Decoding the Cauzin Softstrip: a case study in extracting information from old media
Bibliographic record
Abstract
Having content in an archive is of limited value if it cannot be read and used. As a case study of extricating information from obsolete media, making it readable once again through deep learning techniques, we examine the Cauzin Softstrip: one of the first two-dimensional bar codes, released in 1985 by Cauzin Systems, which could be used for encoding all manner of digital data. Softstrips occupy a curious middle ground, as they were both physical and digital. The bar codes were printed on paper, and in that sense are no different in an archival way than any printed material. Softstrips can be found in old computer magazines, computer books, and booklets of software Cauzin produced. However, managing the digital nature of these physical artifacts falls within the scope of digital curation. To make the information on them readable and useful, the digital information needs to be extracted, which originally would have occurred using a physical Cauzin Softstrip reader. Obtaining a working Softstrip reader is already extremely difficult and will most likely be impossible in the coming years. In order to extract the encoded data, we created a digital Softstrip reader, making Softstrip data accessible without needing a physical reader. Our decoding strategy is able to decode over 91% of the 1229 Softstrips in our Softstrip corpus; this rises to 99% if we only consider Softstrip images produced under controlled conditions. Furthermore, we later acquired another set of 117 Softstrips and we were able to decode nearly 95% of them with no adjustments to the decoder. These excellent results underscore the fact that technology like deep learning is readily accessible to non-experts; we obtained these results using a convolutional neural network, even though neither of the authors are expert in the area.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.015 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".