ContributorImprovements

Currently the contributors table is a battleground where many different data sources fight it out to produce the best possible data representation of a human being who wrote a book. The situation is similar to what used to happen with the editions table. There's no indication as to who created the data, so if the data is wrong, it's not clear which bit of the system is at fault.

This proposal turns the contributors table into a place for every data source to record any information it has about the people who worked on the books it knows about.

It introduces the concept of 'contributor equivalencies' (similar to 'identifier equivalencies'), a place for data sources to record their belief that two distinct Contributors refer to the same individual.

It introduces the concept of a 'presentation contributor' (similar to 'presentation edition'), a place for the LS code itself to record its synthesis of all the available data into the best possible answer to the question: "who wrote this book and which other books did they write?"

Add data_source_id to Contributor

This stops data sources from stepping on each others' toes. If you have something to say about a person, create a Contributor object containing the information, and tag it with the id of the DataSource that provided the information.

Add is_presentation to Contributor

A presentation contributor represents the server's attempt to consolidate data from multiple sources, into an all-round view of an individual. This includes making the determination that, e.g. "Alice McPherson", "Alice Mc Pherson", "Alice Mcpherson", "Alice MacPherson", "Alice Mac Pherson", "Alice McPherson (1980-)", "Allie McPherson", and "Alice Maidenname" are all the same person.

A work's presentation edition will only reference presentation contributors. The WorkContributor join table will only reference presentation contributors.

Add ContributorEquivalency

A contributor equivalency represents an opinion that two Contributors represent the same human being (or corporate entity). It's nearly the same as Equivalency.

• input_id - Foreign key with contributor_id
• output_id - Foreign key with contributor_id
• data_source_id - Data source that has this opinion.
• strength - A number from -1 to 1 indicating the strength of the opinion.

Write the canonicalizer

Create an algorithm similar to the one we have for identifiers, which generates the transitive closure of the ContributorEquivalency graph (adding our own edges like "if two Contributors have the same VIAF number then they are the same person with probability x"), and consolidates a canonical Contributor from the cloud of information.

This will enforce rules like "only one canonical contributor with a given VIAF number".

Unanswered questions

• Should identifiers for people (including their names) be tracked separately from bibliographic information about them? This would allow a data source to say "I think Eric Blair and George Orwell are the same person" rather than "I think my "Eric Blair" record is the same person as my "George Orwell" record. It would be possible to have an opinion about equivalency without also having a record for either party.