Presidio falsely detects "JavaScript" and tech terms as <IN_PAN> when analyzing longer texts

[nurullah7733]

Hi team,

I'm running into an odd behavior with mcr.microsoft.com/presidio-analyzer (running via Docker Compose). In my Express.js application, I'm analyzing user-generated resume-style text using the analyzer API.

I've noticed that when the input contains tech terms like "JavaScript" they are often being falsely detected as the IN_PAN entity type (Indian Permanent Account Number) — but only when the input text is long or paragraph-style.

🧪 Examples:
Input 1 (short sentence):
I love working with JavaScript.
Output: ✅ No PII detected. All good.

Input 2 (longer real-world text):
I worked in web design and wordpress development on Fiverr.
It was like 8 to 9 months. After that I shifted to Javascript.
Now I am working with Node JS and learning more...
Output: ❌ Javascript and similar terms get marked as IN_PAN. like this: I worked in web design and wordpress development on Fiverr. It was like <DATE_TIME>. After that I shift to <IN_PAN>.

Here i use in express.js docker-compose.yml file:
`const axios = require("axios");
exports.analyzerAndAnonymizeController = async (req, res) => {
try {
const { text } = req.body;

// Step 1: Call analyzer
const analyzeResponse = await axios.post("http://analyzer:5001/analyze", {
  text,
  language: "en",
});

const analyzerResults = analyzeResponse.data;

// Step 2: Build dynamic anonymizers based on detected entities
const anonymizers = {
  DEFAULT: {
    type: "replace",
    new_value: "<REDACTED>",
  },
};

for (const entity of analyzerResults) {
  const type = entity.entity_type;
  if (!anonymizers[type]) {
    anonymizers[type] = {
      type: "replace",
      new_value: `<${type}>`,
    };
  }
}

// Step 3: Call anonymizer using analyzer results
const anonymizeResponse = await axios.post(
  "http://anonymizer:5002/anonymize",
  {
    text,
    analyzer_results: analyzerResults,
    anonymizers,
  }
);

return res.status(200).json({
  status: "success",
  analyzer: analyzerResults,
  anonymized: anonymizeResponse.data,
});

} catch (error) {
console.error("analyze and anonymize error:", error.message);
return res.status(400).json({ status: "fail", data: error.message });
}
};
`

[SharonHart]

Hi @nurullah7733, see the answer here:
#1680 (comment)