How to use the tool
Enter text, choose the categories to check, and review the grouped results. The tool updates as you type or load a file.
- Paste text into the input area. You can also select Upload or drop a file onto the input area. The file picker accepts
.txt,.csv,.json,.log, and.mdfiles. - Use Toggle all, or select Person, Organization, Location, Date, Money, Percent, and Time to choose which categories are active.
- Read the grouped output and results table. Each row shows the entity, category, count, and source positions.
- Use Copy or Download to get the results as tab-separated text.
What the results show
The table contains Entity, Category, Count, and Source Position. Results are grouped by category and case-insensitive entity text; the displayed text keeps the first matched form.
Count is the number of occurrences in the source text. Positions are character indexes from the match position. Up to five positions are shown. If there are more, the output adds a +N more indicator.
Entity types detected
- Person: Names with or without supported titles, such as Dr. Jane Smith or Tim Cook.
- Organization: Names with supported suffixes or acronyms, such as Apple Inc. or NASA.
- Location: Supported countries, states, cities, and geographic names such as United States, Austin, and Lake Michigan.
- Date: Numeric, ISO, and month-based forms such as 2024-03-15, 03/15/2024, and March 15, 2024.
- Money: Currency amounts such as \$1 billion or 250,000 USD.
- Percent: Values such as 12% and 3.5%.
- Time: Clock times such as 9:00 AM and 14:30.
How matching works and where it can fail
Recognition is heuristic and regex-based. The implementation checks regular expressions, fixed country, state, and city lists, capitalization, supported titles, organization suffixes, acronyms, and stopword rules. Dates, money, percentages, and times use their own patterns.
Matching depends on those patterns and vocabularies. Unusual names, unsupported wording, and ambiguous capitalization may be missed or misclassified. Overlapping matches are resolved by category priority, so one text span is not reported under multiple categories. No maximum input size is documented.
Privacy and processing
Extraction runs in your browser through JavaScript and browser APIs. Uploaded files are read with FileReader, and the local extraction code does not make a network request. The implementation does not document a maximum input size or specific handling for unreadable files.
Frequently Asked Questions (FAQs)
What entity types does this tool detect?
People, organizations, locations, dates, money amounts, percentages, and times. Each category has its own toggle so you can narrow results to exactly what you need.
How does it recognize people and organizations without an AI model?
It uses regular expressions and fixed rules for titles, capitalization, organization suffixes, acronyms, lists, and formatted values. It does not use an AI model for extraction.
What does the source position column show?
The character index of each occurrence in the source text. Up to five positions are listed; additional positions appear as +N more.
Is my text sent to a server for analysis?
No network request is made for extraction. The browser reads pasted text or uploaded files locally with JavaScript and browser APIs.
Will it catch every entity perfectly?
No. Recognition depends on fixed patterns and vocabularies, so unusual names, unsupported wording, ambiguous capitalization, and overlapping matches can be missed or classified differently. Review the results before relying on them.