Data and analysis pipeline for Williams (2026)

Companion data deposit for:

Williams, G.P. (2026). Identifying Research Gaps and Directions from Published Literature: A Bibliometric and Thematic Synthesis of Utah Lake and Great Salt Lake Research. Water (MDPI), manuscript water-4402795. Accepted; in production. The article DOI is recorded in this resource's Related Resources metadata once issued.

What's in this deposit

• Curated data tables (35 CSVs) corresponding to the manuscript's supplementary references S1-S8 plus core pipeline outputs.
• Deduplicated full corpus in both .csv and .rds, with the abstract column omitted (see "A note on abstract text" below).
• VOSviewer-format network exports (combined search-set, normalized and raw-field variants). Produced with VOSviewer 1.6.20.
• Per-cluster evidence bundle (ZIP, 15 clusters, 30 CSVs), abstract-stripped.
• Redacted LLM prompt records (ZIP, 14 retained clusters) — the full generated prompt for every retained cluster, with each source abstract replaced by [ABSTRACT REDACTED - copyrighted database content]. Instructions, output schema, cluster/model metadata, selection criteria, and representative-paper identifiers are preserved.
• Snapshot of the R analysis pipeline (22 scripts) used to produce the numeric findings in the manuscript. NOT runnable as written; see code/README.md. Run under R 4.5.2 with bibliometrix 5.2.1, igraph 2.2.2, topicmodels 0.2.17 and ggraph 2.2.2.
• Per-subdirectory README files with column-level metadata for every CSV.

Version history

• v2 (2026-07) — restaged for the Water R1 revision. The bibliographic-coupling pipeline was re-run from the final 1,470-record corpus; the coupling subset, community counts, cluster IDs, and all downstream tables changed accordingly. Cluster annotations were regenerated with Claude Opus 5 (Anthropic, accessed 2026-07-28). Adds data/05_clusters_and_annotations/redacted_prompt_records.zip. This is the version the published article corresponds to.
• v1 (2026-05) — original deposit accompanying the first submission, made to Limnology and Oceanography, which declined the manuscript. Its cluster partition and network numbers predate the 2026-07 corpus refresh and do not match the published article. Superseded; use v2.

How to cite

Williams, G.P. (2026). Data and analysis pipeline for "Identifying Research Gaps and Directions from Published Literature: A Bibliometric and Thematic Synthesis of Utah Lake and Great Salt Lake Research." HydroShare. https://doi.org/10.4211/hs.1efc16e179214b3890fad18a7b5c5e1f

Also cite the published manuscript when using these data.

License

Data: CC BY 4.0 Code: MIT

Quick file map

data/ 01_search_and_corpus/ Search outputs + full deduplicated corpus 02_descriptive/ Corpus descriptive stats 03_networks/ Network metrics (Table 3) 04_topic_modeling/ LDA outputs (Table 10) 05_clusters_and_annotations/ Cluster index + AI-assisted annotations + per-cluster evidence bundle + redacted LLM prompt records 06_authors_and_sources/ Author + source ranking tables (S5-S8) 07_temporal_trends/ Year x cluster x topic-share trends 08_normalization_maps/ Keyword / institution normalization rules 09_vosviewer_exports/ VOSviewer-format network exports code/ R-pipeline snapshot (see code/README.md) file_inventory.csv Reader-facing inventory: path, bytes, column count, SHA-256, short notes

Every subdirectory has its own README with per-file metadata.

A note on abstract text

Several analysis steps in the pipeline (topic modeling, keyword co-occurrence on abstract terms, LLM-assisted cluster annotation) rely on the full abstract text of each record. Abstracts are copyrighted publisher content and CANNOT be redistributed via this deposit. Every file in this deposit — including the corpus (.csv and .rds), the VOSviewer exports, the per-cluster evidence bundle, and the redacted prompt records — ships with the abstract columns (AB, AB_raw, and abstract-derived variants) omitted.

To re-run the abstract-dependent analysis steps, download the matching records from Web of Science, Scopus, or ProQuest using your own institutional subscriptions and the search strings in the manuscript's Methods §2.1, then merge abstracts back in by DOI / UT.

The PRISMA counts in data/01_search_and_corpus/ document the corpus build at the metadata level, sufficient for replicating the analysis steps that don't need abstract text (descriptive statistics, network construction, author / source rankings, Bradford-law analysis).

Where the raw inputs live

The pipeline reads from Data/CleanData/ in the original project repo (deduplicated WoS / Scopus / ProQuest exports). The raw search exports themselves are NOT included here because of publisher redistribution terms.

Verification

file_inventory.csv — one row per deposited file with path, bytes, n_columns (CSVs only), sha256, and a short notes field. Use the SHA-256 to verify a downloaded file has not been corrupted.

Contact

G.P. Williams, Department of Civil and Construction Engineering, Brigham Young University, Provo, UT 84602, USA. ORCID: 0000-0002-2781-0738.