"Same as above": Referencing information content in CASTEMO knowledge graphs

Referring to the content of another statement

Basic logic

Statements often make references to other statements to express that the content is the same, different, or related in other ways. In those cases, people are not saying the same statements obviously: they are talking in a different time, and they are also different from the original utterers; just their own statements which the source claims have the same content.

If it is important to note down such chains of reported speech in your research, we recommend a series of properties recording the stated relation to other content.  

Syntax of links to similar content

Let's take the example: "Witness Jane said the same thing as witness Paul." You coded Peter's relevant content in three different statements, let's call them PS1, PS2 and PS3.

You will create two statements: "Witness Jane said" (JS1), and "the same thing as witness Paul". The links will be built in the referring statement "Jane said". In JS1, Jane will be the subject, and the slot for speech content (e.g., actant 2) will be filled in with JS2. JS2 will not be modelled syntactically, but within JS1 (sic), you will attach three properties to it, declaring that JS2 (actant) has same content as, in turn, PS1, PS2 and PS3.

You might want to do some replacements in the "same content" link. For example: "Witness Jane confessed about herself the same things as witness Paul about himself." Now in this case, if it si relevant for your project, you might want to preserve the backlink to same content, rather than repeat them as if Jane's testimony actually listed them, but at the same time, you need to say that Jane is confessing them about herself, not about Peter who features as actant of the original PS1, PS2, and PS3. You can code such things in CASTEMO very easily by setting a replacement rule (a property with proptype C "replacement rule"), which is composed of two subproperties: replacement condition value property, and replacement value property. The former is setting what should be replaced, and the latter for what new entity it should be replaced. E.g., in our example, you might want to take over content from Peter's testimony, but within this same content link, you might want to exchange Peter for Jane, telling your query or data projection script (that you need to develop yourself, this is not InkVisitor's work!): take over Peter's content, but within this link, consider it as concerning Jane. You can see a real-life example in the following screenshot.

While you can code very complex "similar content" references in CASTEMO, (1) it is not always useful, and (2) you need to have a way to extract them back from the coding into the data you will use for analysis.

Differentiate degrees of content similarity

The text (or you, at the inferential epistemic level) can posit different degrees of similarity, which should be properly differentiated. It can be really C same content, but also more generally only C similar content, and for that matter, the text (or you, at "inferential" epistemic level) can also posit that something is C partially similar content, or even C different content. That is, create a set of content similarity Concepts, relate them within a proper taxonomy, and use them to relate statements as required by your research.

Epistemic level in the referencing property

Referring to temporal, spatial and circumstantial information of another document or statement

Referring to temporal, spatial and other circumstantial information of another document ("Written in the same place, on the same date, and with same witnesses present") is a different case - this is not taking over content, but some circumstantial details. These details can be further modified: in the same year as above, but 1st October. In these cases, [... to be continued ...]

The more complicated is the encoding (data-in), the more complicated is writing the script for decoding (data-out). Consider your resources. InkVisitor is a data collection tool: you need to extract the right data set for analysis from it with your own means.