This is the implementation of SECOVARC (The Sentence Encoder with COntextualized Vectors for Argument Reasoning Comprehension), for SemEval-2018 Task 12 - The Argument Reasoning Comprehension Task.
| Model | Dev Acc(%) | Test Acc(%) |
|---|---|---|
| Intra-attention (Habernal, et al., 2018) | 63.8 | 55.6 |
| Intra-attention w/context (Habernal, et al., 2018) | 63.7 | 56.0 |
| SECOVARC-last (w/o heuristics) | 70.1 | 55.9 |
| SECOVARC-last (w/ heuristics) | 70.6 | 55.4 |
| SECOVARC-max (w/o heuristics) | 68.0 | 59.1 |
| SECOVARC-max (w/ heuristics) | 68.4 | 59.2 |
- OS: Ubuntu 16.04 LTS (64bit)
- Language: Python 3.6.2
- Pytorch: 0.3.0
Please install the following library requirements specified in the requirements.txt first.
nltk==3.2.4
tensorboardX==1.0
torch==0.3.0
torchtext==0.2.1
As SECOVARC only accpets one warrant at a time, data manipulation is inevitable. By executing the preprocess.py, you can build a modified version of data (located in .data/arc/preprocessed).
python preprocces.py
python train.py --help
usage: train.py [-h] [--batch-size BATCH_SIZE] [--dropout DROPOUT]
[--epoch EPOCH] [--gpu GPU] [--hidden-size HIDDEN_SIZE]
[--heuristics] [--learning-rate LEARNING_RATE] [--model MODEL]
[--optim OPTIM] [--print-freq PRINT_FREQ] [--pooling POOLING]
[--word_dim WORD_DIM]
optional arguments:
-h, --help show this help message and exit
--batch-size BATCH_SIZE
--dropout DROPOUT
--epoch EPOCH
--gpu GPU
--hidden-size HIDDEN_SIZE
--heuristics
--learning-rate LEARNING_RATE
--model MODEL available: bow, lstm, cove
--optim OPTIM
--print-freq PRINT_FREQ
--pooling POOLING available: max, last, average, min
--word_dim WORD_DIM