TY - JOUR
T1 - Natural language techniques supporting decision modelers
AU - Arco, Leticia
AU - Nápoles, Gonzalo
AU - Vanhoenshoven, Frank
AU - Lara, Ana Laura
AU - Casas, Gladys
AU - Vanhoof, Koen
N1 - Funding Information:
Aiming at creating the collection, we examine many documents where decisions were presented; for instance, documents from the Atlanta police department where working procedures contain business rules. policy documents from insurance companies rules to grant research funding and football rules. Moreover, we included sentences from codes of conduct of diverse companies, such as: AIRBUS, Apple, Citi, Beiersdorf, Walt Disney, IBM, Lidl, PMI, SANDVIK, PEPSICO and Google. These documents are publicly available, which facilitates the reproducibility of results.
Publisher Copyright:
© 2020, The Author(s), under exclusive licence to Springer Science+Business Media LLC, part of Springer Nature.
PY - 2021/1
Y1 - 2021/1
N2 - Decision Model and Notation (DMN) has become a relevant topic for organizations since it allows users to control their processes and organizational decisions. The increasing use of DMN decision tables to capture critical business knowledge raises the need for supporting analysis tasks such as the extraction of inputs, outputs and their relations from natural language descriptions. In this paper, we create a stepping stone towards implementing a Natural Language Processing framework to model decisions based on the DMN standard. Our proposal contributes to the generation of decision rules and tables from a single sentence analysis. This framework comprises three phases: (1) discourse and semantic analysis, (2) syntactic analysis and (3) decision table construction. To the best of our knowledge, this is the first attempt devoted to automatically discovering decision rules according to the DMN terminology from natural language descriptions. Aiming at assessing the quality of the resultant decision tables, we have conducted a survey involving 16 DMN experts. The results have shown that our framework is able to generate semantically correct tables. It is convenient to mention that our proposal does not aim to replace analysts but support them in creating better models with less effort.
AB - Decision Model and Notation (DMN) has become a relevant topic for organizations since it allows users to control their processes and organizational decisions. The increasing use of DMN decision tables to capture critical business knowledge raises the need for supporting analysis tasks such as the extraction of inputs, outputs and their relations from natural language descriptions. In this paper, we create a stepping stone towards implementing a Natural Language Processing framework to model decisions based on the DMN standard. Our proposal contributes to the generation of decision rules and tables from a single sentence analysis. This framework comprises three phases: (1) discourse and semantic analysis, (2) syntactic analysis and (3) decision table construction. To the best of our knowledge, this is the first attempt devoted to automatically discovering decision rules according to the DMN terminology from natural language descriptions. Aiming at assessing the quality of the resultant decision tables, we have conducted a survey involving 16 DMN experts. The results have shown that our framework is able to generate semantically correct tables. It is convenient to mention that our proposal does not aim to replace analysts but support them in creating better models with less effort.
KW - Decision Modeling and Notation
KW - Decision rules
KW - Decision tables
KW - Natural Language Processing
U2 - 10.1007/s10618-020-00718-4
DO - 10.1007/s10618-020-00718-4
M3 - Article
VL - 35
SP - 290
EP - 320
JO - Data Mining and Knowledge Discovery
JF - Data Mining and Knowledge Discovery
SN - 1384-5810
IS - 1
ER -