Week 2
May 30, 2014
Literature review went well last week so I did a little more research and Dr. Bansal and I have narrowed down our project ideas to three. I wrote up a few mini app proposals with the format of: 1) the problem the app addresses 2) features and functionality 3) challenges we might face in development and 4) similar products.
- Educational Resource Aggregator
- Problem: With the increasing number of institutions offering open courseware, it's become necessary to have a way to compare similar courses and evaluate them based on differences in curriculum, time commitment, and student satisfaction.
- Functionality: Comparison of learning materials from several different websites.
- owl:sameAs allows for zero overlap in search results when sampling course descriptions from websites that offer the same course
- Other attributes such as a CS course's programming language can be added to filter results
- Course details can be compared in a single view
- Challenges:
- No guarantee of uniformity between attributes of data i.e. one course may have the programming language in their info box, another may not
- Little to no linked datasets exist for educational resources so different MOOC providers would need to be crawled.
- A small ontology might need to tailored to course attributes if LOM, SCORM or other ontologies aren't appropriate for MOOCs.
- Similar Products
- Crime Mapping
- Problem: Recent police reports are continually a subject of interest in newspapers and other media but blurbs of text are ineffective in determining the areas in which crimes occur most frequently and finding other similarities between incidents.
- Functionality:
- Visualization of police reports (by Google Maps API or CartoDB) would allow users to view incidents with iconography matching different categories of incidents (robberies, murders, etc).
- Filter incidents by category, date range, etc.
- Marker clustering to determine unsafe areas.
- Publishing of linked data would allow for a different lens on police data via SPARQL query. Depending on the robustness of the reports, a query could look for relationships between physical characteristics of a suspect, a general area, date range, etc. A GUI could be produced to make queries specific to this data.
- Challenges:
- Accessibility of police data
- NLP of many sources may be necessary depending on data availability
- Similar Products
- DBpedia Mobile
- Problem: DBpedia, despite being one of the premier data sets of the semantic web and having several applications tailored to it, is largely inaccessible to the general public.
- Functionality: Intuitive GUI for SPARQL queries to find relationships between items on DBpedia
- Challenges:
- In order to display information effectively, we'd need to consider how to graph data with regards to the screen size and limitations of the hardware
- May be too large in scope given the time constraints of the REU
- Similar Products
Given the amount of work that's been done in all of these areas, it's important that we make sure we can contribute something unique to the app's functionality using Linked Open Data rather than innovating solely in GUI or other features of the software. Our ability to do so and availability of data will influence which idea is chosen next week.
-Sebastian