Start Date
12-13-2015
Description
We present a semantic similarity-based recommender service. Our experimental application and validation domain consists of K-12 engineering learning resources. Given a learning resource, we must determine which educational standards it addresses and vice versa, find resources that align with a given standard. One approach to this problem suggests transitively inferring standard alignment from the semantic similarity of other, previously aligned resources. We investigate a bigram-based similarity estimator and a Sammon map-based user interface for visualizing the resulting similarity space. Validation was performed using resources in TeachEngineering.org, a K-12 STEM digital library. Target classifications were derived from author-generated tables of content for these resources. Testing shows good performance of the similarity measure, both in its correspondence to the collection’s table of contents and in the form of a two-dimensional Sammon map. The results provide evidence for the feasibility and practicality of using automated similarity measures in standards alignment and similar problems.
Recommended Citation
Reitsma, Rene; Hsieh, Ping-Hung; and Robson, Robby, "Estimation and Visualization of Digital Library Content Similarities" (2015). ICIS 2015 Proceedings. 5.
https://aisel.aisnet.org/icis2015/proceedings/DecisionAnalytics/5
Estimation and Visualization of Digital Library Content Similarities
We present a semantic similarity-based recommender service. Our experimental application and validation domain consists of K-12 engineering learning resources. Given a learning resource, we must determine which educational standards it addresses and vice versa, find resources that align with a given standard. One approach to this problem suggests transitively inferring standard alignment from the semantic similarity of other, previously aligned resources. We investigate a bigram-based similarity estimator and a Sammon map-based user interface for visualizing the resulting similarity space. Validation was performed using resources in TeachEngineering.org, a K-12 STEM digital library. Target classifications were derived from author-generated tables of content for these resources. Testing shows good performance of the similarity measure, both in its correspondence to the collection’s table of contents and in the form of a two-dimensional Sammon map. The results provide evidence for the feasibility and practicality of using automated similarity measures in standards alignment and similar problems.