Souds like what we have done. Problems are 1) whether you can gather enough labeled data. 2) which feature you assume correlated to age.
About the first problem, use your wisdom or resource. We have been doing age range and gender classification using n-gram features. Accuracy was pretty good. Featuer engineering is extremely important for such simple task. Wenliang Gao 2016年2月29日月曜日、Jeffrey Zemerick<[email protected]>さんは書きました: > This sounds like a document classification problem. I don't think you > need annotated text but instead you need text samples written by > authors of different age ranges. Even with that are you sure there is > a correlation between an author's text and their age (or maybe that's > what you're trying to discover)? I'm personally unaware of any data > like that. > > On Sat, Feb 27, 2016 at 1:16 AM, Madhav Sharan <[email protected] > <javascript:;>> wrote: > > Hey, > > > > I am trying to build a age recognition classifier which will take input a > > text blob and output most likely age of author. > > > > Can someone guide if there is one available already? > > > > Also is there freely available annotated data for age recognition? > > > > Any help is appreciated > > > > Thanks a lot. > > > > -- > > Madhav Sharan >
