>
> |
Collocation Graphs in Three Dimensions
The Point
The idea is to provide a visualization of the vocabulary of a text and how the words are connected to each other. Can such a visualization allow someone to explore a text in an interesting way?
Description
- The top 200 high-frequency words would be displayed. The higher the frequency of a word the larger it would be.
- Each word would be linked by a weighted line to the other words that is collocates with
- For each word we would weigh their overall distance to all the other high-frequency words so as to be able to identify the central words. The arrangement of the HF words would radiate out from the centre with the well connected words in the centre
- We could calculate this for each section of the text and then have a timeline that allows one to see the graph change from one section to another
- If you double-click on a word you could get a KWIC as a menu and full text from there
- If you click on a word its collocates would be highlighted, including collocates not part of the list of HF words.
- If you click and drag you could see more collocates the further out from the word you drag. One could have rings of words - the words 5 words away, the words 10 words away ...
- Alternatively clicking and dragging could rearrange the graph so that the word being focused on is in the centre
- It would be interesting to find ways to show passages that include any two words - perhaps by rolling over the line between two collocating words one could see a KWIC of the passages with both
- It would be nice to animate the emergence of HF words from the beginning of the text
Possible Parameters and Issues to Try
- The frequency of the word could be represented by size or color or both
- The number of HF words to be displayed could be controlled
- We need a stop list and the opportunity to add words not ranked as HF
- We could add the ability to create clusters of words that are similar. Thus pattern and patterns could be merged
- We need to think about how we calculate collocation - distance in words?
- We need to think about how we calculate the weight of each word. Are there MVA techniques?
- What if we did all the words of a text - could we try this with poetry
- Can this work with the scrape of a web site or with spider results
Display and Interaction Issues
- How will this be set up to interact with? Will users walk up to it? Will they sit down?
- What sort of pointing device will they use?
- Can it be designed as an art installation?
- Can a visualization be exported?
Examples, Links and Similar Visualizations
-- GeoffreyRockwell - 31 Aug 2007
|
