hearst s implementation of their algorithm found. Many of the relations found algorithmically were spurious and arose because of word sense ambiguities [Hearst, 1993]. 4.1.2 Hearst Hearst developed a technique to automatically divide long expository texts into segments several paragraphs in length, each of which was about a single subtopic. She chose to linearly segment text partly because of Skorochod ko s work on the structure of texts and because of the ....

Hearst, M. A. (1993). TextTiling: A quantitative approach to discourse segmentation. Technical Report 93/24, University of California, Berkeley.

....groups of sentences, or even fixed length windows of terms, a separate similarity score could be computed for each passage, and the passage level scores might prove to be a better indicator of relevance than document similarity. Several recent experiments have addressed this issue. Hearst [37] proposes a method for partitioning a full text document into multiparagraph units selected according to their coherence as a single topic. The algorithm works by computing the similarity (as for query document similarity described above) between adjacent overlapping text blocks (3 5 sentences ....

Marti A. Hearst. Texttiling: A quantitative approach to discourse segmentation. Technical Report 93/24, University of California, Berkeley, 1993.

....process than centering would be required to explain the relationship between utterances (32) and (33) simply because these utterances span a discourse segment boundary. The second problem is that listeners perceive segment boundaries at various levels of granularity [Passonneau and Litman, 1993; Hearst, 1994; Flammia and Zue, 1995; Hirschberg and Nakatani, 1996] and some segment boundaries are fuzzy [Passonneau and Litman, 1996] For example in discourse A above, 5 out of 7 subjects placed a segment boundary between utterances 29 and 30, while 4 out of 7 subjects placed a segment boundary ....

....and Litman, 1996] This is a naturally occurring exemplar of the first discourse in Figure 9; segment 15 is an interruption and segment 16 is a continuation of segment 14. This analysis is also supported by: 1) the obvious change in content and lexical selection [Morris and Hirst, 1991; Hearst, 1994] and (2) the fact that utterance 9 is an INFORMATIONALLY REDUNDANT UTTERANCE, IRU, In this part of the dialogue, the goal is to put the blue cap and its subcomponents onto the main pump body. The rubber ring is a subcomponent of the blue cap. Thus putting the rubber ring into the blue cap is a ....

Marti Hearst. Text tiling: A quantitative approach to discourse segmentation. In Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics, ACL94, pages 9--16, 1994. 29

....process than centering would be required to explain the relationship between utterances (32) and (33) simply because these utterances span a discourse segment boundary. The second problem is that listeners perceive segment boundaries at various levels of granularity [Passonneau and Litman, 1993; Hearst, 1994; Flammia and Zue, 1995; Hirschberg and Nakatani, 1996] and some segment boundaries are fuzzy [Passonneau and Litman, 1996] For example in discourse A above, 5 out of 7 subjects placed a segment boundary between utterances 29 and 30, while 4 out of 7 subjects placed a segment boundary ....

....and Litman, 1996] This is a naturally occurring exemplar of the first discourse in Figure 2; segment 15 is an interruption and segment 16 is a continuation of segment 14. This analysis is also supported by: 1) the obvious change in content and lexical selection [Morris and Hirst, 1991; Hearst, 1994] and (2) the fact that utterance 9 is an INFORMATIONALLY REDUNDANT UTTERANCE, IRU, which re realizes the content of utterance 3, and reintroduces its content in the current context [Walker, 1993; Walker, 1996] Thus, using hierarchical recency to determine U for the purposes of ....

Marti Hearst. Text tiling: A quantitative approach to discourse segmentation. In Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics, ACL94, pages 9--16, 1994.

....as expected but rather fluctuates rapidly. Attempting to use a time interval for identification of an item would not result in well defined, topically distinct items. 5.2 Method 2: Using a Topic Segmentor Our second approach involved the utilization of Marti A. Hearst s Text Tiling algorithm [7] to locate topic boundaries within expository text. The TextTiling algorithm would report topic (or segment) boundaries, and each segment could be classified as an item consisting of a group of co authors. This would result in acceptable SSB input format. The motivation for this approach was that ....

Hearst, Marti A. TextTiling: A Quantitative Approach to Discourse Segmentation, Technical Report UCB: S2K-93-24, 1993

....of text translation similarity. 1. Introduction The study of large collections of texts and their translations has recently received much attention in the #eld of computational linguistics. In this paper, we extend an application of earlier research on quantitative discourse analysis, by Marti Hearst #Hearst, 1993#, to a trilingual corpus. The interest of the parallel corpus is that it can be used as an alternative resource to evaluate this discourse analysis method. A parallel corpus also o#ers some interesting other uses of the discourse representations than the #monolingual# text segmentation ....

