polygraf-ai/applied-nlp-ner-candidate-starter-100

Applied NLP / NER Candidate Starter Dataset

This dataset contains 100 English NER records prepared for the Polygraf Applied NLP / NER technical task.

The annotations were made messy on purpose. Some spans are wrong, some labels are inconsistent, some records contain traps, and some records are clean. Candidates should load this dataset, review the annotations carefully, and correct them following the Labels and Baseline labeling rules below, plus any extra policy rules they define to close remaining gaps.

Dataset summary

Item	Value
Records	100
Language	English
Task	Named entity recognition / token classification
Labels	PERSON, ORGANIZATION, LOCATION, TIMEDATE, PRODUCT, WORKOFART, JOB, AMOUNT

Record schema

Each record has three fields:

• unique_index (int): the record identifier
• source_text (string): the original sentence or passage
• privacy_mask (list): labeled entities; each entity has:
  • start (int): inclusive character start offset in source_text
  • end (int): exclusive character end offset in source_text
  • label (string): entity class
  • value (string): surface form, which should equal source_text[start:end] when the annotation is correct

Example of one record from this dataset, shown in full:

{
  "unique_index": 51,
  "source_text": "A visit to Flatley Dam in Union County confirmed the alternate lifestyle Kohler maintained outside her role as a International Solutions Analyst at Harber Inc .",
  "privacy_mask": [
    {"start": 11, "end": 22, "label": "LOCATION", "value": "Flatley Dam"},
    {"start": 26, "end": 38, "label": "LOCATION", "value": "Union County"},
    {"start": 73, "end": 79, "label": "PERSON", "value": "Kohler"},
    {"start": 113, "end": 144, "label": "JOB", "value": "International Solutions Analyst"},
    {"start": 148, "end": 158, "label": "ORGANIZATION", "value": "Harber Inc"}
  ]
}

This particular record is shown only to illustrate the format. It is not a statement that every annotation in the dataset is correct.

How to load

from datasets import load_dataset

ds = load_dataset("polygraf-ai/applied-nlp-ner-candidate-starter-100", split="train")
print(ds[0]["source_text"])
print(ds[0]["privacy_mask"])

Labels

• PERSON: a named person, including given names and surnames.
  • Example: <PERSON>Barack Obama</PERSON> / <PERSON>Sarah Chen</PERSON>.
• ORGANIZATION: a named company or institution.
  • Example: <ORGANIZATION>Google</ORGANIZATION> / <ORGANIZATION>Mayo Clinic</ORGANIZATION>.
• LOCATION: a named place such as a city, country, region, street, landmark, or geographic area.
  • Example: <LOCATION>Berlin</LOCATION> / <LOCATION>New York City</LOCATION>.
• TIMEDATE: an expression that places something on a timeline, including dates, clock times, and durations used as time.
  • Example: <TIMEDATE>March 15, 2024</TIMEDATE> / <TIMEDATE>50 minutes</TIMEDATE>.
• PRODUCT: a named commercial product, device, or branded good.
  • Example: <PRODUCT>iPhone 15</PRODUCT> / <PRODUCT>MacBook Pro</PRODUCT>.
• WORKOFART: a named creative or published work such as a book, film, song, or titled publication.
  • Example: <WORKOFART>Oppenheimer</WORKOFART> / <WORKOFART>Spider-Man: Brand New Day</WORKOFART>.
• JOB: an occupational title or formal work role when it functions as such in the sentence.
  • Example: <JOB>software engineer</JOB> / <JOB>CEO</JOB>.
• AMOUNT: a measurable or countable quantity, not a time or date expression.
  • Example: <AMOUNT>50</AMOUNT> tickets / <AMOUNT>120,000</AMOUNT>.

Baseline labeling rules

These are the main rules. They are fixed and should not be edited. Corrections and later labeling decisions should follow these main rules plus any extra policy rules added for gaps.

• A labeled mention should cover the entity itself, not the surrounding grammar.
  • Example: Correct: She visited <LOCATION>Paris</LOCATION> yesterday. Incorrect: labeling visited Paris yesterday as one span.
• Articles, prepositions, conjunctions, and other function words should stay outside the span unless they are truly part of the proper name.
  • Example: Correct: He works at <ORGANIZATION>Google</ORGANIZATION>. Incorrect: <ORGANIZATION>at Google</ORGANIZATION>.
  • Example: <ORGANIZATION>The New York Times</ORGANIZATION> should keep The when it belongs to the established name.
• Bare type-words and category descriptors, meaning words that name a category rather than a specific entity, should not be labeled on their own. Examples include "person", "company", "organization", "team", "hospital", "city", "product", "book", and "quantity".
  • Example: The company hired 200 people. should leave company unlabeled.
  • Example: She joined <ORGANIZATION>Acme Corp</ORGANIZATION>. should still label the named organization.
  • Example: The person who called did not leave a name. should leave person unlabeled.
• A multi-word name should be one span when it forms one real named entity; separate entities should be labeled separately.
  • Example: <PERSON>Barack Obama</PERSON> should be one span.
  • Example: <PERSON>Barack</PERSON> and <PERSON>Michelle</PERSON> should be two spans.
• Coordinated names should be separate spans, unless the conjunction is part of one established name.
  • Example: <ORGANIZATION>Google</ORGANIZATION> and <ORGANIZATION>Microsoft</ORGANIZATION> should be two spans.
  • Example: <ORGANIZATION>Johnson & Johnson</ORGANIZATION> should stay one span because & belongs to the name.
• Ordinary punctuation and stray whitespace should stay outside spans, unless the punctuation belongs to the name or abbreviation.
  • Example: He moved to <LOCATION>Berlin</LOCATION>. should leave the final period outside the span.
  • Example: <ORGANIZATION>AT&T</ORGANIZATION> should keep & inside the span.
• Possessive markers should stay outside the span unless the full possessive form is the name.
  • Example: <PERSON>Maria</PERSON>'s laptop should leave 's outside the span.
• Labels should follow context, not surface form alone. The same words may be an organization in one sentence and a place in another.
  • Example: She works at <ORGANIZATION>Cambridge University</ORGANIZATION>.
  • Example: The conference was held in <LOCATION>Cambridge</LOCATION>.
• Quantities used as time should be labeled as TIMEDATE, not AMOUNT.
  • Example: She bought <AMOUNT>50</AMOUNT> tickets.
  • Example: The train arrives in <TIMEDATE>50 minutes</TIMEDATE>.

These baseline rules will not cover every case. When an ambiguous labeling gap appears, do not guess case by case: add a new policy rule that closes that gap. Every addition or revision should be documented in the report, with a short example and a clear justification. Each new rule must not contradict the other added rules or these baseline rules.

For example, take a phrase such as “about 50 tickets.” The baseline rules already tell you that 50 can belong under AMOUNT, but they do not say whether “about” should sit inside the span or stay outside it. This is not a trivial wording choice: 50 and about 50 are not the same value, so the boundary decision should be thought through carefully. That is exactly the kind of gap a new rule should settle.

License and intended use

The source_text passages are drawn from publicly available English text collections. They are redistributed here in short excerpts, unmodified in wording, solely so that candidates can complete the Polygraf Applied NLP / NER technical task. The annotations in privacy_mask were produced by Polygraf.

You may use this dataset to complete that technical task, including deriving a corrected or expanded dataset from it and publishing that derivative. Any other redistribution or commercial use is not granted here. Rights in the underlying source passages remain with their original holders; if you intend to reuse this material outside the technical task, check the terms of the original sources first.

See the LICENSE file in this repository for the same terms in short form.