Not Every Quantifier Can Be Negated — Dataset Repository

This repository contains the dataset used in our paper:

Not Every Quantifier Can Be Negated
Proceedings of SuB30

The repository is intended to make our data, annotation scheme, and assumptions transparent and reusable.

---

Repository Contents

• Dataset:
  The dataset is stored as JSON arrays, organized by domain and language.
• Paper context:
  The dataset supports the empirical analysis presented in the paper, focusing on the interaction between quantifiers and negation across languages.

---

Dataset Structure

Each JSON file consists of an array of hits.
A hit is a sentence that matches our ruleset for detecting negation–quantifier interactions.

Hit Format

Each element in the array has the following structure:

[
  {
        "id": "114302_0",
        "neg_idx": 20,
        "quantor": "all",
        "quantor_idx": 21,
        "sent": [
            "Birkenstock",
            "sold",
            "32",
            "million",
            "shares",
            "at",
            "$",
            "46",
            "per",
            "share",
            "(",
            "ahead",
            "of",
            "the",
            "green",
            "shoe",
            "option",
            ")",
            "although",
            "that",
            "not",
            "all",
            "proceeds",
            "benefited",
            "the",
            "company",
            "with",
            "some",
            "proceeds",
            "going",
            "to",
            "selling",
            "shareholders",
            ".",
            "\n"
        ],
        "source": {
            "date": "2024-01-19\n",
            "id": 114302,
            "sent": "Birkenstock sold 32 million shares at $46 per share (ahead of the green shoe option) although that not all proceeds benefited the company with some proceeds going to selling shareholders.\n",
            "source": "https://seekingalpha.com/article/4663712-birkenstock-landing-on-its-feet?source=feed_all_articles"
        },
        "type": "neg"
    }
]

Field Description

• id A unique identifier for the hit.

• sent The sentence containing the quantifier and negation.

• neg_idx Token index of the negation marker.

• quantor_idx Token index of the detected target quantifier.

• quantor The surface form of the detected quantifier.

• type The classification of the negation–quantifier interaction following Barwise & Cooper (1981):

  • neg
  • probable_neg
  • un_neg
  • highly_neg
  • un_neg(ne)
  • un_neg(ne-contrast)

• source The original data source (either a sentence from the Wortschatz project or a tweet).

---

Negation Types

• neg Clear cases where the quantifier can be negated.
• probable_neg Cases where negation is possible but context-dependent.
• un_neg Quantifiers that generally resist negation.
• highly_neg Quantifiers that strongly license negation.
• un_neg(ne) / un_neg(ne-contrast) Special cases where the quantifier scopes over a noun that can be interpreted as a named entity, which affects negation behavior (e.g., [Context: Starkey will come.] oh no, not Starkey. Please not Starkey. [Tweet ID 102069626590081024]).

---

Quantifiers by Language

English

neg quantifiers

{ every, all, a, one, two, three, ten, many, much, lot, plenty }

un_neg quantifiers

{ each, some, no, none, any, couple }

probable_neg quantifiers

{ few, little, several, lot }

highly_neg quantifiers

{ most }

---

German

neg quantifiers

{ jeder, jede, jedes, jed, jedem, jeden,
  alle, all, aller, allem, allen,
  ein, eine, eins, einer, einen, einem,
  zwei, drei, vier, fünf, sechs, sieben, acht, neun, zehn,
  viele, viel, vielen, vieler, vielem,
  zahlreich, zahlreiche, zahlreichen, zahlreichem, zahlreicher }

un_neg quantifiers

{ mancher, manche, manches, manchen, manchem,
  einige, einiges, einiger, einigen, einigem,
  kein, keine, keiner, keins, keinen, keinem,
  irgendein, irgendeine, irgendeiner, irgendeinem, irgendeinen, irgendeins }

probable_neg quantifiers

{ wenige, wenig, weniger, wenigere, wenigeren, wenigen, wenigsten, wenigste,
  mehrere, mehreren, mehrerer, mehrerem, mehreres }

highly_neg quantifiers

{ meiste, meisten, meister, meistes, meistem }

---

Citation

If you use this dataset, please cite our paper:

TODO

---

Notes

• Token indices are based on the tokenization used during data extraction.
• The dataset is intended for linguistic and computational semantics research on negation and quantification.