Description
This dataset consists of the 286 publications retrieved from Web of Science and Scopus on July 6, 2023 as citations for Willoughby et al., 2014:
Patrick H. Willoughby, Matthew J. Jansma, and Thomas R. Hoye (2014). A guide to small-molecule structure assignment through computation of (¹H and ¹³C) NMR chemical shifts. Nature Protocols, 9(3), Article 3. https://doi.org/10.1038/nprot.2014.042
We added the DOIs of the citing publications into a Zotero collection. Then we exported all 286 DOIs in two formats: a .csv file (data export) and an .rtf file (bibliography).
<b>Willoughby2014_286citing_publications.csv</b> is a Zotero data export of the citing publications.
<b>Willoughby2014_286citing_publications.rtf</b> is a bibliography of the citing publications, using a variation of the American Psychological Association style (7th edition) with full names instead of initials.
To create <b>Willoughby2014_citation_contexts.csv</b>, HZ manually extracted the paragraphs that contain a citation marker of Willoughby et al., 2014. We refer to these paragraphs as the citation contexts of Willoughby et al., 2014. Manual extraction started with 286 citing publications but excluded 2 publications that are not in English, those with DOIs 10.13220/j.cnki.jipr.2015.06.004 and 10.19540/j.cnki.cjcmm.20200604.201
The silver standard aimed to triage the citing publications of Willoughby et al., 2014 that are at risk of propagating unreliability due to a code glitch in a computational chemistry protocol introduced in Willoughby et al., 2014. The silver standard was created stepwise:
First one chemistry expert (YF) manually annotated the corpus of 284 citing publications in English, using their full text and citation contexts. She manually categorized publications as either at risk of propagating unreliability or not at risk of propagating unreliability, with a rationale justifying each category.
Then we selected a representative sample of citation contexts to be double annotated. To do this, MJS turned the full dataset of citation contexts (Willoughby2014_citation_contexts.csv) into word embeddings, clustered them using similarity measures using BERTopic's HDBS, and selected representative citation contexts based on the centroids of the clusters.
Next the second chemistry expert (EV) annotated the 77 publications associated with the citation contexts, considering the full text as well as the citation contexts.
<b>double_annotated_subset_77_before_reconciliation.csv</b> provides EV and YF's annotation before reconciliation.
To create the silver standard YF, EV, and JS discussed differences and reconciled most differences. YF and EV had principled reasons for disagreeing on 9 publications; to handle these, YF updated the annotations, to create the silver standard we use for evaluation in the remainder of our JCDL 2024 paper (<b>silver_standard.csv</b>)
<b>Inter_Annotator_Agreement.xlsx</b> indicates publications where the two annotators made opposite decisions and calculates the inter-annotator agreement before and after reconciliation together.
<b>double_annotated_subset_77_before_reconciliation.csv</b> provides EV and YF's annotation after reconciliation, including applying the reconciliation policy.
Patrick H. Willoughby, Matthew J. Jansma, and Thomas R. Hoye (2014). A guide to small-molecule structure assignment through computation of (¹H and ¹³C) NMR chemical shifts. Nature Protocols, 9(3), Article 3. https://doi.org/10.1038/nprot.2014.042
We added the DOIs of the citing publications into a Zotero collection. Then we exported all 286 DOIs in two formats: a .csv file (data export) and an .rtf file (bibliography).
<b>Willoughby2014_286citing_publications.csv</b> is a Zotero data export of the citing publications.
<b>Willoughby2014_286citing_publications.rtf</b> is a bibliography of the citing publications, using a variation of the American Psychological Association style (7th edition) with full names instead of initials.
To create <b>Willoughby2014_citation_contexts.csv</b>, HZ manually extracted the paragraphs that contain a citation marker of Willoughby et al., 2014. We refer to these paragraphs as the citation contexts of Willoughby et al., 2014. Manual extraction started with 286 citing publications but excluded 2 publications that are not in English, those with DOIs 10.13220/j.cnki.jipr.2015.06.004 and 10.19540/j.cnki.cjcmm.20200604.201
The silver standard aimed to triage the citing publications of Willoughby et al., 2014 that are at risk of propagating unreliability due to a code glitch in a computational chemistry protocol introduced in Willoughby et al., 2014. The silver standard was created stepwise:
First one chemistry expert (YF) manually annotated the corpus of 284 citing publications in English, using their full text and citation contexts. She manually categorized publications as either at risk of propagating unreliability or not at risk of propagating unreliability, with a rationale justifying each category.
Then we selected a representative sample of citation contexts to be double annotated. To do this, MJS turned the full dataset of citation contexts (Willoughby2014_citation_contexts.csv) into word embeddings, clustered them using similarity measures using BERTopic's HDBS, and selected representative citation contexts based on the centroids of the clusters.
Next the second chemistry expert (EV) annotated the 77 publications associated with the citation contexts, considering the full text as well as the citation contexts.
<b>double_annotated_subset_77_before_reconciliation.csv</b> provides EV and YF's annotation before reconciliation.
To create the silver standard YF, EV, and JS discussed differences and reconciled most differences. YF and EV had principled reasons for disagreeing on 9 publications; to handle these, YF updated the annotations, to create the silver standard we use for evaluation in the remainder of our JCDL 2024 paper (<b>silver_standard.csv</b>)
<b>Inter_Annotator_Agreement.xlsx</b> indicates publications where the two annotators made opposite decisions and calculates the inter-annotator agreement before and after reconciliation together.
<b>double_annotated_subset_77_before_reconciliation.csv</b> provides EV and YF's annotation after reconciliation, including applying the reconciliation policy.
Date made available | Nov 7 2024 |
---|---|
Publisher | University of Illinois Urbana-Champaign |
Keywords
- citations
- citation contexts
- scholarly publications
- reproducibility
- unreliable cited sources
- knowledge maintenance
- scientific digital libraries
- unreliability propagation