Changes to output format

[cmungall]

Related to #11.

Refer to sssom for good practice

• use lowercase
• split entity into two columns
  • ID
  • Property (synonym, label, etc)
• split origin
  • source_version (e.g. "version" : "http://purl.obolibrary.org/obo/go/releases/2021-07-02/go.owl")
  • source_id (e.g. go)
  • source_file (e.g. go.json)
• what is zone?
• sentence id:
  • this is a bit opaque
  • should there be an intermediate file created that is one line per sentence with document ids and sentence ids as the first two columns?

[hrshdhgd]

• what is zone?

This is something OGER spits out in the output. I have no clue what it represents. The documentation does not specify it's significance either. I'll keep looking.

• sentence id:

So when the text has multiple sentences, S1 is Sentence 1, S2 is Sentence 2 so on and so forth. It basically splits sentences by the separator (. for example) and assign these IDs to sentences.

[hrshdhgd]

I looked in the code for OGER and it seems that 'zone' is represented by section_type in the code. This is relevant to clinical notes. I have seen in this in the past that when clinical text is recorded in EHRs , there are sections in the text represented in all caps (for e.g. DIAGNOSIS, TREATMENT PLAN etc.). This basically highlights that. In our case it will always be blank.

[cmungall]

there may be analogs, e.g. a typical journal article will be structured, maybe it is also for the structure/section heading, e.g. methods, abstract, ...?

[hrshdhgd]

That makes sense.

[cmungall]

currently the match_field is sometimes empty sometimes filled

let's change to reuse sssom data dictionary where possible

• object_id: (currently entity_id). The ontology term id that was matched
• object_label: (currently sometimes this is in match field). This is the primary label of the object_id, regardless of whether the match was on the label or synonym
• object_category (currently "type")