Analyse a source

This tutorial takes one source, such as a letter book you have transcribed or a PDF of an edited cartulary, through its first analysis. By the end you will have a reading of the source with its people, places and events, and you will know how to change the way the source is divided and read it again. The terms used here (reading, unit, mention, entity) are explained in How analysis works.

Before you startCopy Anchor Link

You need a library item with text. Analysis reads an item's extracted text, from a PDF or another file, or its transcription. If the item's Analysis tab offers no Analyse… button, the item has no text yet, so add a file with text or transcribe its pages first. A file added a moment ago may still be being read.

A reading uses your daily model allowance, and a long volume can use more than half of a day's default allowance, so check Plans, limits and usage before reading a large source. In an organisation's project you need the editor or owner role.

Step 1: start the readingCopy Anchor Link

1. Open the item in your library and choose its Analysis tab.
2. Choose Analyse…. Pangur asks what you would like to analyse.
3. Choose Analyse just this item. (Select more sources is for reading several sources together; see Work with corpora.)
4. Pangur creates a corpus named after the item, containing only this source, and opens it on its Sources view with the Analyse this corpus dialog already open.
5. Read the estimate. For each source the dialog shows the model that will read it and the estimated cost, and a total line puts them together, for example "1 source would be read, costing about $0.40."
6. Choose Confirm and read.

Every analysis in Pangur's web client runs through a corpus, even for a single source. The corpus is only a grouping, so you can delete it later without losing the reading.

Step 2: follow the readingCopy Anchor Link

Pangur's server reads the source in the background, so you can leave the page. The item's Analysis tab shows the progress: first a line about finding the divisions of the document while the text is being divided, then a line such as "Reading 84 units — 40% through." while the model reads, with a running count of the mentions found. On the corpus page the source's status reads "reading…" until it is done, and the corpus's views fill in as the reading lands.

Step 3: read the resultCopy Anchor Link

When the reading is finished, the item's Analysis tab begins with a summary line, for example "Read as letter book · 212 units · 148 entities", followed by the actual cost once Pangur has it. A second line says how the source was divided, either at each line beginning with the marker, or in even sections because no divisions were marked.

Below that, the tab lists the entities by kind (People, Places, Organisations and so on), each with its mentions, roles and quotations. Show in the text opens the document at a passage. The units are listed with their page or leaf, so you can check where each one begins.

A panel then reports how far the reading has been reviewed, for instance that nobody has been through it yet and six mentions want a look. Review this reading opens the review desk.

If the reading did not finish, the tab says why. A partial reading keeps what was read and tells you how far it got, for example "12 of 30 batches were read before the daily allowance ran out." Read it again when your allowance has reset. A failed reading gives the reason, such as the item having no indexed text, and offers Try again.

Step 4: check how the source was dividedCopy Anchor Link

The division decides what counts as one record, so it matters for every view. Pangur's first reading uses the document's own headings and numbering, or even sections if there are none. A letter book whose letters each begin "To the same" or "Sir" is often better divided at those words, and you can tell Pangur so.

After the first reading, the item's Analysis tab offers How it will be divided.

1. Open How it will be divided.
2. In Start a new unit at each line beginning, type the words that open each unit, such as To the same. The marker can be up to 80 characters long.
3. Choose Show the division. Pangur divides the text at once, without a model, and reports how many units the marker produces, or that no line begins with it. It lists the first eight units with how each one opens and its length, and Show all lists every unit.
4. The estimate line above the preview updates to the new division, showing the model, the estimated cost and the number of units.
5. To return to the document's own structure, choose Use the document's own divisions.

Trying a marker costs nothing. Only reading the source again does.

Step 5: read the source againCopy Anchor Link

Choose Read again on the item's Analysis tab. Pangur asks you to confirm and explains that a new reading is made from scratch, that your corrections carry over, and that anything new is unreviewed. Choose Read it again.

The new reading uses the marker you chose in How it will be divided, or, if you did not choose one, the marker from the last reading. The previous reading stays in the corpus views until the new one is finished. Corrections whose passage the new reading does not contain are listed under Corrections without a place (corrections that outlive a reading).

Filing the units as referencesCopy Anchor Link

A volume of letters is one library item, but you may want to cite each letter on its own. After a reading, File as references… on the item's Analysis tab turns units into library items. The dialog lists the units, marking any that are already filed so that they are skipped. Tick the units you want and choose File with the count, such as File 24.

Each unit becomes a top-level item in your library of a fitting type, such as a letter, a manuscript or a book section, linked back to the volume. If the volume is in your catalogue, the dialog tells you that the new items are also added to the catalogue under the volume's title; otherwise it says that only library items are created. Filing happens only when you choose it.

Exporting one source's readingCopy Anchor Link

The Export menu on the item's Analysis tab downloads the reading. The formats are the same as for a corpus export.