metadata

license: apache-2.0
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.public.merged.json
      - split: validation
        path: valid.public.merged.json
      - split: test
        path: test.public.merged.json
      - split: academic
        path: Academic.public.merged.json
      - split: ood
        path: OOD.public.merged.json
  - config_name: paper
    data_files:
      - split: parts
        path:
          - part_*.public.merged.json
      - split: academic
        path: Academic.public.merged.json
      - split: ood
        path: OOD.public.merged.json

TweetNERD - End to End Entity Linking Benchmark for Tweets

This is the hydrated version of dataset described in the paper TweetNERD - End to End Entity Linking Benchmark for Tweets (to be released soon). It includes the Tweet text based on the Twitter API.

Named Entity Recognition and Disambiguation (NERD) systems are foundational for information retrieval, question answering, event detection, and other natural language processing (NLP) applications. We introduce TweetNERD, a dataset of 340K+ Tweets across 2010-2021, for benchmarking NERD systems on Tweets. This is the largest and most temporally diverse open sourced dataset benchmark for NERD on Tweets and can be used to facilitate research in this area.

TweetNERD dataset is released under Creative Commons Attribution 4.0 International (CC BY 4.0) LICENSE.

The license only applies to the data files present in this dataset. See Data usage policy below.

Usage

We provide the dataset split across the following tab seperated files:

OOD.public.merged.tsv: OOD split of the data in the paper.
Academic.public.merged.tsv: Academic split of the data described in the paper.
part_*.public.merged.tsv: Remaining data split into parts in no particular order.

Official train test splits:

train.public.merged.tsv: Train split as described in paper based on part_* splits.
valid.public.merged.tsv: Validation split as described in paper based on part_* splits.
test.public.merged.tsv: Test split as described in paper based on part_* splits.

Each file is tab seperated and has has the following format:

tweet_id	phrase	start	end	entityId	score
22	[twttr]	[20]	[25]	[Q918]	[3]
21	[twttr]	[20]	[25]	[Q918]	[3]
1457198399032287235	[Diwali]	[30]	[38]	[Q10244]	[3]
1232456079247736833	[NO_PHRASE]	[-1]	[-1]	[NO_ENTITY]	[-1]

For tweets which don't have any entity, their column values for phrase, start, end, entityId, score are set NO_PHRASE, -1, -1, NO_ENTITY, -1 respectively.

Description of file columns is as follows:

Column	Type	Missing Value	Description
tweet_id	string		ID of the Tweet
phrase	string	NO_PHRASE	entity phrase
start	int	-1	start offset of the phrase in text using `UTF-16BE` encoding
end	int	-1	end offset of the phrase in the text using `UTF-16BE` encoding
entityId	string	NO_ENTITY	Entity ID. If not missing can be NOT FOUND, AMBIGUOUS, or Wikidata ID of format Q{numbers}, e.g. Q918
score	int	-1	Number of annotators who agreed on the phrase, start, end, entityId information

In order to use the dataset you need to utilize the tweet_id column and get the Tweet text using the Twitter API (See Data usage policy section below).

Data stats

Split	Number of Rows	Number unique tweets	Number hydrated tweets
OOD	34102	25000	20937
Academic	51685	30119	28694
part_0	11830	10000	6633
part_1	35681	25799	19181
part_2	34256	25000	19876
part_3	36478	25000	20611
part_4	37518	24999	20567
part_5	36626	25000	20667
part_6	34001	24984	20948
part_7	34125	24981	20612
part_8	32556	25000	20610
part_9	32657	25000	21000
part_10	32442	25000	20597
part_11	32033	24972	20583
----------	------------------	------------------------	------------------------
train	349252	255490	207278
valid	6822	5000	4128
test	34129	25000	20274

File Stats are as follows:

part	output_file	orig_rows	unique_tweet_ids	final_rows
Academic	Academic.public.merged.json	51685	30119	28694
OOD	OOD.public.merged.json	34102	25000	20937
part_0	part_0.public.merged.json	11830	10000	6633
part_1	part_1.public.merged.json	35681	25799	19181
part_10	part_10.public.merged.json	32442	25000	20597
part_11	part_11.public.merged.json	32033	24972	20583
part_2	part_2.public.merged.json	34256	25000	19876
part_3	part_3.public.merged.json	36478	25000	20611
part_4	part_4.public.merged.json	37518	24999	20567
part_5	part_5.public.merged.json	36626	25000	20667
part_6	part_6.public.merged.json	34001	24984	20948
part_7	part_7.public.merged.json	34125	24981	20612
part_8	part_8.public.merged.json	32556	25000	20610
part_9	part_9.public.merged.json	32657	25000	21000
test	test.public.merged.json	34129	25000	20274
train	train.public.merged.json	349252	255490	207278
valid	valid.public.merged.json	6822	5000	4128

Data usage policy

Use of this dataset is subject to you obtaining lawful access to the Twitter API, which requires you to agree to the Developer Terms Policies and Agreements.

Cite as:

Mishra, Shubhanshu, Saini, Aman, Makki, Raheleh, Mehta, Sneha, Haghighi, Aria, & Mollahosseini, Ali. (2022). TweetNERD - End to End Entity Linking Benchmark for Tweets (0.0.0) [Data set]. Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks (Neurips), New Orleans, LA, USA. Zenodo. https://doi.org/10.5281/zenodo.6617192 Mishra, S., Saini, A., Makki, R., Mehta, S., Haghighi, A., & Mollahosseini, A. (2022). TweetNERD -- End to End Entity Linking Benchmark for Tweets (Version 1). arXiv. https://doi.org/10.48550/ARXIV.2210.08129

Bibtex:

@inproceedings{TweetNERD,
  doi = {10.48550/ARXIV.2210.08129},
  
  url = {https://arxiv.org/abs/2210.08129},
  
  author = {Mishra, Shubhanshu and Saini, Aman and Makki, Raheleh and Mehta, Sneha and Haghighi, Aria and Mollahosseini, Ali},
  
  keywords = {Computation and Language (cs.CL), Artificial Intelligence (cs.AI), Information Retrieval (cs.IR), Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences, I.2.7, 68T50, 68T07},
  
  title = {{TweetNERD} -- {End to End Entity Linking Benchmark for Tweets}},
  
  publisher = {arXiv},
  
  year = {2022},
    booktitle = "Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks 2 (NeurIPS Datasets and Benchmarks 2022)",

  copyright = {Creative Commons Attribution 4.0 International}
}

@dataset{mishra_shubhanshu_2022_6617192,
  author       = {Mishra, Shubhanshu and
                  Saini, Aman and
                  Makki, Raheleh and
                  Mehta, Sneha and
                  Haghighi, Aria and
                  Mollahosseini, Ali},
  title        = {{TweetNERD - End to End Entity Linking Benchmark
                   for Tweets}},
  month        = jun,
  year         = 2022,
  note         = {{Data usage policy  Use of this dataset is subject
                   to you obtaining lawful access to the [Twitter
                   API](https://developer.twitter.com/en/docs
                   /twitter-api), which requires you to agree to the
                   [Developer Terms Policies and
                   Agreements](https://developer.twitter.com/en
                   /developer-terms/).}},
  publisher    = {Zenodo},
  version      = {0.0.0},
  doi          = {10.5281/zenodo.6617192},
  url          = {https://doi.org/10.5281/zenodo.6617192}
}