examples: add author_comparator simple demo

[chris-asl]

Signed-off-by: Chris Aslanoglou [email protected]

---

I've created this example so that we can see the behaviour of the AuthorComparator ontop of inspire-json-merger.
For making it easier to identify whether the comparator matched the names, I changed the merger_config for this example.

[michamos]

For context: @ksachs @jmartinm the AuthorComparatoris something that is running deep inside the merger, and cannot currently easily used independently. Here @chris-asl and @ammirate configured the json-merger that internally uses the comparator to demonstrate in isolation that it works.

[ksachs]

Do we agree on what should match and what shouldn't?
Do you take name-variants into account?
What about initials vs full first name?
Can we deal with (UTF8) special characters?

E.g.
'Zwikel, C.' should match
(can we make 'Zwikel, Celine' match)

"full_name": "Zwikel, C\u00e9line", 
"name_variations": [
    "C Zwikel", 
    "C\u00e9line Zwikel", 
    "Zwikel", 
    "Zwikel C", 
    "Zwikel C\u00e9line", 
    "Zwikel, C", 
    "Zwikel, C\u00e9line"
]

'Gaspari, Massimo' should match (it should be better than not matching)

"full_name": "Gaspari, M.", 
"name_variations": [
    "Gaspari", 
    "Gaspari M", 
    "Gaspari, M", 
    "M Gaspari"
]

'Gaspari, Massimo' should not match

"full_name": "Gaspari, M.", 
"name_variations": [
    "Gaspari", 
    "Gaspari M", 
    "Gaspari Maria", 
    "Gaspari, M", 
    "Gaspari, Maria", 
    "M Gaspari",
    "Maria Gaspari"
]

[kaplun]

"full_name": "Gaspari, M.", 
"name_variations": [
    "Gaspari", 
    "Gaspari M", 
    "Gaspari Maria", 
    "Gaspari, M", 
    "Gaspari, Maria", 
    "M Gaspari",
    "Maria Gaspari"
]

How can the full_name sport initials, when the name_variations have Maria?

[michamos]

AFAIK, name_variations are used for searching on author as opposed to exactauthor but are not used by the AuthorComparator. They are derived from full_name automatically, so do not contain additional information: https://github.com/inspirehep/inspire-next/blob/f64f0eee6a5a2c0eb996352f1f10371eb64cf530/inspirehep/modules/records/receivers.py#L369-L390.
So only the full_name needs to be considered. And Gaspari, Massimo should match Gaspari, M. but not Gaspari, Maria indeed.

[ksachs]

bad made-up example. Most likely it would be

"full_name": "Gaspari, Maria",   
"name_variations": [
    "Gaspari", 
    "Gaspari M", 
    "Gaspari Maria", 
    "Gaspari, M", 
    "Gaspari, Maria", 
    "M Gaspari",
    "Maria Gaspari"
]

[ksachs]

if name variant can not be used in the AuthorComparator maybe we should strip everything to initials for the matcher. Better some false positives but avoid false negatives.
I was thinking too smart. I thought name variants are taken from the BAI custer.

[kaplun]

Yeah, name_variants are just automatically derived from full_name so can be ignored. The author_comparator will comparate only real information.

[michamos]

do you really want Massimo and Maria to match? those are pretty different names.

[ksachs]

If possible I want
a match for 'M.' - 'Massimo'
no match for 'Maria' - ' Massimo'
But if this is not possible I rather accept a match for 'Maria' - ' Massimo' instead of missing 'M.' - 'Massimo'.
Maybe add an example like ['Ellis, J.', 'Ellis, John'] to the code.

[chris-asl]

So, by adding the examples by @ksachs, we're getting these results:

[match]: "Cox, Brian", "Cox, Brian"
[match]: "O Brien, Dara", "O Briain, Dara"
[NO match]: "O Brien, Dara", "Christos Aslanoglou"
[match]: "John, Ellis", "John ellis"
[match]: "John, Ellis", "John, Richard Ellis"
[match]: "John, Ellis", "John, R. Ellis"
[match]: "John, Ellis", "Ellis, R. John"
[match]: "John, Ellis", "Ellis, R. J"
[match]: "John, Ellis", "j, r. ellis"
[NO match]: "John, Ellis", "j, r, ellis"
[match]: "John, Ellis", "j r ellis"
[match]: "John, Ellis", "j richard ellis"
[match]: "J Ellis", "john r e"
[match]: "Ellis J", "john r e"
[match]: "Ellis, J.", "Ellis, John"
[match]: "John R. Ellis", "j richard ellis"
[match]: "John R. Ellis", "j r ellis"
[match]: "John R. Ellis", "john rich. e"
[match]: "Παπαδόπουλος, Γ", "Παπαδόπουλος Γιώργος"
[match]: "Παπαδόπουλος, Γ", "Παπαδόπουλος Μ Γιώργος"
[match]: "Παπαδόπουλος, Γ", "Παπαδόπουλος Μιχάλη Γ"
[NO match]: "Jhon Brian", "John Brian"
[NO match]: "John Brian", "John C"
[match]: "Gaspari", "Gaspari, Maria"
[match]: "Gaspari M", "Gaspari, Maria"
[match]: "Gaspari Maria", "Gaspari, Maria"
[match]: "M Gaspari", "Gaspari, Maria"
[match]: "Maria Gaspari", "Gaspari, Maria"
[match]: "Sunje, Dallmeier-Tiessen", "Dallmeier-Tiessen, Sunje"
[match]: "Suenje, Dallmeier-Tiessen", "Dallmeier-Tiessen, Sunje"
[match]: "Sünje, Dallmeier-Tiessen", "Dallmeier-Tiessen, Sunje"
[match]: "Sunje, Tiessen", "Dallmeier-Tiessen, Sunje"
[match]: "Sünje, Tiessen", "Dallmeier-Tiessen, Sunje"
[match]: "Suenje, Tiessen", "Dallmeier-Tiessen, Sunje"
[match]: "Sunje, Dallmeier", "Dallmeier-Tiessen, Sunje"
[match]: "Sünje, Dallmeier", "Dallmeier-Tiessen, Sunje"
[match]: "Suenje, Dallmeier", "Dallmeier-Tiessen, Sunje"

[michamos]

@chris-asl she meant Massimo and Maria as first names, so Gaspari, Massimo, Gaspari, Maria and Gaspari, M..

[kaplun]

That's quite strong matching. What about:
"Sunje, Dallmeier-Tiessen", "Sünje, Dallmeier-Tiessen", "Suenje, Dallmeier-Tiessen", "Sunje, Tiessen", "Sünje, Tiessen", "Suenje, Tiessen", "Sunje, Dallmeier", "Sünje, Dallmeier", "Suenje, Dallmeier",

[ksachs]

Is this the same AuthorComparator I'm using in `invenio-matcher-benchmark'?
I get no match for "Gaspari, M", "Gaspari, Massimo".

[match]: "O Brien, Dara, "O Briain, Dara"
How do you know this is a spelling variant and not a different name?

[chris-asl]

@kaplun I've updated my comment with your suggestions.

@ksachs: As discussed with @ammirate, regarding the O Brien, Dara vs O Briain, Dara, this was merely a test to see how the matcher behaves. With the current configuration of the matcher, along with my change here, it seems that the matcher "thinks" that's a match.