Tasks
In the OPAL project, we’re working with DCAT datasets in RDF format (see formats). This is the basis of the hackathon. There are ideas for components that could be used in the project.
Metadata + JavaScript
Visualization: Show the related locations of datasets
Datasets may contain spatial information (geo coordinates). The spatial information could be visualized as a map like at OpenDataMonitor. For this you can use a JavaScript library, framework or API. Some examples are listed in this article.
An example to query spatial data is given in the following code. Note that the returned results have different types and the query has to be refined. You can execute the query a SPARQL endpoint listed at data.
PREFIX dcat: <http://www.w3.org/ns/dcat#>
PREFIX dct: <http://purl.org/dc/terms/>
SELECT * WHERE {
?dataset a dcat:Dataset .
?dataset dct:spatial ?spatial .
}
LIMIT 100
OFFSET 0
Spatial slicing: Allow users to draw a bounding box
Another task is to develop a spatial slicing component. In the available data, there are geo-coordinates for individual datasets. The goal of the spatial slicing task is to get those datasets, which are related to a selected region. To let users define a region, a JavaScript component is required, which allows users to draw a bounding box (rectangle or polygon) and therefore to specify a region. Returning the geo-coordinates of a selected bounding box is the main objective of this task. Afterwards, the bounding box coordinates can be delivered to another component (probably Elasticsearch). Examples can be found at OpenStreetMap Wiki: Bounding Box.
Metadata + Data Science approach
Maybe you already played around with some scientific approaches. That could for instance be a topic extraction or a statistical analysis. You are free to apply those approaches to the available metadata records.
An idea is to first create a set of titles, descriptions and categories of existing datasets. As not for all datasets a category (called theme in DCAT) is set, this could become a supervised (machine learning) classification task. Use the complete datasets to create a model containing used words on the one hand and categories on the other hand. Based on comparisons of word vectors, categories can be set for datasets without categories.
Note that some approaches already have been implemented, e.g. automatic quality assessment, language detection and geodata extraction, geographical topic extraction and data slicing. You can also improve the implementations.
Metadata + Metadata + Data + Data
If you already found multiple datasets, that are related to each other, you could think about combining them into one dataset. This could be two datasets concerning the same topic, e.g. the field of traffic or the same time period or the same city. Feel free to work on combining such datasets. For instance, extract and combine parts of Open Data Paderborn. But before publishing created data, check the underlying licenses!