Skip to main content

CDSL Newsletter — August 2026

Here is the August 2026 edition of the Cologne Digital Sanskrit Dictionaries newsletter, covering the activity of the past month. Significant work this period: a complete literary-source overhaul of the Benfey dictionary, a second and much deeper security hardening pass across the web interfaces, a multi-dictionary endpoint for the new structured API, and the preface digitisation effort maturing from raw OCR into machine-readable abbreviation data.

Benfey dictionary: literary sources linked, sub-headwords promoted​

Benfey's Sanskrit-English Dictionary (BEN) received a month-long structural overhaul. Its literary-source references — previously plain abbreviations in the text — were normalised to the same consistent markup PWG uses, covering bare references, page numbers, and dozens of per-work adjustment passes (Śākuntala, Hitopadeśa, Nala, Kirātārjunīya, and many more). Sub-headwords were then promoted to full headwords, and homonym numbering was added. Thirteen markup and correction issues were closed in the BEN repository in July, and the headword normalisation databases (hwnorm1, hwnorm2) were regenerated to pick up the new headwords. As a result, BEN's literary sources are now linked in the online display, like PWG and PW before it.

Security and reliability: the July sweep of the web interfaces​

Following June's static-analysis scan, July went considerably deeper. Every display surface of the two web frontends (csl-websanlexicon and csl-apidev) now escapes dynamic HTML; standard security headers were added to all server entry points; the bundled jQuery copies across the display sites (GRA, SKD, MW demo pages, mw-dev, the verb tables) were brought up to 3.7.1; and the abandoned jquery.cookie plugin was replaced by js-cookie throughout. The same sweep fixed long-standing user-facing bugs: scan pages now always offer a direct open-PDF link (helping Firefox on Android, where embedded PDFs never rendered), the duplicate display of identical MW entries is gone, Advanced Search highlighting now handles Unicode correctly, and AP90 compound headers display properly. Twelve issues were closed in csl-websanlexicon and six in csl-apidev during the month.

Salt API: one request, many dictionaries​

The structured API introduced last month gained a multi-dictionary endpoint: a single request can now query several dictionaries at once, with parameters for field selection, result size and ordering, and optional Devanagari output (requested in csl-orig issue #2881, "single interface for online requests"). On the user-facing side, the lookup interface shipped its third feature wave: links to the scanned page images, a copy-citation button, a Vedic-accent toggle, and a dark mode.

Prefaces: from scanned pages to machine-readable abbreviation data​

The front-matter digitisation reported in the previous two editions moved from raw OCR to structured data. Page-boundary OCR omissions were repaired for the Böhtlingk–Roth volumes (PWG, PW, PWK), abbreviation keys were aligned to the naming used in the body texts, and machine-readable "legend stores" — the abbreviation legends of each dictionary as data — were emitted for PWG and PW, with a schema and parity checks. The Abbreviations page on the guides site now consumes these legends directly. A new repository was also created for the front matter of the Deccan College dictionary (PD), with its own OCR pipeline.

Measuring the collection: the dictionary atlas​

The csl-atlas repository, which studies the dictionary collection quantitatively, shipped fourteen releases in July. New analyses include a per-dictionary "entry anatomy" radar chart, a data-richness typology grading every dictionary from L0 to L10, a census of orthographic drift in the glosses, and a measured study of completion horizons — how long large dictionaries actually take to finish — comparing the in-progress Deccan College dictionary with AP, PWG and MW. All datasets and visualisations are committed and reproducible in the repository.

KeySwap: a typing helper for IAST​

The sanskrit-util repository grew a new tool: KeySwap, a cross-platform typing helper for entering IAST transliteration (Windows via AutoHotkey, with iPhone and Mac recipes). Type a plain letter and cycle it through its diacritic variants; smart digraph handling, an optional live headword check against the Cologne server, an offline SLP1 wordlist fallback, and an opt-in frequency-ranked mode based on the Digital Corpus of Sanskrit are included. Nine releases (0.8.0 through 0.8.11) landed in July. The library also fixed its iast_to_devanagari function, which previously produced broken output for consonant+vowel sequences.

Text corrections: the daily cadence continues​

The daily correction workflow ran through the month: the daily batches through 10 July were filed and closed (seventeen daily-correction issues), a consolidated AP90 batch was promoted into the source repository, and the offline StarDict dictionaries were regenerated automatically after every correction commit. In the central COLOGNE tracker, twenty-one long-standing discussion issues were reviewed and closed.

To report an error or suggest a correction, open an issue in CORRECTIONS.

To receive future editions by email, subscribe here.