asreview.Project

class asreview.Project(project_path, project_id=None)

Bases: object

Project class for ASReview project files.

This class represents the complete data file for a review project. This data is contained in a single directory with the following files and subdirectories: - project.json: A JSON file containing the configuration and metadata of the project. It’s structure is described in schema.py - data/: A directory containing the input data file exactly as provided by the user. When exporting, this input data is merged with the results of the review to get the export file. - feature_matrices/: A directory containing all the feature matrices that are generated during the review. - results.db: An SQLite database containing all data generated by ASReview: the data parsed from the input file, the labeled records, the last model ranking etc. See asreview/data for information on the parsing of the input. See asreview/state for information on the model and the labeling decisions.

Methods

__init__(project_path[, project_id])
add_dataset(fp[, dataset_id, file_writer])	Add a dataset to the project file.
add_feature_matrix(feature_matrix, name)	Add feature matrix to project file.
add_review([cycle, reviewer, status])	Add new review metadata.
close()	Close the project and release all resources.
create(project_path[, project_id, ...])	Initialize the necessary files specific to the web app.
export(export_fp)
get_feature_matrix(name)	Get the feature matrix from the project file.
get_input_data_reader()
get_model_config()	Get the current model configuration of the review.
get_review_error()
label_priors()	Label prior knowledge from a partially labeled dataset.
load(asreview_file, project_path[, ...])
read_input_data(*args, **kwargs)
remove_dataset()	Remove dataset from project.
remove_review_error()
set_review_error(err)
update_config(**kwargs)	Update project info
update_review([status, model_name, model])	Update review metadata.

Attributes

MODE_SIMULATE
PATH_CONFIG
PATH_CONFIG_LOCK
PATH_DATA_DIR
PATH_DB
PATH_ERROR
PATH_FEATURE_MATRICES
VERSION
config
db
feature_matrices
input_data_fp
review

MODE_SIMULATE = 'simulate'

PATH_CONFIG = 'project.json'

PATH_CONFIG_LOCK = 'project.json.lock'

PATH_DATA_DIR = 'data'

PATH_DB = 'results.db'

PATH_ERROR = 'error.json'

PATH_FEATURE_MATRICES = 'feature_matrices'

VERSION = 3

add_dataset(fp, dataset_id=None, file_writer=None)

Add a dataset to the project file.

Parameters:

fp (str, Path) – Filepath to the dataset. It will be copied to the correct location in the project file.

add_feature_matrix(feature_matrix, name)

Add feature matrix to project file.

Parameters:

• feature_matrix (numpy.ndarray, scipy.sparse.csr.csr_matrix) – The feature matrix to add to the project file.

• name (str) – Name of the feature extractor.

add_review(cycle=None, reviewer=None, status='setup')

Add new review metadata.

Parameters:

• cycle – An active learning cycle object to add to the review. This object is used to store the configuration of the active learning cycle to file.

• reviewer (object) – A reviewer object with to_sql() method.

• status (str) – The status of the review. One of ‘setup’, ‘running’, ‘finished’.

close()

Close the project and release all resources.

Closes the database connection if it was opened. Safe to call multiple times.

property config

classmethod create(project_path, project_id=None, project_mode='oracle', project_name=None, project_tags=None)

Initialize the necessary files specific to the web app.

property db

export(export_fp)

property feature_matrices

get_feature_matrix(name)

Get the feature matrix from the project file.

Parameters:

name (str) – Name of the feature extractor for which to get the cached matrix.

Returns:

numpy.ndarray, scipy.sparse – (Sparse) feature matrix.

get_input_data_reader()

get_model_config()

Get the current model configuration of the review.

Returns:

dict | None – Dictionary containing the model configuration. Returns None if there is no review yet in the project.

get_review_error()

property input_data_fp

label_priors()

Label prior knowledge from a partially labeled dataset.

If the input dataset is partially labeled (some records have an included value of 0 or 1 while others are unlabeled), the labeled records are stored as prior knowledge in the results table.

Fully labeled or fully unlabeled datasets are skipped.

classmethod load(asreview_file, project_path, safe_import=False, reset_model_if_not_found=False)

read_input_data(*args, **kwargs)

remove_dataset()

Remove dataset from project.

remove_review_error()

property review

set_review_error(err)

update_config(**kwargs)

Update project info

update_review(status=None, model_name=None, model=None)

Update review metadata.