Neural Generative Question Answering

Authors: Jun Yin, Xin Jiang, Zhengdong Lu, Lifeng Shang, Hang Li, Xiaoming Li

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical study shows the proposed model can effectively deal with the variations of questions and answers, and generate right and natural answers by referring to the facts in the knowledge-base. The experiment on question answering demonstrates that the proposed model can outperform an embedding-based QA model as well as a neural dialogue model trained on the same data.
Researcher Affiliation Collaboration 1School of Electronic Engineering and Computer Science, Peking University 2Noah s Ark Lab, Huawei Technologies 3Collaborative Innovation Center of High Performance Computing, NUDT, Changsha, China
Pseudocode No The paper describes the model architecture and training process in text and mathematical formulas, but it does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes The data is publicly available online3. 3https://github.com/jxfeb/Generative QA
Open Datasets Yes To facilitate research on the task of generative QA, we create a new dataset by collecting data from the web. ... The data is publicly available online3. 3https://github.com/jxfeb/Generative QA
Dataset Splits No The data is further randomly partitioned into training dataset and test dataset by using triple as the partition key. ... Table 3 shows some statistics of the datasets.
Hardware Specification Yes Our models are trained on an NVIDIA Tesla K40 GPU using Theano [Bastien et al., 2012; Bergstra et al., 2010], with the mini-batch size of 80.
Software Dependencies No The paper mentions using "Jieba Chinese word segmentor" and "Theano" but does not provide specific version numbers for these software components.
Experiment Setup Yes The dimensions of the hidden states of encoder and decoder are both set to 500, and the dimension of the word-embedding is set to 300. Our models are trained on an NVIDIA Tesla K40 GPU using Theano [Bastien et al., 2012; Bergstra et al., 2010], with the mini-batch size of 80. The training of each model takes about two or three days.