Weakly Supervised Multi-task Learning for Semantic Parsing

Authors: Bo Shao, Yeyun Gong, Junwei Bao, Jianshu Ji, Guihong Cao, Xiaola Lin, Nan Duan

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on a newly constructed dataset and Complex Web Questions show that our proposed method outperforms state-of-the-art methods which demonstrates the effectiveness and robustness of our method.
Researcher Affiliation Collaboration Bo Shao1,2 , Yeyun Gong2 , Junwei Bao2 , Jianshu Ji3 , Guihong Cao3 , Xiaola Lin1 and Nan Duan2 1School of Data and Computer Science, Sun Yat-sen University 2Microsoft Research Asia 3Microsoft AI and Research, Redmond WA, USA
Pseudocode No The paper describes its models and methods using mathematical equations and textual descriptions, but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes 3) We construct a relatively large scale semantic parsing dataset to advance the semantic parsing research. We will release the dataset together with our code1. 1https://github.com/shaoboly/wsmtl
Open Datasets Yes In this paper we construct a large scale semantic parsing dataset which contains more than 50,000 <question, logical form> pairs over 9 types of questions.
Dataset Splits No The paper does not explicitly state the training, validation, and test dataset splits with percentages, sample counts, or specific predefined split citations.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory used for running the experiments.
Software Dependencies No The paper mentions using 'Glove word embeddings' and models based on 'GRU-RNN' and 'attention mechanism', but does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes We set the dropout rate to 0.5. The dimension of all hidden vectors and word embedding is 300. Word vocabulary and embedding are not shared between encoder and decoder. In all our experiment, α and β are set to 0.5, λ1 and λ2 are set to 0.1.