Closed

Algorithm to match short text strings (equivalent product names from disparate sources)

This project received 23 bids from talented freelancers with an average bid price of £201 GBP.

Get free quotes for a project like this
Employer working
Project Budget
N/A
Total Bids
23
Project Description

We have an existing MS SQL database which is populated with products retrieved from various sources (portals).

The various portals usually have slightly different names for the products.

E.G

Portal 1:
PortalProduct 1: CocaCola
PortalProduct 2: Manchester United
PortalProduct 3: Maplin Electronics
Portal 2:
PortalProduct 1: Coca-Cola
PortalProduct 2: Man Utd
PortalProduct 3: Maplin Elec Ltd.


- We need to correlate these PortalProducts to a central list of Products.
- One Product can have 1-to-many PortalProducts
- The database is populated with around 10,000 PortalProducts
- The database is not populated with any Products. Products need to be added for each new PortalProduct that doesn't match an existing Product.

This must be implemented as a C# console application so that a) we can see the results when the correlate is run and b) so that we can ultimately run it on a task-schedule.

The developer must develop a text matching algorithm to correlate PortalProduct names with Product names (and create a product where there is no match). The emphasis is on the accuracy of the text matching algorithm. Speed is less important than [url removed, login to view] developer should have a mathematical mind and preferably has experience in developing mathematical text matching algorithms.

Examples of mathematical text matching algorithms that could be used (and potentially refined and combined) to meet our requirements are:

String metric
Locality-sensitive hashing
Needleman–Wunsch algorithm
Smith–Waterman algorithm
Levenshtein distance
Concept Search
Approximate matching with addition of regular expressions ability
Regular expressions for non-fuzzy (exact) matching
Metaphone
Soundex
Agrep
Plagiarism detection

[url removed, login to view]

Ideally the developer will already understand and have used some of these before and able to pick the best algorithm(s) for this application.

Looking to make some money?

  • Set your budget and the timeframe
  • Outline your proposal
  • Get paid for your work

Hire Freelancers who also bid on this project

    • Forbes
    • The New York Times
    • Time
    • Wall Street Journal
    • Times Online