A Human-Like Semantic Cognition Network for Aspect-Level Sentiment Classification
Authors: Zeyang Lei, Yujiu Yang, Min Yang, Wei Zhao, Jun Guo, Yi Liu6650-6657
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To verify the effectiveness of our approach, we conduct extensive experiments on three widely used datasets. The experiments demonstrate that HSCN achieves impressive results compared to other strong competitors. |
| Researcher Affiliation | Academia | 1Graduate School at Shenzhen, Tsinghua University, 2Shenzhen Institutes of Advanced Technology, CAS, 3Technische Universitt Darmstadt, 4TBSI , Tsinghua Universitys, 5Peking University Shenzhen Institute |
| Pseudocode | No | The paper describes the model with mathematical equations and textual explanations, but no explicit pseudocode or algorithm blocks labeled 'Pseudocode' or 'Algorithm' were found. |
| Open Source Code | No | The paper does not contain any explicit statement about releasing source code for the methodology described, nor does it provide a link to a code repository. |
| Open Datasets | Yes | Sem Eval-14 This data is constructed for the aspect based sentiment analysis task (Sem Eval-2014 Task 4)2. Two domain-specific datasets for restaurants (S-res.) and laptops (S-laptop) have been provided for training and testing. Table 1 shows the statistics of these two datasets. ... The original dataset is a collection of tweets from Twitter by (Dong et al. 2014), using keywords (e.g., bill gates , google ) to query the Twitter API. Footnote 2 provides 'http://alt.qcri.org/semeval2014/'. |
| Dataset Splits | No | Table 1 shows statistics for S-res. (train), S-res. (test), S-laptop (train), S-laptop (test). The paper also states: 'The training data consists of 6,248 tweets, and the testing data has 692 tweets.' There is no explicit mention of a validation set split or its use in the paper. |
| Hardware Specification | No | The paper mentions running experiments 'on the same CPU server' but does not provide specific hardware details such as CPU model, GPU model, or memory specifications. |
| Software Dependencies | No | The paper mentions using '300-dimensional GloVe vectors', 'GRU networks', 'RMSprop optimization algorithm', 'dropout strategy', and 'label smoothing technique', but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | Model hyper-parameters are set by a grid search. ... We use 300-dimensional Glo Ve vectors to initialize the word embeddings for words in the context and target words, and all out-of-vocabulary words are initialized by sampling from the uniform distribution U(0.25,0.25). We initialize all the weight matrices as random orthogonal matrices, and all the bias vectors are initialized to zero. The dimension of GRU hidden states is set as 10. We conduct mini-batch (with size 40) training using RMSprop optimization algorithm to train the model. The dropout rate is set to 0.5, and the coefficient λ of L2 normalization is set to 10^-5. Label smoothing coefficient ϵ is set as 0.01. |