people person id_criteria - freebase-schema/freebase GitHub Wiki

• name_match: There is a name_match if the name of a person or their aliases on both the sources match. The match may also be approximate due to the topics frequently using person’s maiden names, nick names, missing/extra middle names, prefix/suffix (such as Jr., Sr.), initials, or names in different languages.

• image_match: There is an image_match if the images on both sources look like the same person.

• place_of_birth_match: There is a place_of_birth_match if the place of birth of the person (down to the city level) listed in Freebase property /people/person/place_of_birth matches with the place on the other source.

• date_of_birth_match: There is a date_of_birth_match if the date of birth of the person (to the day level) listed in Freebase property /people/person/date_of_birth matches with the date (to the day level) on the other source.

• place_of_death_match: There is a place_of_death_match if the place of death (down to the city level) of the person listed in Freebase topic matches with the place on the other source.

• date_of_death_match: There is a date_of_death_match if the date of death of the person (to the day level) from Freebase topic matches with the date (to the day level) on the other source.

• places_lived_match: There is a places_lived_match if one or more places that the person lived in, listed under the Freebase property /people/person/place_lived are the same places on the other source.

• education_match: There is an education_match if one or more of the schools (with years) attended by the person listed in the Freebase property /people/person/education matches with the schools and attending years on the other source. Sometimes if the years are approximate or missing, but the schools match completely, it is a match.

• employment_match: There is an employment_match if one or more records from employment history (employers, job titles such as Visual Graphic Artist and respective years) listed in Freebase property /people/person/employment_history match with the employer, job titles and years on the other source. Sometimes the years or titles may match approximately or will be missing.

• works_match: There is a works_match if any creative work such as art, music, software, movie done by the person match with the creative works on the other source.

• profession_match: There is a profession_match if the profession of the person listed in Freebase property /people/person/profession (such as Artist, Singer, Actor, Politician, Composer, Entrepreneur) matches with the profession on the other source.

• achievements_match: There is an achievements_match if one or more of the achievements or awards (such as major awards, noble prize, honors, company founder, discoveries, patents) achieved by the person listed in Freebase topic match with the achievements listed on the other source.

Real persons should not be matched to fictional characters (Example 11).

Here are some typical patterns for determining identity of two persons:

1. name_match and image_match (Example 1)
2. name_match and date_of_birth_match (Example 2, Example 3). (Example 4 for counterexample)
3. name_match and date_of_death_match (Example 5)
4. name_match and works_match (Example 6)
5. name_match and profession_match and *education_match_ (Example 7)
6. name_match and profession_match and *employment_match_ (Example 8)
7. name_match and achievements_match (Example 9, Example 10)

Pattern 1 Example 1. Here we have a matching image on both the sources. So it is the same person.

Pattern 2 Example 2 Permalink: Here the name, and full date of birth (to the day level) match for the person on both the sources. Hence the persons are identical.

Pattern 2 Example 3 Permalink: Here the names on Left Hand Side (LHS) and Right Hand Side (RHS) topics match partially, since LHS uses her married name while RHS has the maiden name. But since there is a full date of birth match, the persons are identical.

Pattern 2 Example 4. Here both the sources have a name match, but the Google+ person’s date of birth is March 15, 1960 which contradicts the date of birth of the Freebase person (July 26, 1950). Hence these two persons are not identical.

Pattern 3 Example 5 Permalink: Here the name, and complete date of death (to the day level) match for the person on both the sources. Hence the persons are identical.

Pattern 4 Example 6. In this example, Linkedin from LHS and IMDB from RHS both have matching name and cite the person's works on the “Ed, Edd ‘n’ Eddie” TV cartoon series. Hence the persons are identical.

Pattern 5 Example 7 Permalink Here the name, profession, and school attended with corresponding years match for the person between Linkedin from LHS and Freebase from RHS. Hence the persons are identical.

Pattern 6 Example 8 Permalink Here the name, profession and the employment history of the person match on LHS and RHS. Hence the persons are identical.

Pattern 7 Example 9 Permalink Here the name, profession and the achievements (noble prize and year) of the person match on both the sources. Hence the persons are identical.

Pattern 7 Example 10 Permalink Here the name, profession and the achievement (Twitter Founder) match for the person on both the sources. Hence the persons are identical.

Example 11. Here the topic on the LHS is about the real person, while the RHS topic is about a fictional character (a fictional depiction of the person on the LHS). A real person should be matched to the real person only, not to fictional character. Hence these topics are not identical.