Whois Dataset

dblp.zip

This database is used for tasks related to disambiguating author names. It contains 69,574,243 records and 10 columns and was obtained from the DBLP repository and has been preprocessed to extract all possible combinations of pairs of authors (2,665,634) unique authors) from 5,299,929 papers in the database. There are some are in the database where a single author is duplicated.

Attributes

Note that Target Author = Target Author's First Name + Target Author's Last Name + Suffix. The suffix is added to the target author's name to ensure that it refers to a specific, unique person in the real world.

Example: 

Given the following reference string: 

Boukhers, Zeyd, and Asundi, Nagaraj Bahubali. "Deep Author Name Disambiguation Using Bibliographic Data." International Conference on Theory and Practice of Digital Libraries. Springer, Cham, 2022.

The following records are extracted:

data.zip

It contains pickle files in the format <n>_<first_name_acronym> <full_last_name>.pickle, each of which refers to an atomic name (i.e. the acronym of the first name and the full last name), where <n> denotes the number of real-world authors sharing the atomic name. The pickle file contains the indices of these real-world authors in the following format:

[<atomic-name>, <author_index>, <author_index>, ....]

For example, 4_T Akutsu.pickle contains 

['T Akutsu', 2274176, 2276257, 2290454, 2347757]

indices.zip

It contains two dictionaries: 

• index2auth.pickle: The real world author (not the name) is retrieved given the index
• auth2index.pickle: The author index is retrieved given the real world author (not the name)

Utils.zip

It contains other necessary pickle files:

• author_list.pickle: it contains the list of all authors (2665634 authors)
• author_abbvs.pickle: it contains the list of atomic names of all authors (2665634 authors)
• author_names.pickle; it contains the list of full names of all authors (2665634 authors)
• unique.pickle: it contains the list of unique atomic names (1555517 atomic names)
• unique_full_names.pickle: it contains the list of unique full names (2629851 full names)
• indices.pickle: it contains the indices of authors who share the atomic name (1555517 atomic names)

Evaluation_Data.zip

It contains an example of training, validation and test data. 

• x_train.pickle: contains the indices of the training records 
• x_validate.pickle: contains the indices of the validation records 
• x_test.pickle: contains the indices of the testing records 
• y_train.pickle: contains the indices of the corresponding authors to the records in x_train.pickle
• y_validate.pickle: contains the indices of the corresponding authors to the records in x_validate.pickle
• y_test.pickle: contains the indices of the corresponding authors to the records in x_test.pickle

Miscs.zip

It contains data extracted from the database to be used for training and testing. The files contains data for one example atomic name. It contains the following pickle files:

• records_train.pickle: contains the training records in the following format: 

  • combination_id
    reference_id
    target_author
    author_fname
    author_lname
    coauthor_fname
    coauthor_lname
    title
    year
    abbr_journal
    journal

• ref_train.pickle: contains the indices of the training records. 
• old_authors_ids.pickle
• new_authors_ids.pickle