Jump to content

Talk:Named-entity recognition

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

[Untitled]

[edit]

another commonly used name is entity extraction Josh Froelich 02:53, 11 December 2006 (UTC)[reply]

I find that the definition of the term is far fetched compared to its literal representation. "named" "entity" "recognition" means to me what Entity Linking is. What the article describe is that entities are annotated with one of its hypernym. i⋅am⋅amz3 (talk)

Merging with Entity Linking

[edit]

What about merging this article with Entity Linking? i⋅am⋅amz3 (talk)

NER is not a ML specific Task

[edit]

The article speak too much about the ML / statistical approach to NER. They are also dictionary and rule based approaches. i⋅am⋅amz3 (talk)

Classifying software in other ways

[edit]

I'd think there would be more interesting ways to classify NER software than by license type and code access (open, dual, proprietary). How about the types of entity classified? Or the methods used? Of course, the license type is interesting and worth mentioning, but I don't see that it makes sense as the top-level division. --Macrakis (talk) 14:15, 2 January 2008 (UTC)[reply]

I'd like to see a list by type of algorithm. As far as I can tell, that means:
  • Rules and regular expressions
  • Dictionary lookup
  • Linguistic analysis, NLP, etc.
  • Statistical, Bayesian, SVM, etc.
Does anyone do more than two of those?

--Searchtools (talk) 23:33, 24 August 2009 (UTC)[reply]

XML Tags in Examples

[edit]

The XML tags (<ENAMEX> etc.) in the examples, especially the first one, serve more to confuse than to clarify. First, not everyone is familiar with the notation. Second, they add lots of visual noise. Finally, this seems a lot like an implementation detail. Surely we can do better with text formatting? — Preceding unsigned comment added by 94.245.127.69 (talk) 14:22, 18 October 2012 (UTC)[reply]

I just fixed this. The XML was quite distracting. QVVERTYVS (hm?) 12:21, 9 March 2014 (UTC)[reply]

Evaluation

[edit]

Is it worth including a para on eval of NER systems? Covering e.g. strict vs lenient matching, and precision & recall & fscore (w/links to main articles) Leondz (talk) 12:32, 19 March 2014 (UTC)[reply]

I just added a short paragraph on evaluation. Feel free to jump in; it focuses on F1 score for strict matching, and it needs references. QVVERTYVS (hm?) 14:27, 19 March 2014 (UTC)[reply]