# Copyright 2019-2025 The ASReview Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from asreview.learner import ActiveLearningCycleData
AI_MODEL_CONFIGURATIONS = [
{
# ELAS u4 is the default model for this version ASReview LAB. The model
# parameters have been optimized on the SYNERGY dataset. After expert
# elicitation, the model and parameters have been chosen.
# Optimization setup: https://github.com/asreview/asreview-optuna/releases/tag/ASReview2_0b4-nb-tfidf-full-1
"name": "elas_u4",
"label": "ELAS u4",
"type": "ultra",
"value": ActiveLearningCycleData(
querier="max",
classifier="svm",
classifier_param={"loss": "squared_hinge", "C": 0.11},
balancer="balanced",
balancer_param={"ratio": 9.8},
feature_extractor="tfidf",
feature_extractor_param={
"ngram_range": (1, 2),
"sublinear_tf": True,
"min_df": 1,
"max_df": 0.95,
},
),
},
{
# ELAS u3 mimics the performance of ASReview LAB version 1. The model
# has an updated and optimized balancer, as "double" balance has been
# deprecated in this version of ASReview LAB.
"name": "elas_u3",
"label": "ELAS u3",
"type": "ultra",
"value": ActiveLearningCycleData(
querier="max",
classifier="nb",
classifier_param={"alpha": 3.822},
balancer="balanced",
balancer_param={"ratio": 1.2},
feature_extractor="tfidf",
feature_extractor_param={"stop_words": "english"},
),
},
# ELAS u2 refers to the model that was used in the ASReview LAB version 1. The
# model is not available in the current version of ASReview LAB.
# ELAS u1 refers to the model that was used in the ASReview LAB version 0. The
# model is not available in the current version of ASReview LAB.
{
# ELAS l2 is the multilingual model for this version ASReview LAB. The model
# parameters have been optimized on the SYNERGY dataset. After expert
# elicitation, the model and parameters have been chosen.
# Optimization setup: https://github.com/asreview/asreview-optuna/releases/tag/ASReview2_0b4-svm-e5-full-1
"name": "elas_l2",
"label": "ELAS l2",
"type": "lang",
"extensions": ["asreview-dory"],
"value": ActiveLearningCycleData(
querier="max",
classifier="svm",
classifier_param={"loss": "squared_hinge", "C": 0.106, "max_iter": 5000},
balancer="balanced",
balancer_param={"ratio": 9.707},
feature_extractor="multilingual-e5-large",
feature_extractor_param={"normalize": True},
),
},
{
# ELAS h3 is the heavy model for this version ASReview LAB. The model
# parameters have been optimized on the SYNERGY dataset. After expert
# elicitation, the model and parameters have been chosen.
# Optimization setup: https://github.com/asreview/asreview-optuna/releases/tag/ASReview2_0b4-svm-mxbai-full-1
"name": "elas_h3",
"label": "ELAS h3",
"type": "heavy",
"extensions": ["asreview-dory"],
"value": ActiveLearningCycleData(
querier="max",
classifier="svm",
classifier_param={"loss": "squared_hinge", "C": 0.067, "max_iter": 5000},
balancer="balanced",
balancer_param={"ratio": 9.724},
feature_extractor="mxbai",
feature_extractor_param={"normalize": True},
),
},
]
[docs]
def get_ai_config(name=None):
"""Get the AI configuration.
Parameters
----------
name: str
The name of the AI configuration. If None, the default AI
configuration is returned.
Returns
-------
dict:
The default AI configuration.
"""
if name is None:
return filter(
lambda x: x["type"] == "ultra", AI_MODEL_CONFIGURATIONS
).__next__()
for model_config in AI_MODEL_CONFIGURATIONS:
if model_config["name"] == name:
return model_config
raise ValueError(f"Model configuration {name} not found.")
