Academic
Publications
SeLeCT: a lexical cohesion based news story segmentation system

SeLeCT: a lexical cohesion based news story segmentation system,Ai Communications,Nicola Stokes,Joe Carthy,Alan F. Smeaton

SeLeCT: a lexical cohesion based news story segmentation system   (Citations: 31)
BibTex | RIS | RefWorks Download
In this paper we compare the performance of three dis- tinct approaches to lexical cohesion based text segmen- tation. Most work in this area has focused on the dis- covery of textual units that discuss subtopic structure within documents. In contrast our segmentation task requires the discovery of topical units of text i.e. dis- tinct news stories from broadcast news programmes. Our approach to news story segmentation (the Se- LeCT system) is based on an analysis of lexical cohe- sive strength between textual units using a linguistic technique called lexical chaining. We evaluate the rela- tive performance of SeLeCT with respect to two other cohesion based segmenters: TextTiling and C99. Using a recently introduced evaluation metric WindowDifi, we contrast the segmentation accuracy of each system on both 'spoken' (CNN news transcripts) and 'writ- ten' (Reuters newswire) news story test sets extracted from the TDT1 corpus.
Journal: Ai Communications - AICOM , vol. 17, no. 1, pp. 3-12, 2004
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...Major approaches involve TextTiling [9], C99 [4], and lexical chaining [10]...
    • ...Stokes et al. [10] embodied word cohesion by lexical chaining for story segmentation...
    • ...In their SeLeCT system [10], related words are linked into chains and a high concentration point of chain starting and ending is an indication of a story boundary...

    Lei Xieet al. Laplacian Eigenmaps for Automatic Story Segmentation of Broadcast News

    • ...Rather th an adopting such a video-centric approach, like [21], our segmentation efforts on news video operate primarily on the text (transcribed speech), augmented by features from the video and audio m odalities...

    Gert-Jan Poulisseet al. News story segmentation in multiple modalities

    • ...Lexical chaining [6], [7] chains up related words and locates the positions where the count of chain starts and ends achieve local maxima...
    • ...Therefore, a high concentration of starting and/or ending chains is an indicator of story boundaries [6]...
    • ...Lexical Chaining [6] 0.5168 NCuts [11] 0.6197 Multi-feature Integration 0.7722...
    • ...Finally, we compared our method with several state-ofthe-art topic segmentation approaches on the TDT2 corpus and the results are shown in Table 3. We can see that the proposed multi-feature integration approach significantly outperform the TextTiling approach [5], the lexical chaining approach [6] and the NCuts approach [11]...

    Lei Xieet al. Integrating acoustic and lexical features in topic segmentation of Chi...

    • ...Three categories of cues have been explored for automatic story segmentation, namely acoustic/prosodic cues [1], visual cues [2] and lexical cues [3, 4]. Many typical acoustic or visual cues, such as significant pauses, speaker changes, color histograms and anchor face appearances, rely heavily on editorial rules...
    • ...TDT2 corpus are shown in Fig. 4. For comparison, F1measures of several state-of-the-art lexical-similaritybased approaches on the same corpus are shown together, including word-based and subword-based Text-Tiling [3,10], Lexical Chaining [4,11], LSA-based Text-Tiling [12, 13]...

    Zihan Liuet al. Laplacian Eigenmaps for automatic news story segmentation

    • ...Another group of works in this category depend on additional semantic knowledge extracted from dictionaries and thesauruses [9, 14] or from collocations collected in large corpora [2, 3, 12]...

    Raji R. Pillaiet al. Linear text segmentation using classification techniques

Sort by: