Topic segmentation with an aspect hidden Markov model(15)

We present a novel probabilistic method for partially unsupervised topic segmentation on unstructured text. Previous approaches to this problem utilize the hidden Markov model framework (HMM). The HMM treats a document as mutually independent sets of words

12REFERENCESexploreatemporalanalysisofourdataandmodellongtermtopicshiftsinthehiddenfactorsandlanguagemodels.

References

[1]DougBeeferman,AdamBerger,andJohnLafferty.Statisticalmodelsfortext

segmentation.MachineLearning,1999.

[2]A.P.Dempster,ird,andD.B.Rubin.Maximumlikelihoodfromincom-

pletedataviatheemalgorithm.JournaloftheRoyalStatisticalSociety,SeriesB(Methodological),39(1):1–38,1977.

[3]DanielGildeaandThomasHofmann.Topic-basedlanguagemodelsusingem.

EuroSpeech-99,pages2167–2170,1999.

[4]MartiA.Hearst.Contextandstructureinautomatedfull-textinformationaccess.

puterScienceDivisionTech-nicalReport,1994.

[5]ThomasHofmann.Probabilisticlatentsemanticindexing.Proceedingsofthe

Twenty-SecondAnnualInternationalSIGIRConferenceonResearchandDevel-opmentinInformationRetrieval,1999.

[6]P.vanMulbregt,I.Carp,L.Gillick,S.Lowe,andJ.Yamron.Textsegmentation

andtopictrackingonbroadcastnewsviaahiddenmarkovmodelapproach.1998.

[7]AndrewJ.Viterbi.Errorboundsforconvolutionalcodesandanasymptotically

optimaldecodingalgorithm.IEEETransactionsonInformationTheory,13:260–269,1967.