Brede non-negative matrix factorization

DTU Informatics >>> Neuroinformatics > Services > Brede non-negative matrix factorization from string

Number of topics:  Document separator: Newline Two newlines

1 documents
0 terms:


(no results)

Result details

(no results)


This web service will perform topic mining with sentiment analysis, see the script (formatted).


  1. Type in texts: one in each line.
  2. (Set the number of topics if not you want it automatically determined)
  3. Press "Analyze" and wait.

The results

The results with the texts will be grouped in topics. The value in parentheses after a word or a text is the "load" of the word or the text on the topic.


  1. The topic mining is performed with non-negative matrix factorization.
  2. The sentiment analysis via the AFINN word list. The sentiment of a topic is found by summing sentiment of the individual texts weighted by the number of texts in the topic. The sentiment of each individual text is found by summing the sentiment strength of each word weighted by the number of words. The weighting is by the square root.
  3. A word list excludes common English words ("stopwords")
  4. The analysis will only work on up to a few hundreds short texts.
  5. The results may change each time you run the algorithm. This is due to random initializations and the issues in the factorization algorithms.
  6. The value shown after each topic, each word and each document is the load telling how important they are.
  7. Texts and words are assigned exclusively to one topic even if some of the texts are load partially on two or more topics.
  8. If the load of some texts or words are zero then they show up in the "not assigned" category.
  9. The example texts are taken as the first (few) line(s) in Wikipedia articles about companies and countries. If the topic mining works well it should separate country and company texts.


  1. Mining the posterior cingulate: Segregation between memory and pain components. Finn Årup Nielsen, Daniela Balslev, Lars Kai Hansen. NeuroImage, 27(3):520-532, 2005.
  2. A new ANEW: Evaluation of a word list for sentiment analysis in microblogs, Finn Årup Nielsen, Proceedings of the ESWC2011 Workshop on 'Making Sense of Microposts': Big things come in small packages, May 2011.

Finn Årup Nielsen, DTU Informatics — 0.077 seconds