Andornot Blog

Image Annotator

Andornot — Wed, 25 Feb 2026 18:16:10 GMT

Andornot has made an Image Annotator desktop application! It's for creating and editing annotation layers on large-format images, such as maps, architectural drawings, manuscripts and other high-resolution images. Annotations can be saved and later displayed to the public in the Andornot Discovery Layer web application, which now supports annotations in its Zoomable Image Viewer.

In the Annotator you can draw a variety of shapes, pointers and labels overtop an image, customize each shape's fill colour, stroke colour and opacity, and provide a detailed description about the shape.

Supported shapes include:

rectangle
polygon
ellipse
freehand path (multi-segmented line)
arrow
formatted text label

Technical Details

The Image Annotator is a Vue 3 application running in Electron, the desktop Chromium/NodeJS browser. Image zooming is handled by OpenSeadragon, and the annotation library is based on Annotorious. The annotations are stored in JSON format and are a hybrid of W3C (text labels) and Annotorious standards (shapes) for performance reasons, but can be easily crosswalked to IIIF.

A Peek Behind the Curtain

Hi, it's Peter. I'm the lead developer at Andornot. I have to tell you - this was a lot of fun to make! It's the first time I used Claude Opus in a serious way on a sizeable project from beginning to end, and I thoroughly enjoyed it. I have been programming for 25+ years, and coding with an AI assistant has made things so interesting: it's so smart and so dumb at the same time. If programming before AI was like plowing with a mule, now it's like putting a harness on a jumbo jet: you're going to get a lot done but you also need to know what you're doing so it doesn't introduce performance or security flaws. Or, you know, plow a six foot furrow a mile long through your farmhouse.

Is AI now good enough at handwriting recognition to use with your archives?

Andornot — Thu, 04 Dec 2025 23:09:18 GMT

This recent article (https://generativehistory.substack.com/p/gemini-3-solves-handwriting-recognition) suggests that AI is now at a level where it can reliably transcribe handwritten text, something Andornot has monitored for years. Our Andornot Discovery Interface excels at searching the full text of documents, but only if those documents contain a text layer. A text layer is inherent in born-digital documents, and for digitized or scanned ones, can be created through an OCR (optical character recognition) process. But for years, no tool we tried was quite good enough at recognizing handwriting to warrant the cost and effort. The error rate was too high and the quality of the textual output too poor to support meaningful full text searching.

Now, however, the authors of this article suggest they the latest version of Google's Gemini 3 is more than good enough. If true, then running it against the hundreds, thousands or millions of handwritten documents held in archives could make a wealth of new materials and information more available to researchers, without the effort of human transcription. Exciting!

We put this to the test and found that if we, as humans, could read the handwriting, then so too could Gemini and that the textual output was excellent, as good as a human would achieve.

But for handwriting that is hard for a human to read, because it's a looping, slanted, sideways swirl of ink across a page, barely recognizable as letters, so too did Gemini struggle.

So, is it worth it, or better to wait a bit longer, before working on your own collections? If the handwritten materials in your collections are legible, we'd say the time has come. But if you have to rotate the page 90 degrees, then again, then again, trying to figure out which way is up, what language it's written in, if the lines are even letters or maybe a treasure map, better to wait. When the day comes that AI exceeds human handwriting recognition, which is surely not far off, then it'll be even more worthwhile to use.

Wrangell Newspaper Archive

Andornot — Thu, 12 Dec 2024 16:13:23 GMT

A digitized collection of newspapers from the city of Wrangell, Alaska, dating from 1898 to the present, is now available online at https://wrangellnewspapers.andornot.net

This full-text searchable site is powered by our Andornot Discovery Interface and allows users to search for people, places, events and more using keywords, then refine their search by newspaper name and date of publication.

Sophisticated algorithms ensure the most relevant issues appear first in search results.

Each issue is presented as a PDF, viewable immediately and with search words highlighted for quick navigation to pages and articles of interest.

A flipbook-style viewer is also available, allowing users to simulate on-screen the act of turning the pages of the physical edition.

The collection includes the historical newspapers The Fort Wrangell News (1898), The Stikeen River Journal (1898-1899), The Alaska Sentinel (1902-1909), and the Wrangell Sentinel (1909-present). The Sentinel is the oldest continuously published newspaper in Alaska. The collection will be updated annually to ensure continued access to Wrangell's ongoing story.

Andornot hosts the site as part of our Managed Hosting service.

Riverview Hospital Artifact Collection

Andornot — Wed, 11 Dec 2024 17:57:59 GMT

The Riverview Hospital is a significant part of the City of Coquitlam's history. The mental health hospital was constructed on 1,000 acres of land within the unceded and ancestral territory of the kʷikʷəƛ̓əm (Kwikwitlem) people. When it opened in 1913, it was known as the Essondale Branch of the Provincial Hospital for the Insane and consisted of a single asylum building with extensive grounds and a productive farm known as ƛ̓éxətəm (formerly Colony Farm). Over the years the hospital grew, eventually consisting of dozens of building and housing thousands of patients. In 1965, it was renamed the Riverview Hospital, which remained its name until it closed in 2012. In 2021, the Riverview lands were renamed səmiq̓wəʔelə, which translates to Place of the Great Blue Heron in hən̓q̓əmin̓əm.

