A Tri-Role Topic Model for Domain-Specific Question Answering

Authors: Zongyang Ma, Aixin Sun, Quan Yuan, Gao Cong

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental As a case study, we conducted experiments on ranking answers for questions on Stack Overflow... Experimental results show that TRTM is effective in facilitating users getting ideal rankings of answers, particularly for new and less popular questions. Evaluated on n DCG, TRTM outperforms state-of-the-art methods.
Researcher Affiliation Academia School of Computer Engineering, Nanyang Technological University, Singapore 639798 {zma4, qyuan1}@e.ntu.edu.sg {axsun, gaocong}@ntu.edu.sg
Pseudocode Yes The graphical representation of the TRTM model is shown in Figure 2 and the generative process is summarized as follows: For each question q Qu ... The Inference Algorithm of TRTM Given question set Q composed by U and answer set A by answerers from V , we obtain the likelihood of the data in Equation 1. The exact inference of Equation 1 is intractable. We propose an Expectation-Maximization (EM) algorithm to appropriately infer TRTM. The EM algorithm has two steps: E-step and M-step.
Open Source Code No The paper provides a link to a Stack Overflow data dump (http://blog.stackoverflow.com/category/cc-wiki-dump/), but no explicit statement or link to the authors' own open-source code for the methodology was found.
Open Datasets Yes We evaluate the proposed TRTM model on Stack Overflow data1. 1http://blog.stackoverflow.com/category/cc-wiki-dump/. Questions and answers from Stack Overflow posted between Jan 01, 2011 and Mar 31, 2011 are used as training data; questions and answers published from Apr 01, 2011 to Sep 06, 2013 are used as test data.
Dataset Splits No Questions and answers from Stack Overflow posted between Jan 01, 2011 and Mar 31, 2011 are used as training data; questions and answers published from Apr 01, 2011 to Sep 06, 2013 are used as test data. (No explicit mention of a separate validation set for hyperparameter tuning or model selection was found, although hyperparameters were set experimentally.)
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts, or cloud instance types) used for running experiments were mentioned.
Software Dependencies No No specific software dependencies with version numbers (e.g., library or solver names with versions) were mentioned.
Experiment Setup Yes We experimentally set the hyperparameters of TRTM: α = 100, β = 100, τ = 100. We evaluated different number of topics |Z| = 10, 20, and 40.