[Corpora-List] help with n-grams

William Fletcher fletcher at usna.edu
Mon Oct 27 18:03:33 CET 2008


Hello Marc,

My free Windows program kfNgram http://www.kwicfinder.com/kfNgram/ can generate character n-grams ("chargrams") in addition to word n-grams ("wordgrams"). You can specify length (n), position (initial, final, medial) etc. It was used to produce chargrams from the BNC at http://pie.usna.edu/explorec.html or http://phrasesinenglish.org/explorec.html

Let me know if you have any questions about how to use it on your data, e.g. how to define a custom character set for the phonemic data.

Regards, Bill Fletcher

-----Original Message----- From: corpora-bounces at uib.no [mailto:corpora-bounces at uib.no] On Behalf Of Marc FRYD Sent: Sunday, October 26, 2008 4:19 AM To: corpora at uib.no Subject: [Corpora-List] help with n-grams

Hi all, I wonder if anyone could help a linguist with moderate programming abilities with the following task. I am currently working on a corpus of aligned grapheme-to-phoneme isolated words. I would like to produce an N-gram parsing of both levels of data (the graphemic and the phonemic) with a view to extracting trends favouring realisations (i.e. this grapheme will realise as that phoneme with an x rate of occurrence if preceded/followed by such and such graphemes). The db is currently c3000 words, but it will keep growing. Cheers, Marc

-- Dr. Marc FRYD Senior Lecturer in English Linguistics

Faculté des Lettres et des Langues Université de Poitiers 95 avenue du Recteur Pineau 86022, Poitiers, France

Office: 05 49 45 48 11 Cell: 06 76 28 18 50

_______________________________________________ Corpora mailing list Corpora at uib.no http://mailman.uib.no/listinfo/corpora



More information about the Corpora mailing list