Settings and defaults

Each setting of each function, its default, and where the default comes from. The reference pages give every permitted value.

function	setting	default	source of the default
`harvest`	`min_words` / `max_words`	4 / 80	old cleaning notebook
`extract_dois`	`first_pages` / `save`	2 / `None`	the paper’s own DOI is on page 1 (checked by hand: 7 of 7)
`extract_references`	`save` / `cache`	`None` / `True`	214 of 215 DOIs of the sample papers exist at doi.org
`download_papers`	`licences`	`None`	`["cc-by"]` keeps only CC BY papers (see A snowball round)
`download_papers`	`sources`	`None`	`None` = the nine free sources, from `openalex` to `landing_page`; add `tmsr-doi-downloader`
`download_papers`	`respect_robots`	`True`	obey `robots.txt`; set `False` only with the right to do so
`download_papers`	`email`	`None`	the polite pool of OpenAlex and Crossref, and Unpaywall; `None` keeps `UNPAYWALL_EMAIL`
`causenet`	`threshold`	0.8	old pipeline sweep 0.5–0.8
`causenet`	`mode`	`neural`	model card default
`causenet`	`decision`	`span_only`	since 0.2.0 (`cls+span` before): a cause span and an effect span are enough
`causenet`	`avoid_ambiguous`	`True`	since 0.2.0: skip a relation with a pointer span (“this”, “it”)
`causenet`	`batch_size` / `device`	64 / `auto`	`auto` = an NVIDIA GPU, else an Apple GPU, else the CPU
all harmonizer	`embeddings`	`allmpnet`	Paper 3; `bge_base` scores 0.02 F1 higher (Thesaurus strategies, section 5)
`standardize_*`	`taxonomy`	`elsst`	project choice; the only built-in thesaurus
`standardize_*`	`strategy`	`enriched`	Paper 3 web tool; confirmed by the held-out test
`standardize_*`	`top`	5	project choice
`standardize_*`	`leaves_only`	`False`	since 0.2.0 (`True` before): broad concepts too, such as TRUST and MOTIVATION
`label_groups`	`by` / `top`	`keywords` / 3	since 0.2.0 (`both` before): the TF-IDF keywords of the group
`standardize_*`	`threshold`	`None`	since 0.2.0; `None` gives every span its best concept
`standardize_groups`	`name_by`	`mean_sim`	old pipeline’s cluster-naming comparison
`pairwise_grouping`	`threshold`	0.69	Paper 3 `BEST_PAIRWISE_THRESHOLDS`
`pairwise_grouping`	`linkage`	`average`	since 0.2.0 (`single` before): single linkage chains (see Group and name constructs)
`clusterer`	`min_cluster_size` / `min_samples`	2 / 2	Paper 3 `BEST_CLUSTERING_PARAMS`
`clusterer`	`umap`	`None`	Paper 3; the best run had no reduction
`causal_map`	`min_papers` / `min_sentences`	1 / 2	project choice; 1 paper works for one long book
`causal_map`	`min_share`	0.67	project choice: a two-way pair needs 2 papers to 1 for one direction
`draw_map`	`save` / `height`	`None` / 720	a notebook shows the page; a script writes `causal_map.html`

τ, the score that a span needs for a thesaurus concept, is the threshold of the standardize_* functions. It is off by default. Choose it per model, per strategy, per thesaurus. The evidence behind the defaults is in Thesaurus strategies, section 5.

Keyboard shortcuts

theoryminer

Settings and defaults