Latent Dirichlet Markov allocation for sentiment analysis

Ayoub Bagheri, Mohamad Saraee, Franciska M.G. de Jong

    Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

    146 Downloads (Pure)


    In recent years probabilistic topic models have gained tremendous attention in data mining and natural language processing research areas. In the field of information retrieval for text mining, a variety of probabilistic topic models have been used to analyse content of documents. A topic model is a generative model for documents, it specifies a probabilistic procedure by which documents can be generated. All topic models share the idea that documents are mixture of topics, where a topic is a probability distribution over words. In this paper we describe Latent Dirichlet Markov Allocation Model (LDMA), a new generative probabilistic topic model, based on Latent Dirichlet Allocation (LDA) and Hidden Markov Model (HMM), which emphasizes on extracting multi-word topics from text data. LDMA is a four-level hierarchical Bayesian model where topics are associated with documents, words are associated with topics and topics in the model can be presented with single- or multi-word terms. To evaluate performance of LDMA, we report results in the field of aspect detection in sentiment analysis, comparing to the basic LDA model.
    Original languageUndefined
    Title of host publicationThe Fifth European Conference on Intelligent Management Systems in Operations (IMSIO 5)
    Place of PublicationSalford, UK
    PublisherThinkLab, University of Salford
    Number of pages6
    ISBN (Print)0903440547
    Publication statusPublished - Jul 2013

    Publication series

    PublisherThinkLab, University of Salford


    • EWI-23676
    • METIS-300002
    • IR-87440

    Cite this

    Bagheri, A., Saraee, M., & de Jong, F. M. G. (2013). Latent Dirichlet Markov allocation for sentiment analysis. In The Fifth European Conference on Intelligent Management Systems in Operations (IMSIO 5) (pp. 90-96). Salford, UK: ThinkLab, University of Salford.