The Riverview Hospital Artifact Collection was originally compiled by staff and former staff of the Riverview Hospital. A staff committee started collecting historical artifacts from the institution in the 1980s, and this group was officially incorporated as the Riverview Hospital Historical Society in 1997. They collected equipment, furniture, objects and documents to preserve and share the story of the hospital, its staff and its patients.

When the hospital closed in 2012, the City became the custodians of this collection, and in 2024, it became fully searchable online at https://www.riverviewhospitalartifacts.ca, powered by our Andornot Discovery Interface, with back-end management using Inmagic DB/TextWorks software.

30 Years of Helping Cultural Institutions

Andornot — Thu, 05 Dec 2024 18:30:27 GMT

In 1995, our founding partners teamed up to sell and support Inmagic software and provide consulting services, database design and hosting to libraries, archives, museums, and a diversity of other clients.

Floppy disks! DOS! CRT Monitors! Horse-drawn modems! Such were the tools our intrepid pioneering partners used to build their business, and then oh boy, along came the internet. So they adapted and innovated. It's been constant adaptation and innovation ever since, to be honest. We are now experimenting with AI, named entity recognition and machine transcription or oral history recordings, to name just a few current projects.

In 2025, we celebrate our 30th year in business and feel incredibly grateful for the support of our clients, and the interesting projects we’ve been privileged to work on.

To all of you, we express a heartfelt Thank You for the wonderful opportunities you have provided us over these many years.

While we're celebrating this milestone anniversary, we also say a fond farewell to our partner Kathy Bryce. Kathy has worked with so many of you at one point, while guiding the business and mentoring new staff and partners. Kathy will formally leave Andornot at the end of 2024 and enjoy a well-deserved retirement, working in her large garden and on local community initiatives, especially those related to technology training for seniors.

Peter, Jonathan and other members of the Andornot team look forward to providing all the same service and support you are used to.

Norfolk County Newspapers

Andornot — Fri, 29 Nov 2024 00:25:52 GMT

The Heritage and Culture department of Norfolk County, which includes the Archives, has digitized over 60,000 pages of county newspapers from the 1860s to the 1970s, preserving the activities and changes of the county over that century. This collection is fully searchable online at https://norfolkcountynewspapers.andornot.net, a site powered by our Andornot Discovery Interface.

Users may search for people and places by name, and search by any words found in the text, then limit their results by year, decade and newspaper name. Each page may be immediately viewed online, downloaded, or shared with friends and family by email or on social media.

Norfolk County joins Bruce County and Glengarry County in making newspaper collections searchable in this way, through dedicated sites, and other archives that Andornot works with where newspapers are searchable alongside other archival collections.

Using AI Tools to Transcribe Oral History Recordings

Andornot — Sun, 01 Sep 2024 14:54:12 GMT

Traditionally, oral history recordings made by archives and museums were stored on tape, then perhaps digitized, and now are of course recorded digitally initially, often saved in MP3 format for public access.

Descriptive information such as the names of the interviewers and interviewees and interview date, plus keywords or subject headings representing people, places and topics mentioned in the interview may be recorded in a database record along with the digital file name or original physical format. Time permitting, individual passages within recording may be tagged with timestamps and keywords, using a tool such as the Oral History Metadata Synchronizer.

For researchers interested in finding and listening to oral histories on specific topics or places, they are highly reliant on the descriptive cataloguing, as most search engines don’t have access to transcriptions of the interview. Such transcriptions have traditionally been very time consuming to create, requiring staff or volunteers to carefully listen to each recording (i.e. hours of listening) and accurately type the text into a document.

Fortunately, the current level of machine listening and transcription is such that these hours of work can be outsourced to computers. For recordings where the language is predominantly English and where the speech is clear, the accuracy of the transcribed text is as close to 100% as a human could be performing the same work manually. When the recording has audio artifacts, or the speakers are less clear, the accuracy declines, but again, in the same way a human transcription would be less than perfect.

For languages other than English, the results vary from language to language, but may still yield searchable results, even if not a perfect word for word transcription.

When a recording has been transcribed to text, that text can then be indexed into and made searchable through a search engine such as our Andornot Discovery Interface.

Machine transcription of oral history recordings is akin to optical character recognition of digitized print materials, both yielding a large quantity of text that helps users find resources of interest. Further steps can be taken on both these kinds of text to automatically identify the names of people and places within them, which can then be used to limit a search, or to create browsable indexes of terms.

If you have existing oral history recordings, physical or digitized, as well as print materials, that you’d like to make searchable, or make more accessible, contact Andornot to discuss using tools mentioned in this blog post.

