- Interview with the Director of AI teams:
Case study: Dataset is as below:
| itemId | item_name | item_description | manufacturer_name | supplier_name | target_variable(home/workplace) |
|---|---|---|---|---|---|
| 1000 | monitor | ram: 32gb,2.8 GHZ,HDD | HP | senty pvt ltd | workplace |
| 1001 | |||||
| 1003 |
Given only sample size of 10 of above schema (with target variable assigned), training dataset(without target_variable assigned) size is 100, how would you avoid the effort of manual tagging of target variable to training data?
- How would you select the variables?
- How would u use the opentext column
item_description?
The round is for just 10 minutes, and I didn't get selected.
- k-NN/decision tree algorithm for dependent variable tagging (even associaton rule mining algorithms that include apriori and eclat).
- initial selection of variables is based on domain knowledge as number of variables is very less here.
- preprocess the text column
item_descriptionmaybe do text summarization of the full text and extract keywords using bi-grams and collocations and use that tag in further processing.