Applying ChatGPT to Genealogical Data Processing: Constructing Family Trees and Tracing Ancestral Lineages

Processing and standardizing genealogical data is a complex problem, as lineage records in Vietnam are often scattered across stele inscriptions, rough translations of family books, or unstructured oral traditions. This article demonstrates how to leverage ChatGPT and structured prompting techniques to convert raw historical text into standardized family trees formatted as JSON Graph, Mermaid diagrams, and international GEDCOM files.

1. Challenges in Processing Traditional Genealogy Data

Traditional Vietnamese family records often face three major technical barriers when brought into the digital environment:

• Complex hierarchical relationships: Generational naming and addressing conventions (sexagenary cycle, generation number, sub-branches, senior and junior lines) are extremely difficult to map automatically without standardized relational logic.
• Missing data: Many lineage branches have lost records of birth years, death dates, or ancient geographic names due to war and migration.
• Unstructured format: Records are written in a free-form narrative style, interspersed with historical anecdotes and archaic language, lacking any tabular structure.

2. Step 1: Extract Person Entities and Chronology (NER)

The first stage is Named Entity Recognition (NER). We instruct the AI to scan the raw text to extract all individuals, their generation numbers, and key chronological milestones:

You are an expert in Vietnamese history and genealogy.
Below is a raw historical excerpt from a family branch:
[PASTE GENEALOGY RECORD TEXT HERE]

TASK:
Extract a list of all individuals appearing in the text above and format it as a Markdown table with the following columns:
1. ID: Unique identifier (P01, P02...)
2. Full Name: Full name along with courtesy name, pseudonym (if available)
3. Generation: Generation number within the clan
4. Birth / Death Year: Explicit Gregorian calendar years (if sexagenary cycle information is present, convert it approximately)
5. Family Relation: Eldest/younger son of whom, spouse of whom
6. Hometown / Burial Place: Specific geographical locations

Once the AI outputs the consolidated table, run a quick cross-check to ensure that archaic names and feudal era reign titles have been converted accurately.

3. Step 2: Set Up Parent-Child and Spousal Relationship Matrices

The entity data needs to be linked into a relationship network. We instruct the AI to clearly define two core types of relationships: Parent-Child relationships and Spousal relationships:

Based on the list of individuals extracted in step 1, convert all information into a standard JSON Graph data structure:
{
  "nodes": [
    {
      "id": "P01",
      "name": "Nguyen Van To",
      "generation": 1,
      "gender": "male",
      "birth": 1820,
      "death": 1895
    }
  ],
  "links": [
    {
      "source": "P01",
      "target": "P02",
      "relationship": "parent-child"
    }
  ]
}

Requirements:
- Strictly prevent any illogical relationship loops (e.g., a child born before a parent).
- Export clean JSON data, ready to be ingested into graph visualization libraries.

4. Step 3: Visualizing the Family Tree with a Mermaid JS Diagram

To allow family members to view the diagram intuitively without installing complex software, we convert the data into Mermaid JS syntax. This syntax is natively supported on GitHub, Notion, and modern websites:

Based on the established family relationship structure, generate the Mermaid JS (graph TD) source code to display the family tree:
- Use the graph TD syntax (Top-Down from top to bottom).
- 1st generation members are placed on the top row, branching down to the 2nd and 3rd generations.
- Format each block to include: Full name along with year of birth - year of death.
- Apply a prominent background color to the main/eldest branch.

Sample Diagram Code Generated by AI

graph TD
    classDef truong fill:#fde047,stroke:#000,stroke-width:2px;
    classDef thu fill:#FDF8EC,stroke:#000,stroke-width:1.5px;

    P01["Cụ Khởi Tổ: Nguyễn Quý Công (1820 - 1895)"]:::truong
    P02["Đời 2 (Trưởng): Nguyễn Văn Giáp (1850 - 1920)"]:::truong
    P03["Đời 2 (Thứ): Nguyễn Văn Ất (1855 - 1930)"]:::thu

    P01 sang P02
    P01 sang P03

    P04["Đời 3: Nguyễn Văn Bính (1880 - 1950)"]:::truong
    P02 sang P04

5. Step 4: Export to Internationally Standardized GEDCOM Format

GEDCOM (Genealogical Data Communication) is the globally recognized standard for exchanging genealogical data. After standardizing with ChatGPT, you can export to a .ged file to import into professional software such as Gramps, MyHeritage, or Ancestry:

0 HEAD
1 SOUR CHATGPT_GENEALOGY
1 GEDC
2 VERS 5.5.1
2 FORM LINEAGE-LINKED
1 CHAR UTF-8
0 @I1@ INDI
1 NAME Nguyễn Quý /Công/
1 SEX M
1 BIRT
2 DATE 1820
1 DEAT
2 DATE 1895
0 @F1@ FAM
1 HUSB @I1@
1 CHIL @I2@
0 TRLR

6. Common Errors and Troubleshooting

A compilation of solutions for technical issues during the clan family tree digitization process:

7. Frequently Asked Questions (FAQ)

Answers to common questions about applying technology to preserve family history:

Read next: A Comprehensive 9-Step Roadmap to Mastering Claude in 2 Days