Parsing title and last name, e.g. "Mr XXX" should be last name, not first name

[danhartropp]

Love this library - very useful.

However, I've noticed that parsing names in the format Prefix Lastname (e.g. Mr Magoo) parse with a blank name.last and the last name in the first name position (e.g. name.first == Magoo, name.last == ''). I think this should be the other way round!

I may have time to fix this in your code later, but for now i'm using the following kludge to make it work in my code...

if name.title != '':
if name.last == '':
name.last = name.first
name.first = ''

[derek73]

I agree with you, it probably should be the other way around. That sounds like a reasonable improvement. 👍

[danhartropp]

I've ben thinking about this and it has occurred to me that the problem may apply to all prefixes. For instance, while a name in the format Mr Magoo should probably parse to prefix = Mr and lastname = Magoo, another name in the format Sir Gerald should parse to prefix = Sir and firstname - Gerald.

Are you planning an update release any time soon ... I won't have time to look at this for a couple of weeks, but I may be able to after that.

[derek73]

There is no rule-based approach to know that "Gerald" is a first name in the case of "Sir Gerald" (if indeed it is). We must choose either last or first name. If we just feed it one name, we have to assume it's a first name. If we feed it two names and one is a title, it seems the other name will more often be a last name than a first name.

Just for my own curiosity, I wonder what kind of data you're using the library for?

[derek73]

I added my code to treat a title and name as a last name instead of a first name. Let me know if you have a chance to try it out or run into any problems.

[danhartropp]

Hi Derek,

I'm using it as part of a scraper, extracting questions and answers from the UK parliament website (so 'sir' comes up quite a lot, along with the occasional 'Viscount').

I think the rule with 'sir' is rigid. You would say sir Gerald, or sir Gerald Kaufman, but never sir Kaufman.

I haven't looked into your code, but is it possible to add an exception to the normal last name assumption when the title is 'sir' ?

Dan

Sent from my HTC

----- Reply message -----
From: "derek73" notifications@github.com
To: "derek73/python-nameparser" python-nameparser@noreply.github.com
Cc: "danhartropp" danhartropp@yahoo.co.uk
Subject: [python-nameparser] Parsing "Mr XXX" (#7)
Date: Wed, May 7, 2014 05:51
There is no rule-based approach to know that "Gerald" is a first name in the case of "Sir Gerald" (if indeed it is). We must choose either last or first name. If we just feed it one name, we have to assume it's a first name. If we feed it two names and one is a title, it seems the other name will more often be a last name than a first name.

Just for my own curiosity, I wonder what kind of data you're using the library for?

—
Reply to this email directly or view it on GitHub.

[derek73]

It would certainly be possible and preferable to create a special case for "sir" if the assumption that it is most-typically followed by a first name is true. "Sir" is not a word comes up often here in the US so I don't have a very informed understanding of how it's used.

Looks like Wikipedia agrees with you.

Sir is an honorific address used as a courtesy title to address a man without using his given name or family name in many English speaking cultures

In formal protocol Sir is the correct styling for a knight or a baronet [...] used with (one of) the knight's given name(s) or full name, but not with the surname alone ("Sir James Paul McCartney", "Sir Paul McCartney", or "Sir Paul", but never "Sir McCartney").

Seems we should also include "dame" in the same exception? Does that sound right to you?

It also seems maybe "king" and "queen" would have the same exception? Kind of an edge case though. Any others you can think of?

[derek73]

What about 'Maid Marion'? Is that another one?

Master?

Also seems like "Aunt", "Auntie" and "Uncle" also fall in this category.

[derek73]

Looking through all of the titles this is the list I came up with. Let me know what you think.

# When these titles are used with a single name, that name is a
# first name rather than a last name.
FIRST_NAME_TITLES = set([
    'sir',
    'dame',
    'king',
    'queen',
    'master',
    'maid',
    'uncle',
    'auntie',
    'aunt',
    'brother',
    'sister',
    'mother',
    'father',
    'pope',
])

[danhartropp]

I've had a look at the TITLES set ... the set of FIRST_NAME_TITLES looks
right to me.

A couple of other things you might be interested in ...

'The Right Honourable' is a title that appears a lot in the UK
parliament. It is also possible for someone to be 'Honourable' without
being 'Right Honourable'. The short form is 'Rt Hon' so I therefore add
'hon' and 'honourable' to titles_c in my code, which takes care of it.
According to wikipedia, 'honorable' is an alternative spelling sometimes
used for US judges.

I've also come across an issue with suffixes ... members of the UK
parliament have the suffix 'MP', so I add that to suffixes_c. The issue
is that some MPs are also QCs, so they have chained suffixes like 'QC
MP'. As far as I can see, the parser doesn't allow chained suffixes, so
just adding 'qc' to suffixes_c doesn't solve the problem. In the UK
suffixes like 'III' are rare, so my workaround just looks for cases
where the lastname has been parsed as 'QC' and adjusts accordingly.

I'm using a workaround, but you might like to consider adding chained
suffixes to the code in the same way that titles can already be chained.

Thanks again for a great tool - it has saved me hours of time!

Dan

On 08/05/2014 22:08, derek73 wrote:

Looking through all of the titles this is the list I came up with. Let
me know what you think.

|# When these titles are used with a single name, that name is a

first name rather than a last name.

FIRST_NAME_TITLES = set([
'sir',
'dame',
'king',
'queen',
'master',
'maid',
'uncle',
'auntie',
'aunt',
'brother',
'sister',
'mother',
'father',
'pope',
])
|

—
Reply to this email directly or view it on GitHub
#7 (comment).

[danhartropp]

I've just realised I should have ben clearer - I understand that suffixes are chained when comma separated - is there a way to chain them without the commas, in the way that titles are?

[derek73]

Thanks for the suggestions Dan. That sounds like a good idea.

It might be possible to recognize chained suffixes since we're not relying on their position in the string to recognize them. The concept sort of touches on #2 (MP is actually a title even though it comes at the end of the name), which I was planning to try to address in a v0.3. I think I'll open a new issue for that because it would be nice and seems possible.

"Hon" has come up before. Unfortunately "Hon" is also a first name. Currently the parser will recognize "Hon." as a title if it ends in a period but assume its a first name if it has no period. I don't know if that's the right solution. Seems it's more common as a title. I don't know how many Hon's we'd be upsetting though by always counting it as title. Hon Solo is probably not going to care.

Another thing I want to do in the next big version bump is to make it easier to adjust the constants by adding some add() and delete() methods. That might be a good time to flip flop on "Hon" and count it as a title since you could do easily something like name HumanName.constants.titles.delete("hon") if you had a data set with more "Hon"s than judges.

Seems we should definitely add "MP" and "QC" to the ... I guess SUFFIXES_C, for now. And "right", "honorable" and "honourable" to TITLES_C.

Technically, if you wanted to hack yours to deal with "QC MP" you could override parse_nicknames() on a subclass or something. It's a method that happens before any other name processing, so before it's split on spaces. You could search for "QC MP", if found remove it from the string and do self.title_list.append('QC MP") or self.title_list.append(['QC','MP']). Another thing I want to do with next version bump is to change that method name to a pre_process() or something.

Thinking about it, that actually might be a better approach than chaining these titles/suffixes. "Rt Hon" is always that, never "Hon Rt" nor "Rt MD". Though, I guess if we came across "Rt MD" might still be better to count both as titles/suffixes.