Sentiment induction for attention and lexicon regularized neural sentiment analysis

• Autores: Lingxian Bao
• Directores de la Tesis: Patrik Lambert (codir. tes.), Toni Badia Cardús (codir. tes.)
• Lectura: En la Universitat Pompeu Fabra ( España ) en 2021
• Idioma: español
• Tribunal Calificador de la Tesis: Leo Wanner (presid.), Rodrigo Agerri Gascón (secret.), Montserrat Cuadros Oller (voc.)
• Programa de doctorado: Programa de Doctorado en Traducción y Ciencias del Lenguaje por la Universidad Pompeu Fabra
• Materias:
  • Lingüística
    • Lingüística aplicada
      • Lingüística informatizada
• Enlaces
  • Tesis en acceso abierto en: TDX
• Resumen

  • Deep neural networks as an end-to-end approach lacks flexibility and robustness from an application point of view, as one cannot easily adjust the network to fix an obvious problem, especially when new training data is not available: e.g. when the model consistently predicts `positive` when seeing the word "disappointed". Meanwhile, it is less stressed that the attention mechanism is likely to "over-focus" on particular parts of a sentence, while ignoring positions which provide key information for judging the polarity.

    In this thesis, we describe a simple yet effective approach to leverage lexicon information so that the model becomes more flexible and robust. We also explore the effect of regularizing attention vectors to allow the network to have a broader "focus" on the input sequence. Moreover, we try to further improve the proposed lexicon enhanced neural sentiment analysis system by applying sentiment domain adaptation.

    This thiesis is made of 7 chapters:

    1. Introduction: overview of the thesis content.

    2. Literature Review: review of the filed sentiment analysis and related works.

    3. Objective: core research questions of this thesis.

    4. Theoretical Framework: brief review of the essential knowledge () to understand this thesis.