This research introduces a novel model architecture designed to tackle the challenges of short-text topic modelling on corpora of open-ended survey questions. The proposed model combines variational autoencoder (VAE) and negative multinomial sampling techniques to effectively perform topic modelling on short texts, providing a unique and insightful understanding of customers' opinions and preferences. The VAE helps manage the high dimensionality and sparsity of short texts, while the negative multinomial sampling enables the generation of unique topics from the learned latent space. The ultimate goal is to leverage this knowledge to improve customer segmentation and guide strategic decision-making within the organization. The model is compared to other methodologies, such as frequentist, probabilistic, and neural approaches. These comparisons help to identify the strengths and weaknesses of each approach, providing valuable insights for future research and applications.
This research introduces a novel model architecture designed to tackle the challenges of short-text topic modelling on corpora of open-ended survey questions. The proposed model combines variational autoencoder (VAE) and negative multinomial sampling techniques to effectively perform topic modelling on short texts, providing a unique and insightful understanding of customers' opinions and preferences. The VAE helps manage the high dimensionality and sparsity of short texts, while the negative multinomial sampling enables the generation of unique topics from the learned latent space. The ultimate goal is to leverage this knowledge to improve customer segmentation and guide strategic decision-making within the organization. The model is compared to other methodologies, such as frequentist, probabilistic, and neural approaches. These comparisons help to identify the strengths and weaknesses of each approach, providing valuable insights for future research and applications.
Neural Topic Modelling of Short Text Survey Responses in Italian Market Research
CASARELLA, CECILIA
2023/2024
Abstract
This research introduces a novel model architecture designed to tackle the challenges of short-text topic modelling on corpora of open-ended survey questions. The proposed model combines variational autoencoder (VAE) and negative multinomial sampling techniques to effectively perform topic modelling on short texts, providing a unique and insightful understanding of customers' opinions and preferences. The VAE helps manage the high dimensionality and sparsity of short texts, while the negative multinomial sampling enables the generation of unique topics from the learned latent space. The ultimate goal is to leverage this knowledge to improve customer segmentation and guide strategic decision-making within the organization. The model is compared to other methodologies, such as frequentist, probabilistic, and neural approaches. These comparisons help to identify the strengths and weaknesses of each approach, providing valuable insights for future research and applications.File | Dimensione | Formato | |
---|---|---|---|
CasarellaCeciliaTesiDataScienceFinalA.pdf
accesso riservato
Dimensione
32.37 MB
Formato
Adobe PDF
|
32.37 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/64787