reuters — Reuters news stories (1996-1997)
- Description
A bipartite network of news stories and the words they contain, from RCV1-v2, the version by Lewis, Yang, Rose and Li (2004) of the Reuters Corpus, Volume 1 (RCV1), an archive of English-language news stories written by Reuters journalists between August 1996 and August 1997. RCV1-v2 consists of 804,414 stories, which the Reuters editors had assigned to categories of three kinds: topics, industries and regions. For their text categorization experiments, Lewis et al. split them into 23,149 training stories (those of 20-31 August 1996) and 781,265 test stories, and published the tokenized stories: the words of each story, with the words of a large stop list (essentially all function words) removed, and the remaining ones replaced by their stems, in scrambled order, so that the texts of the stories cannot be reconstructed.
Each node is a story or a stem, and an edge connects each story to each stem it contains, weighted by the number of occurrences of the stem in the story. The network contains all stories of RCV1-v2, those of the training set (marked by
is_train) and of the test set. The 2,377 stories of the original RCV1 that are not in RCV1-v2 (because of invalid category codes), which the authors also published, are left out.The data are derived from the Reuters Corpus, Volume 1 (Reuters Ltd.); the authors of the data ask that publications using it give an attribution to Reuters.
The network has the following properties:
- (Node)
name: the stem (empty for stories). - (Node)
is_word: whether the node is a stem (otherwise a story). - (Node)
doc_id: the Reuters id of the story (-1 for stems). - (Node)
is_train: whether the story belongs to the training set of Lewis et al. (otherwise to the test set; false for stems). - (Node)
topics,industries,regions: the codes of the topic, industry and region categories assigned to the story in RCV1-v2 (e.g.CCAT,I81402,USA; empty for stems). - (Edge)
weight: the number of occurrences of the stem in the story.
- (Node)
- Tags
- Informational Language Bipartite Weighted Metadata
- Added
- 2020-08-11
- Citation
-
- D. D. Lewis, Y. Yang, T. G. Rose, F. Li, "RCV1: A New Benchmark Collection for Text Categorization Research." Journal of Machine Learning Research 5, 361-397 (2004). www.jmlr.org
- Upstream URL OK
-
https://www.jmlr.org/papers/v5/lewis04a.html
http://www.ai.mit.edu/projects/jmlr/papers/volume5/lewis04a/lyrl2004_rcv1v2_README.htm - Upstream license
- Available without license; attribution to Reuters requested, http://www.ai.mit.edu/projects/jmlr/papers/volume5/lewis04a/lyrl2004_rcv1v2_README.htm
- Networks
-
Tip: hover over a column header for a description.
Column legend
- $\left<k\right>$
- Mean degree
- $\sigma_k$
- Standard deviation of the degrees
- $\lambda_h$
- Largest eigenvalue of the Hashimoto (non-backtracking) matrix
- $\tau$
- Random walk mixing time
- $r$
- Degree assortativity
- $c$
- Global clustering coefficient
- $\oslash$
- (Pseudo-) diameter
- $S$
- Relative size of the largest component
- NPs
- Node properties
- EPs
- Edge properties
- Ridiculograms*
-