Scoring on the basis of Job Postings and Job Applications - Text Miner/NLP

[TET_34]

Hello everyone,

I have a question about Text Mining.

We have a data that includes whole job applications with candidate information and we have also data includes whole job postings with employer information and these datas are historical. These data sets will be a live data in the future but we currently proceed over a specific reporting date(snapshot).

For Job Postings we have two information;

Employer ID
Job Description

We created a single column that is called Job Description by pulling the following values;

Gender
Location
Military Status
Educational Status
Social Status (Disabled, Veteran etc.),
Marital Status
Driver's License
Foreign Language

For Job Applications we have two information;

Employee ID
Candidate Information

We created a single column that is called Candidate Information by pulling the following values;

Gender
Location
Military Status
Educational Status
Social Status (Disabled, Veteran etc.),
Marital Status
Driver's License
Foreign Language

As you know, the nodes I can use for Text Miner on Enterprise Miner are as follows;

Import, Parsing, Filter, Topic, Cluster, Profile and Rule Builder

The targeted operations with Text Mining are as follows;

1. Our goal is to find out how well the Candidate Information and the Job Description table match.
2. Assigning a score according to the match rate.
3. We want to protect the top 30 candidates with the best score by avoiding multiplication.

As a result we want to get an output similar to the table below.

KEY	RANK	Employer_ID	JOB_DESCRIPTION  (TEXT)	Employee_ID	CANDIDATE_INFORMATION (TEXT)	MATCH SCORE
1	1	1	A B c d e	10	A B C D E F	100
2	2	1	A B c d e	20	A B C D E	100
3	3	1	A B c d e	30	A B C	60
4	4	1	A B c d e	70	C UK	20
:	:	1	:	:	:	:
30	30	1	A B c d e	40	A	20
31	1	2	TR uK usA ITA	90	TR UK USA ITA	100
32	2	2	TR uK usA ITA	50	TR UK	50
33	3	2	TR uK usA ITA	60	TR USA	50
34	4	2	TR uK usA ITA	80	TR	25
35	:	2	:	:	:	:
:	30	2	TR uK usA ITA	70	C UK	25

It would be nice if we could get guidance on this case with the Text Miner perspective.