Ulster University Logo

Increasing Topic Coherence by Aggregating Topic Models

Blair, Stuart, Bi, Yaxin and Mulvenna, Maurice (2016) Increasing Topic Coherence by Aggregating Topic Models. In: Proceedings of the 9th International Conference on Knowledge Science, Engineering and Management (KSEM-2016). (Eds: Lehner, F and Fteimi, N), Springer International Publishing AG, Heidelberg, pp. 69-81. ISBN 978-3-319-47650-6 [Book section]

[img] Text - Published Version
Indefinitely restricted to Repository staff only.

355kB
[img] Text - Supplemental Material
Indefinitely restricted to Repository staff only.

33kB

DOI: 10.1007/978-3-319-47650-6_6

Abstract

In this paper, we introduce a novel method for aggregating multiple topic models to produce an aggregate model that contains top- ics with greater coherence than individual models. When generating a topic model a number of parameters must be specified. Depending on the parameters chosen the resulting topics can be very general or very specific. In this paper the process of aggregating multiple topic mod- els generated using different parameters is investigated; the hypothesis being that combining the general and specific topics can increase topic coherence. The aggregate model is created using cosine similarity and Jensen-Shannon divergence to combine topics which are above a sim- ilarity threshold. The model is evaluated using evaluation methods to calculate the coherence of topics in the base models against those of the aggregated model. The results presented in this paper show that the aggregated model outperforms standard topic models at a statistically significant level in terms of topic coherence when evaluated against an external corpus.

Item Type:Book section
Keywords:Topic models, Semantic coherence, Ensemble methods
Faculties and Schools:Faculty of Computing & Engineering
Faculty of Computing & Engineering > School of Computing and Mathematics
Research Institutes and Groups:Computer Science Research Institute
Computer Science Research Institute > Artificial Intelligence and Applications
ID Code:36092
Deposited By: Professor Maurice Mulvenna
Deposited On:28 Oct 2016 10:41
Last Modified:17 Oct 2017 16:26

Repository Staff Only: item control page