AI in Business and Society
Loading...
Paper Number
2210
Paper Type
Completed
Description
Identifying innovative users and their ideas is crucial, for example, in crowdsourcing. But, analyzing large amounts of unstructured textual data from such online communities poses a challenge for organizations. Therefore, researchers started developing automated approaches to identify innovative users. Our study introduces an advanced machine-learning approach that minimizes manual work by combining transfer learning with a transformer-based design. We train the model on separate datasets, including an online maker community and various internet texts. The maker community posts represent need-solution pairs, which express needs and describe fitting prototypes. Then, we transfer the model and identify potential user innovations in a kitesurfing community. We validate the identified posts by manually checking a subsample and analyzing how words affect the model's classification decision. This study contributes to the growing portfolio of user innovation identification by combining state-of-the-art natural language processing and transfer learning to improve automated identification.
Recommended Citation
Resch, Christian; Streibel, Julian; Feiter, Tim; and Kock, Alexander, "Identifying User Innovations through AI in Online Communities– A Transfer Learning Approach" (2023). ICIS 2023 Proceedings. 17.
https://aisel.aisnet.org/icis2023/aiinbus/aiinbus/17
Identifying User Innovations through AI in Online Communities– A Transfer Learning Approach
Identifying innovative users and their ideas is crucial, for example, in crowdsourcing. But, analyzing large amounts of unstructured textual data from such online communities poses a challenge for organizations. Therefore, researchers started developing automated approaches to identify innovative users. Our study introduces an advanced machine-learning approach that minimizes manual work by combining transfer learning with a transformer-based design. We train the model on separate datasets, including an online maker community and various internet texts. The maker community posts represent need-solution pairs, which express needs and describe fitting prototypes. Then, we transfer the model and identify potential user innovations in a kitesurfing community. We validate the identified posts by manually checking a subsample and analyzing how words affect the model's classification decision. This study contributes to the growing portfolio of user innovation identification by combining state-of-the-art natural language processing and transfer learning to improve automated identification.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
10-AI