Tackling Data Sparseness in Recommendation using Social Media based Topic Hierarchy Modeling

Authors: Xingwei Zhu, Zhao-Yan Ming, Yu Hao, Xiaoyan Zhu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our evaluation on two real world datasets demonstrates the superiority of our method over state-of-the-art methods.
Researcher Affiliation Collaboration Xingwei Zhu1, Zhao-Yan Ming2 , Yu Hao1 and Xiaoyan Zhu1 1State Key Laboratory of Intelligent Technology and Systems, Tsinghua National Laboratory for Information Science and Technology, Department of Computer Sci. and Tech., Tsinghua University 2Department of Computer Science, Digi Pen Institute of Technology
Pseudocode No The paper describes the parameter estimation process verbally and with equations but does not include any pseudocode or algorithm blocks.
Open Source Code Yes Our datasets and codes are available at data.csaixyz.org/ijcai2015/ijcai_data.rar
Open Datasets Yes The first is the Movie Lens 1M dataset (Movie) 7. It contains one million ratings on 3, 706 movies produced by 6, 040 users. The second dataset we used is an i Tunes app rating dataset (App). It contains 88, 253 ratings on 1, 485 apps produced by 4, 483 users.
Dataset Splits No For the evaluation, we first split the user-item ratings into two parts, i.e., 80% for model training and 20% for test.
Hardware Specification No The paper does not specify the hardware used for experiments, such as specific CPU or GPU models.
Software Dependencies No Finally, we adopt My Media Lite 8 toolkit to implement both our THRec model and the baseline methods 9. The version number for My Media Lite is not specified.
Experiment Setup Yes The parameter L in the constraint is an adjustable hyper-parameter. ... we first initiate the weight of each topic based on its popularity as follows: wtm = p(tm) L (8) ... in which γ is the learning rate.