New anal­y­sis pin­points what makes a wine unique

Weekend Post (South Africa) - - NEWS -

GRASS may re­mind some of cer­tain pas­tures. Vanilla and cus­tard sound like pud­ding in­gre­di­ents for a cold day. But while these aro­mas are not typ­i­cally as­so­ci­ated with drink­ing wine‚ a study has found that ex­perts man­aged to use these de­scrip­tions among a to­tal of 266 and 250 dif­fer­ent aroma sen­sory words for South African Chenin blanc and Sauvi­gnon blanc wines re­spec­tively.

The study ex­plored sen­sory de­scrip­tions of Chenin wines with ma­chine learn­ing tech­niques.

Col­lec­tively‚ the wine va­ri­etals Chenin blanc‚ Sauvi­gnon blanc and Chardon­nay hold the largest share (per vol­umes pro­duced) in the lo­cal white wine cat­e­gory.

In this study‚ de­scribed as the first of its kind‚ re­searchers em­ployed tech­niques that are usu­ally used to an­a­lyse “big data” and pin­point com­puter hack­ers‚ among oth­ers.

They used sen­sory de­scrip­tors of 7 000 lo­cal wines to vi­su­alise the sen­sory space of Chenin wines.

Sci­en­tists from Stel­len­bosch Uni­ver­sity’s In­sti­tute of Wine Biotech­nol­ogy‚ the Fac­ulty of AgriS­ciences and the Depart­ment of In­for­ma­tion Science con­trib­uted to the study.

The re­searchers used a data vi­su­al­i­sa­tion tech­nique known as for­mal con­cept lat­tices in com­bi­na­tion with tra­di­tional sta­tis­ti­cal meth­ods to vi­su­alise the sen­sory at­tributes of Chenin blanc and Sauvi­gnon blanc wines.

This tech­nique is nor­mally used in in­for­ma­tion science and ma­chine learn­ing (a method of data anal­y­sis that au­to­mates an­a­lyt­i­cal model build­ing)‚ ac­cord­ing to the uni­ver­sity.

The re­searchers also con­sid­ered the un­struc­tured sen­sory data on 4 352 Sauvi­gnon blanc wines pub­lished in the John Plat­ter Wine Guide to South African Wines for the same pe­riod (2008-2014) to con­firm their find­ings.

The to­tal num­ber of words used by Plat­ter wine pan­el­lists to de­scribe the 2 746 Chenin blanc wines in this study amounted to 38 403‚ and 71 892 words for 4 352 Sauvi­gnon blanc wines.

Since the re­searchers were only in­ter­ested in aroma at­tributes‚ they re­moved all non-sen­sory words from the orig­i­nal data set and elim­i­nated du­pli­cates.

A to­tal of 266 and 250 dif­fer­ent aroma sen­sory words were iden­ti­fied for Chenin blanc and Sauvi­gnon blanc wines re­spec­tively.

These find­ings had im­por­tant im­pli­ca­tions for South Africa’s Chenin blanc wine style la­belling‚ the Uni­ver­sity of Stel­len­bosch said.

Newspapers in English

Newspapers from South Africa

© PressReader. All rights reserved.