Keyword Matching Algorithm

Keyword Matching & Scoring

The Japanese ATS includes an automated resume analyzer designed to help administrators quickly identify qualified candidates. When an applicant uploads a PDF resume, the system extracts the text and compares it against a standardized dictionary of professional keywords.

The Keyword Dictionary

The algorithm scans for a wide range of professional competencies, specifically tailored for technical and language-oriented roles. The pre-defined dictionary includes:

• Japanese Proficiency: JLPT levels (N1, N2, N3, N4, N5) and general language skills.
• Programming Languages: Java, JavaScript, TypeScript, Python, C++, Ruby, etc.
• Web & Cloud Technologies: React, Node.js, AWS, Azure, Docker, and CI/CD.
• Databases: SQL, PostgreSQL, MongoDB, and others.
• Professional Experience: Titles like "Senior," "Lead," "Architect," and "Manager."
• Education: Verification of degrees (Bachelor, Master, PhD).

Scoring Mechanism

The Keyword Match Score provides a quantitative measure of how well a resume aligns with the system's target criteria.

• Match Increment: Each unique keyword found in the resume adds 10% to the total score.
• Maximum Score: The score is capped at 100%.
• Precision Matching: The algorithm uses word-boundary detection (Regex) to ensure accuracy. For example, it distinguishes "Java" from "JavaScript" and correctly identifies terms with symbols like "C++" or "C#".

Analysis Process

1. Text Extraction: Upon submission, a background Edge Function processes the uploaded PDF using pdfjs-dist to convert document layouts into searchable text.
2. Normalization: The extracted text and keywords are normalized to lowercase to ensure the matching is case-insensitive.
3. Keyword Identification: The system runs a comparison between the resume text and the global dictionary (plus any custom keywords provided).
4. Database Update: The resulting score and a list of specific "Matched Keywords" are stored alongside the applicant's profile.

Administrator View

Administrators can view these metrics directly within the Admin Dashboard and Applicant Detail pages:

• Match Score Badge: A visual percentage indicator that helps in rapid shortlisting.
• Matched Keywords List: A set of tags showing exactly which relevant terms were found in the candidate's CV.
• Extracted Text: For transparency, administrators can view the raw text extracted from the PDF to verify the context of specific keywords.

// Example of how the score is calculated internally:
const score = Math.min(100, matchedKeywords.length * 10);

Limitations

• File Format: Currently, the keyword matching algorithm only supports PDF files.
• Image-based PDFs: Resumes that are saved as images (scanned documents) without an OCR layer cannot be read by the current extractor. Candidates are encouraged to upload text-based PDFs for the best results.