Assisted declarative process creation from natural language descriptions

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

In this paper, we report recent advances on user support for declarative process generation from natural language descriptions. The Process Highlighter is a hybrid-modelling tool that facilitates the (manual) creation of Dynamic Response Condition (DCR) graphs directly from text documents, supporting non-technical users in the adoption of declarative process models. While some process descriptions are a few paragraphs long, others, such as the ones coming from municipal governments and legal bodies might contain several pages. Some aspects that undermine the adoption of hybrid modelling techniques and their promised one-to-one correspondence between texts and process models are the length of the texts, the inconsistent use of terms, and the difficulty in identifying textual elements that correspond to elements in a declarative process model. To mitigate these risks, we have implemented major additions in the Process Highlighter for industrial usage. The principal change is the inclusion of Natural Language Processing (NLP) techniques to support users in the identification of roles, activities and constraints. This, combined with the modelling, simulation and verification tools already existing in the framework, support the users in providing process models that are better aligned with their specifications, in a shorter time. These features are motivated from empirical observations of the use of the Process Highlighter in groups of caseworkers and students of process engineering in Danish universities.

TitelProceedings - 2019 IEEE 23rd International Enterprise Distributed Object Computing Workshop, EDOCW 2019
Publikationsdatookt. 2019
ISBN (Elektronisk)9781728145983
StatusUdgivet - okt. 2019
Begivenhed23rd IEEE International Enterprise Distributed Object Computing Workshop, EDOCW 2019 - Paris, Frankrig
Varighed: 28 okt. 201931 okt. 2019


Konference23rd IEEE International Enterprise Distributed Object Computing Workshop, EDOCW 2019

ID: 235145080