....section 2 we brie#y discuss the concept of lexical cohesion, which is used to express similarity of pairs of adjacent text units. The resulting similarity measurements can be viewed as a division of a discourse #signal into a sequence of frames. The discussion summarizes earlier reseach of Marti Hearst #Hearst, 1993# and pinpoints some di#erences between our re implementation and her original proposal. We will then #section 3# brie#y discuss the trilingual corpus that we used for experimentation and evaluation and the types of linguistic and other analysis we applied to it. Section 4 discusses the ....

[Article contains additional citation context not shown here]

Hearst, M. 1993. TextTiling: a quantitative approach to discourse segmentation. Technical Report 93#24, Project Sequoia, University of California, Berkeley.

....process than centering would be required to explain the relationship between utterances (32) and (33) simply because these utterances span a discourse segment boundary. The second problem is that listeners perceive segment boundaries at various levels of granularity [Passonneau and Litman, 1993; Hearst, 1994; Flammia and Zue, 1995; Hirschberg and Nakatani, 1996] and some segment boundaries are fuzzy [Passonneau and Litman, 1996] For example in discourse A above, 5 out of 7 subjects placed a segment boundary between utterances 29 and 30, while 4 out of 7 subjects placed a segment boundary ....

....and Litman, 1996] This is a naturally occurring exemplar of the first discourse in Figure 2; segment 15 is an interruption and segment 16 is a continuation of segment 14. This analysis is also supported by: 1) the obvious change in content and lexical selection [Morris and Hirst, 1991; Hearst, 1994] and (2) the fact that utterance 9 is an INFORMATIONALLY REDUNDANT UTTERANCE, IRU, which re realizes the content of utterance 3, and reintroduces its content in the current context [Walker, 1993; Walker, 1996] Thus, using hierarchical recency to determine Un Gamma1 for the purposes of ....

Marti Hearst. Text tiling: A quantitative approach to discourse segmentation. In Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics, ACL94, pages 9--16, 1994.

....process than centering would be required to explain the relationship between utterances (32) and (33) simply because these utterances span a discourse segment boundary. The second problem is that listeners perceive segment boundaries at various levels of granularity [Passonneau and Litman, 1993; Hearst, 1994; Flammia and Zue, 1995; Hirschberg and Nakatani, 1996] and some segment boundaries are fuzzy [Passonneau and Litman, 1996] For example in discourse A above, 5 out of 7 subjects placed a segment boundary between utterances 29 and 30, while 4 out of 7 subjects placed a segment boundary ....

....and Litman, 1996] This is a naturally occurring exemplar of the first discourse in Figure 2; segment 15 is an interruption and segment 16 is a continuation of segment 14. This analysis is also supported by: 1) the obvious change in content and lexical selection [Morris and Hirst, 1991; Hearst, 1994] and (2) the fact that utterance 9 is an INFORMATIONALLY REDUNDANT UTTERANCE, IRU, which re realizes the content of utterance 3, and reintroduces its content in the current context [Walker, 1993; Walker, 1996] Thus, using hierarchical recency to determine Un Gamma1 for the purposes of ....

Marti Hearst. Text tiling: A quantitative approach to discourse segmentation. In Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics, ACL94, pages 9--16, 1994.

....process than centering would be required to explain the relationship between utterances (32) and (33) simply because these utterances span a discourse segment boundary. The second problem is that listeners perceive segment boundaries at various levels of granularity [Passonneau and Litman, 1993; Hearst, 1994; Flammia and Zue, 1995; Hirschberg and Nakatani, 1996] and some segment boundaries are fuzzy [Passonneau and Litman, 1996] For example in discourse A above, 5 out of 7 subjects placed a segment boundary between utterances 29 and 30, while 4 out of 7 subjects placed a segment boundary ....

....and Litman, 1996] This is a naturally occurring exemplar of the first discourse in Figure 9; segment 15 is an interruption and segment 16 is a continuation of segment 14. This analysis is also supported by: 1) the obvious change in content and lexical selection [Morris and Hirst, 1991; Hearst, 1994] and (2) the fact that utterance 9 is an INFORMATIONALLY REDUNDANT UTTERANCE, IRU, 16 In this part of the dialogue, the goal is to put the blue cap and its subcomponents onto the main pump body. The rubber ring is a subcomponent of the blue cap. Thus putting the rubber ring into the blue cap ....

Marti Hearst. Text tiling: A quantitative approach to discourse segmentation. In Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics, ACL94, pages 9--16, 1994.