Andornot’s 2024 Summer Intern

Andornot — Sun, 01 Sep 2024 14:40:51 GMT

Andornot has a long-standing commitment to professional development and to supporting our colleagues and clients in the cultural heritage sector. We attend trade shows and conferences and sponsor other events either through in-person booths or financial support to help the event proceed. For many years we provided an annual professional development grant to help clients attend conferences themselves.

This year we took on a summer intern, both to assist us with some research and development projects, and to provide some practical tasks and professional guidance to a new graduate. Our intern this year is not a graduate of a traditional library or archival studies program, but rather has just completed a Master of Data Science in Computational Linguistics degree, a field that is burgeoning but also of great interest and applicability to the work that libraries, archives and museums do.

We tasked Jarret with several research projects, including:

using LLMs such as ChatGPT to improve the output of optical character recognition processes on digitized newspapers;
using machine learning tools to transcribe oral history recordings to better enable full text searching;
using named entity recognition tools to identify the names of people and places in digitized newspapers, other documents, and oral history recordings, to allow researchers to browse term lists and refine searches using facets, in addition to full text searching; and
researching whether any current tools could improve searching of Indigenous language text.

Jarrett’s work on LLMs can be read here, while his work on named entity recognition is available in our blog here. Information on machine transcription of oral history recordings is availabe here.

We are grateful to have had Jarrett’s expertise this summer and will be making use of his research in upcoming projects.

Improving Search Filters with Named Entity Recognition (NER)

Andornot — Sun, 01 Sep 2024 14:33:36 GMT

Named Entity Recognition (NER) is a valuable tool for identifying names within a text, such as the names of peoples and places. Our summer 2024 intern, Jarrett MacFarlane, explored uses of NER with full text resources that have not been catalogued or described by a librarian or archivist. The intention was to extract names that could be used to filter a search, or to present a browsable index of names, for use in our Andornot Discovery Interface search engine. For the purpose of this project, he worked with local Ontario newspapers dating from the late 19th and early 20th century.

Jarrett developed an NER script which uses the natural language processing (NLP) library spaCy to automatically pick out names from documents. The script can then group similar names that are likely to refer to the same thing, and then can check place names against existing datasets of place names for accuracy.

Depending on the quality of some historical documents, names may appear in many variations, with misspellings, differing formats, or partial matches. For example, “M. A. Smith,” “Mary A. Smith,” and “Mary A. Srnith” (an OCR error) are likely to all refer to the same person.

Some challenges with performing NER on historical documents include:

Extracting names from historical newspapers where optical character recognition (OCR) may have introduced errors in spelling due to the poor condition of the original document.
Ensuring the accurate representation of place names in cultural collections.

Our NER approach addresses these issues by first processing the text using spaCy’s transformer model to extract the names. We can then group together names that are partial matches, or likely to represent the same thing, using a technique called Levenstein distance to identify likely matches. For example, in historical newspapers, “Wm. Hart” would be a common shortening for “William Hart.” Because these two forms of the same name share a lot of letters in common and in the same order, they will have a very close Levenstein distance, which means we can identify them as likely to be the same name and group them together. This allows us to eliminate minor variations in the final name filters, and ideally present the most relevant form of the name for user filtering. For place names, we can check them against external datasets of place names, such as the Canadian Geographical Names Database created by Natural Resources Canada, which includes the Indigenous Place Names dataset. This helps ensure that culturally significant names are properly represented.

Jarrett’s work with NER this summer will appear in upcoming projects with digitized newspaper collections, and is available to be applied retroactively to existing sites powered by our Andornot Discovery Interface.

Burnaby Art Gallery Upgrades Collections Search

Andornot — Fri, 16 Aug 2024 15:09:19 GMT

Andornot has worked with the Burnaby Art Gallery for over 15 years and originally converted data from a legacy system into DB/TextWorks databases used to manage the collections internally.

A publicly-accessible search interface to the collections was first launched in 2008 using Inmagic WebPublisher PRO software, then was upgraded to the Andornot Discovery Interface (AnDI) in 2012. The Public Art collection was added with some updates in 2015.

Only minor modifications have been made since then, so the site was due for an upgrade to incorporate the latest accessibility and security updates as well as new features that have been added to AnDI.

An upgraded site powered by the current version of AnDI was launched in 2024 at https://collections.burnabyartgallery.ca with a bright and welcoming home page. Images representing areas of the collection invite users to click to browse, while AnDI's search features allow researchers to locate items of interest from the gallery's permanent and art education collections, library, and public art registry, brought together into this single search interface.

As with the recently-upgraded Heritage Burnaby website also powered by AnDI, this site allows users to both search and display Indigenous terms, names, languages and places not only in their Anglicized forms, using Roman letters, but also using the APA or NAPA phonetic character sets. While the underlying DB/TextWorks databases cannot store the necessary characters natively, Andornot developed a means to substitute Roman letters with APA and NAPA phonetic character sets on-the-fly during indexing into AnDI, allowing both versions to be fully searched and displayed. For example, users may search for either Musqueam or ʷməθkʷəy̓əm, and view both terms side by side in all records that formerly contained only Musqueam.

Contact Andornot to discuss upgrades and enhanced features for your own heritage collections.