Reproducibility and Corpus Certificates
Why certificates?
Every bibliometric analysis starts with a corpus, but most analyses don’t document how the corpus was built. scimapR’s corpus certificates solve this: they are structured YAML documents that record the corpus hash, the steps behind the corpus and the local files it was built from, so that another researcher can check a corpus against them and run its searches again.
Provenance tracking
Every record in a scimapR corpus knows where it came from:
head(sm_provenance(corpus))
#> # A tibble: 6 × 8
#> work_id source source_id_external fetch_date query engine
#> <chr> <chr> <chr> <dttm> <chr> <chr>
#> 1 W000000001 synthetic <NA> 2026-11-03 18:20:50 sm_example… native
#> 2 W000000002 synthetic <NA> 2026-11-03 18:20:50 sm_example… native
#> 3 W000000003 synthetic <NA> 2026-11-03 18:20:50 sm_example… native
#> 4 W000000004 synthetic <NA> 2026-11-03 18:20:50 sm_example… native
#> 5 W000000005 synthetic <NA> 2026-11-03 18:20:50 sm_example… native
#> 6 W000000006 synthetic <NA> 2026-11-03 18:20:50 sm_example… native
#> # ℹ 2 more variables: scimapR_version <chr>, prompt_hash <chr>Corpus hashing
Creating a certificate
cert <- sm_certificate(corpus)
#> ✔ Certificate created. Corpus hash: 61c0637eede7
str(cert, max.level = 1)
#> List of 20
#> $ certificate_version: chr "1.0"
#> $ created : POSIXct[1:1], format: "2026-11-03 18:20:51"
#> $ scimapR_version : chr "0.4.0.9000"
#> $ r_version : chr "4.6.2"
#> $ platform : chr "unix"
#> $ corpus_hash : chr "61c0637eede72e86bf489cf7d8a27949789b82485ea6646cc707da9ec1945bb6"
#> $ hash_version : int 2
#> $ n_works : int 200
#> $ n_authors : int 80
#> $ n_institutions : int 0
#> $ n_references : int 1869
#> $ year_range : int [1:2] 2015 2024
#> $ question_id : chr NA
#> $ queries :List of 1
#> $ provenance_summary :List of 1
#> $ screening_summary : list()
#> $ embedding_info :List of 13
#> $ local_inputs : tibble [0 × 9] (S3: tbl_df/tbl/data.frame)
#> $ is_locked : logi FALSE
#> $ metadata :List of 2
#> - attr(*, "class")= chr "sm_certificate"Verifying a certificate
result <- sm_verify_certificate(corpus, cert)
result$pass
#> [1] TRUE
result$checks[, c("check", "expected", "actual", "status")]
#> # A tibble: 5 × 4
#> check expected actual status
#> <chr> <chr> <chr> <chr>
#> 1 corpus_hash 61c0637eede72e86 61c0637eede72e86 pass
#> 2 n_works 200 200 pass
#> 3 n_authors 80 80 pass
#> 4 n_institutions 0 0 pass
#> 5 n_references 1869 1869 passLocal input files
A corpus built from local files, such as a database snapshot, an
institutional export or an affiliation dictionary, records each file
with its SHA-256 hash. sm_corpus_from_tables(inputs = )
registers the files the tables come from;
sm_register_input() adds any other file.
tracker <- tempfile(fileext = ".csv")
writeLines(c("doi,unit", "10.9999/example.1,Unit A"), tracker)
corpus <- sm_register_input(corpus, tracker, role = "publication tracker",
builder = "R/01_import.R::read_tracker")
cert <- sm_certificate(corpus)
#> ✔ Certificate created. Corpus hash: 61c0637eede7
sm_verify_certificate(corpus, cert)$checks[, c("check", "status")]
#> # A tibble: 6 × 2
#> check status
#> <chr> <chr>
#> 1 corpus_hash pass
#> 2 n_works pass
#> 3 n_authors pass
#> 4 n_institutions pass
#> 5 n_references pass
#> 6 input:file1e971bff7494.csv pass
# once the file changes, its check fails
cat("10.9999/example.2,Unit B\n", file = tracker, append = TRUE)
sm_verify_certificate(corpus, cert)$pass
#> [1] FALSEA file that verification cannot find is reported as
"not checked", with a warning; its relative path is looked
for in base_dir, in the folder of the certificate file and
in the working directory. sm_rebuild_from_cert() replays
the recorded searches of OpenAlex, PubMed and Crossref, but it cannot
replay local files or file reads: for a corpus built only from them it
stops and names the files and the code that read them.