CompareLDA: A Topic Model for Document Comparison
Authors: Maksim Tkachenko, Hady W. Lauw7112-7119
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Evaluation on several public datasets underscores the strengths of Compare LDA in modelling document comparisons. and Our experimental objective is to validate the efficacy of Compare LDA in deriving topics that are well-aligned to document comparisons. |
| Researcher Affiliation | Academia | Maksim Tkachenko, Hady W. Lauw School of Information Systems Singapore Management University maksim.tkatchenko@gmail.com, hadywlauw@smu.edu.sg |
| Pseudocode | No | The paper describes the generative process and model fitting steps in paragraph form, but does not include a clearly labeled 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | No | The paper only references the implementation used for a baseline model ('We used the following implementation: https://github.com/ vietansegan/segan/'). It does not provide any statement or link for the open-source code of their proposed method, Compare LDA. |
| Open Datasets | Yes | For experiments, we rely on public text corpora... Wikipedia The first is a set of three datasets constructed from Wikipedia1 pages... The second dataset is from Amazon as described in (Mc Auley, Pandey, and Leskovec 2015; Mc Auley et al. 2015). ... The third dataset contains movie reviews (Pang and Lee 2005). |
| Dataset Splits | No | Each dataset is split into training and testing folds in 80:20 proportion respectively. The paper does not explicitly state a separate validation split. |
| Hardware Specification | No | The paper does not provide any specific hardware details (such as GPU or CPU models, or cloud computing specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions using a specific implementation for a baseline model ('https://github.com/ vietansegan/segan/') but does not provide any specific version numbers for software dependencies or libraries used in their own methodology or experiments. |
| Experiment Setup | No | The paper describes the model fitting procedure and mentions varying the number of topics (default is 80), but it does not specify concrete experimental setup details such as learning rates, batch sizes, or optimizer settings typically required for full reproducibility. |