Resume Parsing

Resume parsing

How reveilio extracts text from resumes and what formats are supported.

---

Supported file formats

Extension	Method	Notes
.pdf	PyPDF2	Works with text-based PDFs. Scanned/image PDFs require OCR (not built in).
.docx	python-docx	Full support for modern Word files (Office 2007+).
.doc	antiword / heuristic	Legacy Word binary format. Best results with antiword installed. Falls back to UTF-16LE extraction and ASCII runs.
.txt	UTF-8 decode	Plain text with graceful handling of encoding errors.
Free text	Direct string	Pass text directly without any file. Useful for database rows or API responses.

---

The Resume class

A Resume is a lightweight wrapper that holds the extracted plain text and a filename. Unlike JobDescription, the Resume class does not call the LLM on its own. Structured extraction happens inside analyze_resume() as part of a single combined LLM call.

From a file

resume = reveilio.Resume.from_file("resumes/alice.pdf")
print(resume.filename)   # "alice.pdf"
print(resume.text[:200]) # extracted plain text

From free text

resume = reveilio.Resume.from_text(
    "Alice Example, Senior Python Engineer, 6 years experience...",
    filename="alice.txt",  # optional, used in output reports
)

From bytes

resume = reveilio.Resume.from_bytes(
    uploaded_file.read(),
    filename="alice.pdf",  # extension determines the parser
)

---

What gets extracted during analysis

When you call reveilio.analyze_resume(resume, jd), the scorer extracts the following structured data from the resume in a single LLM call:

Field	Type	Description
name	string	Candidate full name.
email	string	Email address.
phone	string	Phone number.
linkedin	string	LinkedIn profile URL if present on the resume.
skills	list	All technical and non-technical skills found.
experience	list	Each position with title, company, duration, description, and location.
education	list	Each degree with institution, year, and location.
certifications	list	Professional certifications.
total_experience	float	Calculated total years of experience.
top_keywords	list	Most relevant keywords from the resume.
career_gaps	list	Detected gaps in employment history.

---

Handling .doc files (legacy Word binary)

The .doc format is a legacy binary format from Word 97–2003. Reveilio handles it using a multi-step fallback strategy:

1. RTF detection. Some tools save RTF files with a .doc extension. Detected by the {\rtf header and parsed as RTF.
2. OOXML detection. Some tools save DOCX (ZIP-based) files with a .doc extension. Detected by the PK header and parsed as DOCX.
3. Antiword. If the file is truly binary Word, reveilio tries the antiword command-line tool. Install it with sudo apt-get install antiword on Linux.
4. UTF-16LE extraction. Word binary files store body text as UTF-16LE runs. Reveilio extracts these directly.
5. Printable ASCII runs. As a last resort, reveilio scans for sequences of printable ASCII characters in the raw binary.

---

Using the text extractor directly

The text extraction logic is available as a standalone utility:

from reveilio.parsing.text_extract import extract_text

text = extract_text("document.pdf")                    # from file path
text = extract_text(raw_bytes, filename="doc.docx")    # from bytes

---

Convenience: skip the Resume class

You do not have to create a Resume object manually. analyze_resume() accepts file paths and free text directly:

# All three are equivalent
reveilio.analyze_resume("resumes/alice.pdf", jd)
reveilio.analyze_resume(reveilio.Resume.from_file("resumes/alice.pdf"), jd)

# Free text works too
reveilio.analyze_resume("Alice, 6 years Python, FastAPI, AWS", jd)

---

Continue to Scoring & analysis.