I Know Something You Don’t Know: The annotation saga continues…
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Over the past 20 years, the biodiversity informatics community has pursued components of the digital annotation landscape with varying degrees of success. We will provide an historical overview of the theory, the advancements made through a few key projects, and will identify some of the ongoing challenges and opportunities. The fundamental principles remain unchanged since annotations were first proposed. Someone (or something): (1) has an enhancement to make elsewhere from the source where original data or information are generated or transcribed; (2) wishes to broadcast these statements to the originator and to others who may benefit; and (3) expects persistence, discoverability, and attribution for their contributions alongside the source. The Filtered Push project (Morris et al. 2013) considered several use cases and pioneered development of services based on the technology of the day. The exchange of data between parties in a universally consistent way necessitated the development of a novel draft standard for data annotations via an extension of the World Wide Web Consortium’s Web Annotation Working Group standard (Sanderson et al. 2013) to be sufficiently informative for a data curator to confidently make a decision. Figure 2 from Morris et al. (2013), reproduced here as Fig. 1, outlines the composition of an annotation data package for a taxonomic identification. The package contains the data object(s) associated with an occurrence, an expression of the motivation(s) for updating, some evidence for an assertion, and a stated expectation for how the receiving entity should take action. The Filtered Push and Annosys (Tschöpe et al. 2013) projects also considered implementation strategies involving collection management systems (e.g., Symbiota) and portals (e.g., European Distributed Institute of Taxonomy, EDIT). However, there remain technological barriers for these systems to operate at scale, the least of which is the absence of globally unique, persistent, resolvable identifiers for shared objects and concepts. Major aggregation infrastructures like the Global Biodiversity Information Facility (GBIF) and the Distributed System of Scientific Collections (DiSSCo) rely on data enhancement to improve the quality of their resources and have annotation services in their work plans. More recently, the Digital Extended Specimen (DES) concept (Hardisty et al. 2022) will rely on annotation services as key components of the proposed infrastructure. Recent work on annotation services more generally has considered various new forms of packaging and delivery such as Frictionless Data (Fowler et al. 2018), Journal Article Tag Suite XML (Agosti et al. 2022), or nanopublications (Kuhn et al. 2018). There is risk in fragmentation of this landscape and disenfranchisement of both biological collections and the wider research community if we fail to align the purpose, content, and structure of these packages or if these fail to remain aligned with FAIR principles. Institutional collection management systems currently represent the canonical data store that provides data to researchers and data aggregators. It is critical that information and/or feedback about the data they release be round-tripped back to them for consideration. However, the sheer volume of annotations that could be generated by both human and machine curation processes will overwhelm local data curators and the systems supporting them. One solution to this is to create a central annotation store with write and discovery services that best support the needs of all stewards of data. This will require an international consortium of parties with a governance and technical model to assure its sustainability.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.000 | 0.003 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it