RTHN: A RNN-Transformer Hierarchical Network for Emotion Cause Extraction

Authors: Rui Xia, Mengran Zhang, Zixiang Ding

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We finally achieve the best performance among 12 compared systems and improve the F1 score of the state-of-the-art from 72.69% to 76.77%.
Researcher Affiliation Academia Rui Xia , Mengran Zhang and Zixiang Ding School of Computer Science and Engineering, Nanjing University of Science and Technology, China {rxia, zhangmengran, dingzixiang}@njust.edu.cn
Pseudocode No The paper describes the model architecture and mathematical formulations but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes 1The source code can be obtained at https://github.com/NUSTM/RTHN
Open Datasets Yes We evaluate our RTHN model on the benchmark ECE corpus [Gui et al., 2016a], which was the mostly used corpus for emotion cause extraction.
Dataset Splits Yes The same as [Gui et al., 2017], we randomly divide the data with the proportion of 9:1, with 9 folds as training data and remaining 1 fold as testing data. The following results are reported in terms of an average of 10-fold cross-validation.
Hardware Specification Yes In Table 3, we report their performance as well as the training time on a GTX-1080Ti GPU server.
Software Dependencies No The paper mentions using "word2vec toolkit" and "Adam optimizer" but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes The dimension of word embedding, RP embedding and GP embedding is set to be 200, 50 and 50, respectively. The hidden units of LSTM in word-level encoder is set to be 100. The dimension of the hidden states in Tranformer is 200, and the dimensions of query, key and value are 250, 250, and 200 repectively. The maximum numbers of words in each clause and clauses in each document are set to be 75 and 45, respectively. The network is trained based on the Adam optimizer with a mini-batch size 32 and a learning rate 0.005.