....into a larger number of smaller segments might be possible and be necessary for the given texts. And so we will have to consider the evaluation method that the agreement with human subjects is tested in future. However, since human subjects do not always agree with each other on segmentation[5, 3, 14], our evaluation method using the texts in the questions with model answers is considered to be a good simplification. Several other methods to text segmentation have been proposed. Kozima[7] and Youmans[17] proposed statistical measures(they are named LCP and VMP respectively) which indicate the ....

M.A. Hearst. Texttiling: A quantitative approach to discourse segmentation. Technical Report 93/24, University of California, Berkeley, 1993.

.... on where segment boundaries are, either because they construct different mental representations of the segmentation of a discourse, or because segments are naturally defined at varying levels of granularity (Passonneau and Litman, 1993; Grosz and Hirschberg, 1992; Passonneau and Litman, 1994; Hearst, 1994). To illustrate the problem, consider the continuation in 14 of the discourse excerpt in 9 from (Walker and Prince, In Press) 14) it was an emergency for her j to pick up the phone right away. Her i sister] j not being home, she i hung up. Her i sister] j came home a short time later, ....

Hearst, Marti. 1994. Text tiling: A quantitative approach to discourse segmentation.

....on the Internet. The FAQ Finder project has shown that when there is an existing collection of questions and answers, as found in the FAQ files, question answering can be reduced 12 Non frivolous examples of such files also exist in the RTFM archive. 13 The TextTiling technique proposed in (Hearst, 1993) seems particularly suitable here. SOLAR STAR SYSTEMS Acamar TNG The Vengeance Factor Alpha Centauri TOS Metamorphosis . Minos Korva (11 lightyears from McAlister C5 ....

Hearst, M. A. 1993. TextTiling: A quantitative approach to discourse segmentation. Technical Report Sequoia 93/94, Computer Science Department, University of California, Berkeley.

.... conclusion:premise Phi Phi H H Also D core:evidence conclusion:premise Phi Phi H H and E Figure 2: The RDA analysis of (1) These rates of agreement are similar to those found in studies of (nonembedded) segmentation agreement (Grosz Hirschberg 1992; Passonneau Litman 1993; Hearst 1993). However, our assessment of RDA reliability differs from this work in several key ways. For one thing, our subjects coders are not naive about their task and the data is not spoken. Further, the task is more complex than identifying locations of segment boundaries. Example hypotheses and initial ....

Hearst, M. 1993. Texttiling: A quantitative approach to discourse segmentation. Technical Report 93/24, U. of Californa, Berkeley.

....needs. In this paper we describe such an effort: we add associational information to a hierarchically structured lexicon in order to better serve a text labeling task. An algorithm for partitioning a full length expository text into a sequence of subtopical discussions is described in [9]. Once the partitioning is done, we need to assign labels indicating what the subtopical discussions are about, for the purposes of information retrieval and hypertext navigation. One way to label texts, when working within a limited domain of discourse, is to start with a pre defined set of ....

Marti A. Hearst. TextTiling: A quantitative approach to discourse segmentation. Technical Report 93/24, Sequoia 2000.

No context found.

Hearst, M. A. (1993). TextTiling: A Quantitative Approach to Discourse Segmentation. Technical Report UCB:S2K-93-24, Berkeley, CA.

No context found.

Hearst, M. : TextTiling: a Quantitative Approach to Discourse Segmentation. Technical Report, UCB:S2K93 -24, pp.33-64. (1993)

No context found.

Hearst, M. : TextTiling: a Quantitative Approach to Discourse Segmentation. Technical Report, UCB:S2K-93-24, pp.33-64. (1993)

No context found.

M.A. Hearst, `TextTiling: A Quantitative Approach to Discourse Segmentation ', Technical Report 93/24, University of California, Berkeley, 1993.

No context found.

Hearst, M. A. (1993). "TextTiling: A quantitative approach to discourse segmentation". Tech. Rep. 93/24, Dept. of Computer Science, University of California, Berkeley.

No context found.

Marti Hearst. 1993. Text tiling: A quantitative approach to discourse segmentation. Technical report, University of California, Berkeley, Sequoia.

No context found.

Marti Hearst. 1993. Text tiling: A quantitative approach to discourse segmentation. Technical report, University of California, Berkeley, Sequoia.

No context found.

Marti Hearst. 1993. Text tiling: A quantitative approach to discourse segmentation. Technical report, University of California, Berkeley, Sequoia.

No context found.

M.A. Hearst, TextTiling: A quantitative approach to discourse segmentation, Technical Report 93/24, University of California, Berkeley.

No context found.

Hearst, M. 1993. TextTiling: A quantitative approach to discourse segmentation. Berkeley, CA: University of Berkeley, 1993.

No context found.

Marti Hearst. 1993. Text tiling: A quantitative approach to discourse segmentation. Technical report, University of California, Berkeley, Sequoia.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC