Dataset Open Access

Temporally-Informed Analysis of Named Entity Recognition

Rijhwani, Shruti; Preoțiuc-Pietro, Daniel


Citation Style Language JSON Export

{
  "publisher": "Zenodo", 
  "DOI": "10.5281/zenodo.3899040", 
  "language": "eng", 
  "title": "Temporally-Informed Analysis of Named Entity Recognition", 
  "issued": {
    "date-parts": [
      [
        2020, 
        6, 
        17
      ]
    ]
  }, 
  "abstract": "<p>This repository contains the data set developed for the paper:</p>\n\n<p>&ldquo;Shruti Rijhwani and Daniel Preo\u021biuc-Pietro. <em>Temporally-Informed Analysis of Named Entity Recognition.</em> In Proceedings of the Association for Computational Linguistics (ACL). 2020.&rdquo;</p>\n\n<p>It includes 12,000 tweets annotated for the named entity recognition task. The tweets are uniformly distributed over the years 2014-2019, with 2,000 tweets from each year. The goal is to have a temporally diverse corpus to account for data drift over time when building NER models.</p>\n\n<p>The entity types annotated are locations (LOC), persons (PER) and organizations (ORG). The tweets are preprocessed to replace usernames and URLs with a unique token. Hashtags are left intact and can be annotated as named entities.</p>\n\n<p><strong>Format</strong></p>\n\n<p>The repository contains the annotations in JSON format.</p>\n\n<p>Each year-wise file has the tweet IDs along with token-level annotations. The Public Twitter Search API (<a href=\"https://developer.twitter.com/en/docs/tweets/search\">https://developer.twitter.com/en/docs/tweets/search</a>) can be used extract the text for the tweet corresponding to the tweet IDs.</p>\n\n<p><strong>Data Splits</strong></p>\n\n<p>Typically, NER models are trained and evaluated on annotations available at the model building time, but are used to make predictions on data from a future time period. This setup makes the model susceptible to temporal data drift, leading to lower performance on future data as compared to the test set.</p>\n\n<p>To examine this effect, we use tweets from the years 2014-2018 as the training set and random splits of the 2019 tweets as the development and test sets. These splits simulate the scenario of making predictions on data from a future time period.</p>\n\n<p>The development and test splits are provided in the JSON format.</p>\n\n<p><strong>Use</strong></p>\n\n<p>Please cite the data set and the accompanying paper if you found the resources in this repository useful.</p>", 
  "author": [
    {
      "family": "Rijhwani, Shruti"
    }, 
    {
      "family": "Preo\u021biuc-Pietro, Daniel"
    }
  ], 
  "id": "3899040", 
  "type": "dataset", 
  "event": "The 58th Annual Meeting of the Association for Computational Linguistics (ACL2020)"
}
309
81
views
downloads
All versions This version
Views 309309
Downloads 8181
Data volume 15.0 MB15.0 MB
Unique views 287287
Unique downloads 7878

Share

Cite as