0
$\begingroup$

I was looking for some ML/NLP advice. I have 50,000 newsletters (emails) that are labelled either “event” or “not event”. Here is an example of each:

Not event:

COMPLIMENTARY STANDARD SHIPPING ON ALL DOMESTIC ORDERS. 

ENDS SOON 
TIMELESS SKIN SERUM DUO
http://enews.ernolaszlo.com/q/bmhc31K1AEaZwmn8quhjEVD2Gn6bhKECHXID2Z_-oGFLUexGdsY3pzTtf

Protect your youthful glow; stock up now.
http://enews.ernolaszlo.com/q/Y16cHUQU_1l761il6e1xyZBSB5Kl1GNI9uyRdaZhOJoy7obGhj-Hm_eoW

Transphuse Rapid Renewal Cell Protocol 
This month only, get our full-size, revolutionary age-reversing serum free when you spend $400
http://enews.ernolaszlo.com/q/W5UbQ8S85EcX85jCVa01TKcQIR9f0F27SnHlrel6xmpasf7GYw-gCjpHQ

Event:

First Street Gallery Logo [http://r20.rs6.net/tn.jsp?f=001HzXGe-7Pq8- 
zlt4lLYuOB6gUgJEJfvwghspnK22oQib-   APXPI0NJvw4FOyJ- 




OPENING RECEPTION THIS THURSDAY

JULY 23, 6-8 PM
2015
MFA NATIONAL EXHIBITION

Thomas Wharton, Martini, digital print on canvas, 9 x 8 inches

JULY 23 - AUGUST 14, 2015

First Street Gallery proudly presents our fifth annual MFA NATIONAL 
EXHIBITION.

This exhibition highlights the vast reach of fine arts teaching throughout the United
States. The competition is open to all current and former MFA graduates within the
past three years. Our Juror, David A. Ross, former Director of The Whitney Museum
of American Art and The San Francisco Museum of Modern Art, has a 40-year career
 as a museum professional and educator. He is currently the Chair of the MFA Art
 Practice Program at SVA.

ARTISTS: Danny Baskin (UARK), Kimberly Becker (Heartwood), William 
Chambers (Mass
Art), Donna Cleary (SVA), Sarah Dahlinger (Ohio), Jason Egitto 
(Syracuse), Lindsey
Elsey (Clemson), Dan Fenstermacher (SJSU), En Iwamura (Clemson), 
Richard James (KU),
Annie Johnston (UT-Austin), June Korea (SVA), James Lambert (Mass 
Art), Junko Ledneva
(UAF), J. Myszka Lewis (UW-Madison), Wilson Parry (Parsons), Veronica 
Perez (MECA),
Danette Pratt (Ohio), Jason Schwab (CCAD), Thomas Wharton (UT-Knoxville).

For more info please visit the 2015 MFA NATIONAL EXHIBITION album on our 
Facebook
page. [http://r20.rs6.net/tn.jsp?f=001HzXGe-7Pq8- 
zlt4lLYuOB6gUgJEJfvwghspnK22oQib- 
APXPI0NJv4vTi_KIZjl5AqU4Nfz6z3iB6MWBRsf5QHv- 
4T8IUSmtSVEItfj8TEaoluebDWPmWv8D8ayfdc-wgMtgkfoRJvH4e6- 
s0HG2jeuCsdSgT2q8uK4gve2K-u9S0X-AI-dtxpp72dGVY3orPDp-aCm- 
Gm8b_bp2B_Kh4JXLtEQn1qL6y- 



 T3w2wKAO_Ijb4WKJM54g==&c=ThoQa9dOxQlD2qj4Bk8tMgN
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Summer Gallery Hours: 11 am - 6 pm, Monday through Friday

526 West 26th Street, Suite 209, New York, New York 10001

646-336-8053 · 646-336-8054 (fax)

First Street Gallery is located in the heart of Chelsea, NYC between 
10th & 11th
 Avenues.

 [Nearest Subways: C,E,R,1,F,V to 23rd St. - crosstown bus to 10th Ave.

Nearest Buses:  9th Avenue (#11), 8th Avenue (#10)

FSG Gallery Location (Larger) [http://r20.rs6.net/tn.jsp?f=001HzXGe-7Pq8-zlt4lLYuOB6gUgJEJfvwghspnK22oQib- 




~~~~~~~~~~~~~~~~~~~~~~~~~~

 Join Our Mailing List [http://visitor.constantcontact.com/email.jsp?m=1103049216003]

 © 2001-2011 First Street Gallery

 Artist images copyrighted by the individual artists. All rights reserved.

~~~~~~~~~~~~~~~~~~~~~~~~~~
Forward email
http://ui.constantcontact.com/sa/fwtf.jsp? 
llr=dauwnjdab&m=1103049216003&[email protected]$&a=1121719149089

This email was sent to [email protected] by 
[email protected].

Update Profile/Email Address
http://visitor.constantcontact.com/do? 
p=oo&m=001JtllHKXFqd_bBq0Avb7lPA%3D%3D&ch=0d2a4930-e9f2-11e4-af5b-d4ae52754aa9&ca=cd656f5e-7d82-4a66-9e64-9ee1bb170ae9


Instant removal with SafeUnsubscribe(TM)
http://visitor.constantcontact.com/do? 
p=un&m=001JtllHKXFqd_bBq0Avb7lPA%3D%3D&ch=0d2a4930-e9f2-11e4-af5b-d4ae52754aa9&ca=cd656f5e-7d82-4a66-9e64-9ee1bb170ae9


Privacy Policy:
http://ui.constantcontact.com/roving/CCPrivacyPolicy.jsp

Online Marketing by
Constant Contact(R)
www.constantcontact.com



First Street Gallery | 526 W. 26th Street, Suite 209 | New York | NY | 10001

I want to create a model that can triage new newsletters as either event or not. My current plan was to:

  1. Strip all HTML
  2. TFIDF on events vs non-events
  3. Create something that searches for a date/time (events have to have date/time), and then labels the post either 1 or 0
  4. Use some type of spam email code to assess if an email looks like spam, and label it as such

Then throw that data, with the labels, and train using XGBoost or something. Does that sound reasonable? I don't have a strong background in NLP, but I know a lot of you are NLP experts, what kind of architecture would you recommend for this problem? Should I be using LDA or something to identify the parts of speech? Any recommended way to clean HTML?

$\endgroup$

1 Answer 1

0
$\begingroup$

If the only distinction between an event and a non event is the presence of day/time element you might not even require any classification algorithm. Wouldn’t just mining for the day or time elements in the text be sufficient?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.