Testing resume parsing

To ensure that resume parsing is working as required run some tests before full implementation. Sovren suggests a number of rules for successfully testing resume parsing including:

  1. Do not use fake or disguised resumes.

    The engine is designed to recognize and reject resumes containing:

    • Test or a similar string as part or all of the candidate name.
    • Employer 1 or similar in a career history.
    • Text translated by an online translation engine for the purposes of testing.
  2. Do not use resumes supplied by a parsing engine vendor.

    They are typically selected to perform well with the vendor's own engine and poorly with others.

  3. When testing resumes in more than one language, test a number of resumes per language or locale.
  4. Test resumes from more than one source and that apply to a representative range of industries and classes, relevant to the needs of your organization.
  5. Validate parsed resumes manually, checking the parsed results against the originals.
  6. Test for completeness and accuracy.