VAND Multi: Feature Overview - PMeschenmoser/Visual-Analyzer-for-News-Dependencies GitHub Wiki
VAND Multi is a research-oriented visualisation system which aims at parameter evaluation and optimization for segment-based detection of information reuse.
- Article Import
- Client-side import
- Offer server-sided sets
- Highly parameterisable file parsers for articles in .txt and .JSON formats
- Visualization
- 1:n article view to fastly detect information reuse from different sources, taking temporal flow into account
- A main article is given on the right. Segments for n articles are given in a compressed presentation on the left view.
- Scalable pixel visualisation for statistical comparison, e.g. by encoding segment permutation matrices
- Linking graph area to identify major flow directions
- Further encoding of tf-idf similarity or 'copy ratio', using border thickness
- Arrangement onto self-adapting time scales
- Vertical ordering according to step function which aims at avoiding overlap layout
- Focus+Context visualisations for effective navigation in visualisation space
- For a more detailed analysis: 1:1 article comparison view
- Segment filtering by feature ordering, count, type,...
- etc.
* Permutation matrices are used to illustrate relative segment orders in a scalable view. For this, segments of a reference article are ordered according to (a) their position in the reference article and (b) to their counterparts' position in the main article. Two lists are resulting. We iterate over the lists and check whether we have segments of the same ID at a particular list index (-> correct relative order). Correct relative orders are visualized by small green squares, permutations are given by lightblue squares. To make permutation matrices comparable between multiple reference articles, we order these squares from right to left, and from bottom to top.
- Code
- Easy setup, as project dependencies rely at most on node modules
- Central Error Handling
- Reusable and Modularised OOP Javascript