Protocol: how the pool of sources was built

This file records every search, count and rule of the document-model review. Shared rules: evidence README. Searches ran on 2026-10-29. Terms are defined in frame.

1. Why grey literature

Garousi 2019, Table 4, answered for document-model: all seven questions get a yes.

# Question Answer for this review
1 Complex, not solvable from formal literature alone? yes: the rules for nesting, reuse and cycles live in standards and tools (XML, XInclude, DITA, TEI)
2 Lack of volume or quality of evidence? yes: no reader study of reuse was found; the corpus studies are from 2002 to 2011
3 Context important? yes: literary encoding, technical documentation and wikis meet different structures
4 Corroborate research with practice? yes: the OHCO critique against the standards that kept the tree
5 Challenge practice with research, or the reverse? yes: reuse was standardised without a measured effect (position.md)
6 Would a synthesis serve both communities? yes
7 Large practitioner volume? yes: prose linters and document tools with thousands of stars each (state-of-the-art.md)

2. Questions

The six sub-questions Q1 to Q6 are stated in frame. In short: document models and their recursive rule; text that is not a tree; a unit with two parents; self-reference and cycles; checking by program; the cost of strict models.

3. Existing reviews (Garousi Guideline 2)

  • Searched first, as strings F7, F13, F25 and the review terms of F26, and by the leads the commissioner named.
  • Terms: integrative literature review, systematic literature review, mapping study, meta-analysis, review, with component content management, content strategy, design patterns, overlap, schema languages, automated writing evaluation.
  • Found: 11 reviews and taxonomies; 2 read in full (DeRose 2004; Batova 2014 in its abstract and method), 4 by abstract, 5 by title and venue. They are placed in review.md strand 0.

6. Stopping rule

  • Stated before searching: effort bounded, top N per string as in section 4 (15, 12, 10 or 8), with saturation, k = 5 consecutive results that add no new concept.
  • Formal: evidence exhaustion for the small sets (F1 to F3, F5, F15, F17, F22, F33); effort bounded for the rest.
  • Grey web: all results shown per string; producer visits stopped when every system named in the frame had been read once.
  • Q/A: as in the shared protocol, k = 10 within a string; the run stopped after NM13.

7. Selection

  • Include: a source on how documents are modelled as trees or graphs, how units are reused, how cycles are handled, or how structure is checked by program. English, and readable on the search date.
  • Exclude:
    • uses of the words in other fields (geology, chemistry, medicine, user-interface frameworks);
    • query languages, storage and database mapping of XML with no bearing on the document model;
    • vendor pages selling a content management product;
    • sources whose text and abstract could not be reached and whose title does not settle relevance (located only).

8. Quality of grey sources (Garousi Guideline 11, Table 7)

  • The 20 criteria and the threshold: evidence README.
  • Outlet tier: books and standards = tier 1 (1); project documentation, specifications of community projects, pull requests, wiki and Q/A pages = tier 2 (0.5); company blogs = tier 3 (0).
  • Impact is judged from stars, the size of the user base, or votes and views.
  • Threshold: 10 of 20. Below it a source stays only as an example of practice; the text that cites it says so.
  • A part of a source that was not seen scores 0 on its criteria (Gamma 1995, read only through its table of contents and a secondary page).

9. Extraction and synthesis

  • Extracted per source: kind, year, the claim, the evidence offered, any number with its location, and whether it was read in full, by abstract or by title.
  • Synthesis is narrative, by question Q1 to Q6, and every claim traces to a row of sources.

10. Limits

  • One first rater for document-model; the second rater's blind sample is reported in the evidence README.
  • Shared limits: evidence README. Here: arXiv refused some calls on the day.
  • Several works were reachable only by title and venue: publisher sites refused automated access, and no identity was sent to any service to reach them. Those works carry no number in the text.
  • Forward snowballing covered two works; backward snowballing three.