Unproven points about your family no longer get lost in notes

After a few months of family research, a lot of unproven material piles up. The cantor's son named in the 1795 confession roll vanishes from later documents: did he die or leave the household? A wife's age in a confession list does not match the revision. A search found a similar surname in a neighbouring parish, and you need to look at the entry yourself.

Until now all of this lived in notes. A month later the notes showed a closed question as open, and findings waiting for your look lay there for weeks.

Now each such item is a separate hypothesis in the family tree.

What a hypothesis records

• What exactly is claimed, in one sentence.
• What it rests on: one document or several, and whether you saw the entry itself or only the machine reading.
• Which document will settle it. For example, the village death register for 1795–1800 or the lists of the church school.
• Findings waiting for your word. No more than five, each with the address of the sheet and a crop of the line.
• For a closed hypothesis: what was decided, by whom and when, and which new document could reopen it.

Example:

What to decide first

Tell the assistant: “what do I need to decide about the family”. It will show no more than three decisions. The first ones are those that already have a finding and on which more people in the family tree depend. Each comes with a crop of the line from the scan.

You can answer in one word: “ours”, “not ours”, “unreadable”. Your answer is recorded in the hypothesis, the finding disappears from it, and what is confirmed goes into the family tree.

Disputed points are no longer recorded silently

When two documents contradict each other, such a fact cannot be recorded in the family tree without a hypothesis. Nyshporka checks this itself. So every disputed fact has an answer to the question of what will settle it.

Closed hypotheses also stay in the family tree. You decided that the Kovalenkos from the neighbouring village are a different family, and recorded why. The assistant sees this and will not start the same investigation a second time. Only a new document named in the hypothesis will reopen the question.

Newspapers, journals and directories: a separate print engine

A district newspaper of the 1930s, diocesan gazettes, a provincial address calendar. The engine for manuscripts glued together lines from neighbouring columns on a newspaper page, and an article was read in fragments.

Now a printed publication is read by a separate print engine on a rented machine. Columns and articles are separated and taken in reading order. Errors are below 1% per character.

• Arrange the run of issues in folders: Title/year/issue/ or simply Title/issue/. Write the issue the way you already call it: 3 (770) 1936-01-06, 2 1934-02-26, Додаток … 1935-08-12.
• Tell the assistant “read the run in the cloud”. It will ask for a short code of the publication, and if you do not know one, it will suggest its own.
• If a folder holds several newspapers, for example all the press of a city, the assistant will name them and suggest a code for each.
• After reading, the issues assemble into years by themselves, and a surname can be searched in the newspaper at once. A newspaper year goes to Supriaha as one book, not as a thousand separate issues.
• The assistant will show which issues are missing from your run.

Print is not read on your own computer: the engine needs a graphics card with 20 GB or more.

What else appeared after 0.28.0

• Share everything that has been read with Supriaha in one go. Say “share with Supriaha everything that is ready”. 1,079 files of ЦДІАК ф.224 went this way in 49 minutes.
• Before reading a file, the assistant first looks in Supriaha. If someone else has already read the file, the text will be there in a minute, not overnight.
• Bessarabian revision lists by village name. Name a village, and Nyshporka will tell you in which files of the 1824–1886 catagraphies it lies, on which sheets and on which film. The films of this fond are on a free mirror, and no FamilySearch login is needed.
• Where to dig in Galicia. The assistant knows where the Galician registers are (ЦДІАЛ, the archive in Ivano-Frankivsk, AGAD), the cadastres and tax books, and register indexes such as GeneTeka.
• Books from the archives of Lublin and Przemyśl are shown with the name of the archive, in the “Poland” group.
• Nyshporka in Docker. If it runs in your container and the assistant on the computer itself, search and line crops work through the container.
• Documentation is available in English, Polish, Romanian, French and German.

Fixed

• Cloud reading on Windows no longer fails at start when the Nyshporka folder has a Cyrillic name or the laptop runs on battery.
• Moving scans to a rented machine. Without storage, Nyshporka shows before renting how much time and money this will take. If it takes too long, it suggests storage, a smaller batch, or paying deliberately.
• A machine that fits your budget and time. Before, the estimate was calculated for a typical $3 and 14 hours, and suitable machines were rejected.
• Hundreds of small files (for example, newspaper issues) start in the cloud in minutes, not hours.
• A file made of scan links goes to the cloud whole, not as empty links.
• Newspapers of the same day. Different newspapers published on the same day, and an issue with a supplement of that day, no longer merge into one file in the file list.
• Manually linking a reading to a file no longer turns a read file into an unread one. If a link has already landed wrongly, the file is shown as read again after the update.
• Files with a letter in the inventory number (РДАДА 210-6д-199) are visible in the file list.
• A reference code with labels “ф.315 оп.1 спр.8433” is understood by search and notes, not only by file registration.
• Search in Docker works.
• Installation. Manuscript reading installs completely even when Git is not yet installed. The machine check tells you what is missing if Visual C++ is absent on Windows. If you installed 0.28.4 from scratch, tell the assistant “install the reading engines”.

How to update

nysh update — or ask your assistant to update Nyshporka. After the update the assistant will review what changed by itself and offer to create hypotheses from what already lies in your notes.