Modelling Form-Meaning Systematicity with Linguistic and Visual Features
Authors: Arie Soeteman, Dario Gutierrez, Elia Bruni, Ekaterina Shutova8870-8877
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimenting on the English lexicon, we find that the inclusion of visual features allows us to identify more form-meaning systematicity than when using a textbased model alone. |
| Researcher Affiliation | Collaboration | Arie Soeteman University of Amsterdam awsoeteman@hotmail.com Dario Gutierrez IBM Research elkin.gutierrez@ibm.com Elia Bruni University of Amsterdam e.bruni@uva.nl Ekaterina Shutova University of Amsterdam e.shutova@uva.nl |
| Pseudocode | No | The paper describes the methodology using mathematical equations and diagrams (e.g., Figure 1, Figure 2) but does not include any pseudocode or explicitly labeled algorithm blocks. |
| Open Source Code | No | The paper mentions a third-party model's availability ('The model is freely available as part of the Word2Vec system release1, along with 300-dimensional vector representations for 3 million words and phrases.'), but does not state that the authors' own implementation code for their methodology is open source or publicly provided. |
| Open Datasets | Yes | We use the same lexicon as Gutierrez et al. (2016), which has been constructed by cross-referencing monomorphemic English words in the CELEX lexical database (Baayen, Piepenbrock, and Gulikers 1996) with monomorphemic words in the Oxford English Dictionary Online (Simpson, Weiner, and others 1989).; We used skip-gram with negative sampling (Mikolov et al. 2013) trained on the Google News dataset as our text-based model.; a deep convolutional neural network that was trained on the Image Net classification task (Russakovsky et al. 2015). |
| Dataset Splits | No | The paper mentions training a neural network 'on the entire lexicon' for one of its multimodal models, but it does not specify any explicit train/validation/test dataset splits with percentages or counts for its primary experiments. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory, or cloud instances) used to run the experiments. |
| Software Dependencies | No | The paper mentions software tools like 'Word2Vec system release' and 'Caffe deep learning framework' but does not specify their version numbers or other software dependencies with specific versions required for reproducibility. |
| Experiment Setup | Yes | We initialize edit weights to 1 and optimize them by minimizing the MSE until convergence.; The optimal weighting factor for our two monomodal models was 0.75.; The first three layers of both branches are fully connected Re LU layers.; We train the network on the entire lexicon using MSE as the loss function. |