Modelling Semantic Categories Using Conceptual Neighborhood
Authors: Zied Bouraoui, Jose Camacho-Collados, Luis Espinosa-Anke, Steven Schockaert7448-7455
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our main results for the category induction task are summarized in Table 2. In this table, we show results for different choices of the number of selected conceptual neighbors k, ranging from 1 to 5. As can be seen from the table, our approach substantially outperforms all baselines, with Multi being the most competitive baseline. |
| Researcher Affiliation | Academia | Zied Bouraoui CRIL U. Artois CNRS zied.bouraoui@cril.fr Jose Camacho-Collados Cardiff University, UK camachocolladosj@cardiff.ac.uk Luis Espinosa-Anke Cardiff University, UK espinosa-ankel@cardiff.ac.uk Steven Schockaert Cardiff University, UK schockaerts1@cardiff.ac.uk |
| Pseudocode | No | The paper describes procedures and steps for its model (e.g., 'our approach involves the following three steps'), but these are presented in natural language paragraphs rather than structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions that 'We used the implementation available at https://github.com/ huggingface/pytorch-pretrained-BERT' for the pre-trained BERT model, which is a third-party tool. However, it does not provide any link or explicit statement about releasing the source code for the methodology developed in this paper. |
| Open Datasets | Yes | To this end, we used the NASARI vectors4, which have been learned from Wikipedia and are already linked to Babel Net (Camacho-Collados, Pilehvar, and Navigli 2016). 4Downloaded from http://lcl.uniroma1.it/nasari/. ... As the text corpus to extract sentences for category pairs we used the English Wikipedia. In particular, we used the dump of November 2014, for which a disambiguated version is available online5. 5Available at http://lcl.uniroma1.it/sw2v. |
| Dataset Splits | Yes | For each of these categories, we split the set of known instances into 90% for training and 10% for testing. To tune the prior probability λA for these categories, we hold out 10% from the training set as a validation set. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used for running the experiments, such as GPU models, CPU specifications, or memory. |
| Software Dependencies | No | The paper mentions using '300-dimensional pretrained GloVe word embeddings (Pennington, Socher, and Manning 2014)' and 'pre-trained 768-dimensional BERT-base model (Devlin et al. 2019)'. While specific models are named, no version numbers for underlying software dependencies like Python, deep learning frameworks (e.g., PyTorch, TensorFlow), or other libraries (e.g., scikit-learn for SVM) are provided. |
| Experiment Setup | No | The paper describes high-level model components (Gaussian Classifier, GLR Classifier, Multinomial Logistic Regression) and some data processing parameters (e.g., 'maximum window size of 10 tokens', 'k categories'), but it does not specify concrete hyperparameters like learning rates, batch sizes, number of epochs, or optimizer settings for training any of the models. |