Global Belief Recursive Neural Networks
Authors: Romain Paulus, Richard Socher, Christopher D. Manning
NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We present a qualitative and quantitative analysis of the GB-RNN on a contextual sentiment classification task. The main dataset is provided by the Sem Eval 2013, Task 2 competition [17]. We outperform the winners of the 2013 challenge, as well as several baseline and model ablations. |
| Researcher Affiliation | Collaboration | Romain Paulus, Richard Socher Meta Mind Palo Alto, CA {romain,richard}@metamind.io Christopher D. Manning Stanford University 353 Serra Mall Stanford, CA 94305 manning@stanford.edu |
| Pseudocode | No | The paper describes the mathematical formulations and steps of the model but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository for the described methodology. |
| Open Datasets | Yes | The Sem Eval competition dataset is composed of tweets labeled for 3 different sentiment classes: positive, neutral and negative. The tweets in this dataset were split into a train (7862 labeled phrases), development (7862) and development-test (7862) set. The final test set is composed of 10681 examples. The main dataset is provided by the Sem Eval 2013, Task 2 competition [17]. |
| Dataset Splits | Yes | The tweets in this dataset were split into a train (7862 labeled phrases), development (7862) and development-test (7862) set. The best models were selected by cross-validating on the dev set for several hyper-parameters (word vectors dimension, hidden node vector dimension, number of training epochs, regularization parameters, activation function, training batch size and dropout probability) and we kept the models with the highest cross-validation accuracy. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU or CPU models used for running the experiments. |
| Software Dependencies | No | The paper mentions software like the 'Stanford Parser' and 'Ada Grad' but does not specify their version numbers or other key software dependencies with version information. |
| Experiment Setup | Yes | Optimal performance for the single best GB-RNN was achieved by using vector sizes of 130 dimensions (100 pre-trained, fixed word vectors and 30 trained on sentiment data), a mini-batch size of 30, dropout with p = 0.5 and sigmoid non-linearity. |