In this project, we constructed an Association Network from Wikipedia and a psychological process called "free association", and then utilized this network to solve various real-world applications. An application in which we have already achieved good results is the "semantic relatedness computation" task, where we proposed algorithms based on the constructed network to compute the semantic relatedness between terms and short texts, both reaching state-of-the-art performance on standard benchmark data sets. This work has been accepted as a full paper by AAAI 2015. Our current focus is utilizing this "association network" to improve both the accuracy and serendipity of Recommender System, by introducing into it semantic knowledge provided in the network.
- An Association Network for Computing Semantic Relatedness. Keyang Zhang, Kenny Q. Zhu, and Seung-Won Hwang. In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2015), Austin, TX, January 2015.
Our Association Network can be downloaded here. Please fill in the form before downloading the data.
Our online demo is now available here. Due to limited memory resources, currently this demo only employs a subgraph of the association network.
The proof of Lemma 1 in An Association Network for Computing Semantic Relatedness is available here.
- Kenny Q. Zhu, Distinguished Research Professor at SJTU, email@example.com
- Keyang Zhang, Undergraduate at SJTU, firstname.lastname@example.org