A Hierarchical Framework for Relation Extraction with Reinforcement Learning
Authors: Ryuichi Takanobu, Tianyang Zhang, Jiexi Liu, Minlie Huang7072-7079
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our model was evaluated on public datasets collected via distant supervision, and results show that it gains better performance than existing methods and is more powerful for extracting overlapping relations. |
| Researcher Affiliation | Academia | Ryuichi Takanobu,1,3 Tianyang Zhang,1,3 Jiexi Liu,2,3 Minlie Huang1,3 1 Dept. of Computer Science & Technology, 2 Dept. of Physics, Tsinghua University, Beijing, China 3 Institute for Artificial Intelligence, Tsinghua University (THUAI), China 3 Beijing National Research Center for Information Science & Technology, China {gxly15, zhang-ty15, liujx15}@mails.tsinghua.edu.cn, aihuang@tsinghua.edu.cn |
| Pseudocode | Yes | Algorithm 1: Training Procedure of HRL |
| Open Source Code | Yes | Data and code are publicly available at: https://github.com/truthless11/HRL-RE. |
| Open Datasets | Yes | We evaluated our model on the New York Times corpus which is developed by distant supervision and contains noisy relations. The corpus has two versions: 1) The original version generated by aligning the raw data with Freebase relations (Riedel, Yao, and Mc Callum 2010); 2) A smaller version of which the test set was manually annotated (Hoffmann et al. 2011). |
| Dataset Splits | Yes | For each dataset, we randomly chose 0.5% data from the training set for validation. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'Glove vectors' and 'Bi-LSTM' but does not specify software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x). |
| Experiment Setup | Yes | All hyper-parameters are tuned on the validation set. The dimension of all vectors in Eq. (1), (2) and (6) is 300. The word vectors are initialized using Glove vectors (Pennington, Socher, and Manning 2014) and are updated during training. Both relation type vectors and entity tag vectors are initialized randomly. The learning rate is 4e 5, the mini-batch size is 16, α = 0.1 in Eq. (9), β = 0.9 in Eq. (5), and the discount factor γ = 0.95. |