Compare Proposal

Nothing to compare.

Aspect Classification using MonkeyLearn or similar

  • Posted at : 1 month ago
  • Post Similar Project
20

Budget
2
Proposals
57
Views
Awarded
Status
Skills Required

Posted By -

KL

4.9
Projects Posted : 133
Projects Paid : 50
Services Purchased : 2
Total Spent :
1095
Feedbacks : 73 %

Project Details show (+) hide (-)

Looking for someone who claims to be a Data Scientist or interested in AI.

To use the Aspect Classifier at https://monkeylearn.com/blog/aspect-based-sentiment-analysis/ or similar.

I uploaded some 25000 rows representing the tables and fields of a few dozen Open Source applications, in CSV, JSON, and XML formats to Icedrive.

This problem I would call "classification", I believe it is specifically Aspect Classification.

For example, say SugarCRM has Contacts and vTigerCRM has People, which serve the same function, which we can tell because both are related to their respective Accounts tables.  Those would probably be Aspect tagged as People tables as a Parent Class, and Contact as a function class.

Regarding fields: for example, email, phone, address are what I am calling Contact Fields, and further, "Classic" Contact fields, one has a phone field and one phone number, which are equivalent, and to be categorized as Contact Method-Classic, as opposed to Contacts - Social, which would be all the FB, Twitter, LIn, etc. social networks.

So the category could be either hierarchical or hyphenated, Contacts - Classic, vs. Contacts - Social

Idea is that eventually, we can compare-  this contacts table has 15 fields and the other one 14, this one has 3 Classic Contacts and the other only 2, the other one has 7 Socials, and this one, only 4, etc.

When you train MonkeyLearn, the idea is that the model will become more robust over time, so that you might only have to classify 1000 records to get good classification of all 25000.

Write an intelligent report of findings, suggestions of next steps, etc. and transfer the MonkeyLearn or similar account to one of my emails.