Each of the approximately 17,000 documents are represented in an 8-dimensional vector space (eight is the number of topics) where the value of the first coordinate point is the percentage share of document of the first topic, the second coordinate point is the percentage share of document of the second topic, etc… . We chose eight number of topics through trial and error. These eight number of topics resulted in the best interpretable themes of the texts.
The 8-dimensional vector representation of each of the documents is projected onto a Cartesian plane (via Multidimensional Scaling) in such a way as to preserve, as much as possible, the distances between pairs of documents in the 8-dimensional space. To view the document projections from the DEEDS corpus and corresponding metadata, click below:
Using the document ID number, search below for the document's topic mixture and its metadata information
Example: 580081
Search Result:
Topic Mixture:
* 'Internal' indicates date is included in the text. .. 'External' indicates date is provided by editor based on textual context.