Idea Mining From Online Reviews Using Transformation-Based Natural Language Processing Tasks
- Resource Type
- Conference
- Authors
- Mahdi, Hussain Falih; Gupta, Lav Kumar; Choudhury, Tanupriya; Bansal, Nikunj
- Source
- 2022 International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT) Multidisciplinary Studies and Innovative Technologies (ISMSIT), 2022 International Symposium on. :894-899 Oct, 2022
- Subject
- Communication, Networking and Broadcast Technologies
Computing and Processing
Engineered Materials, Dielectrics and Plasmas
Engineering Profession
Photonics and Electrooptics
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Technological innovation
Machine learning algorithms
Bit error rate
Transformers
Natural language processing
Feedforward neural networks
Data mining
Text Mining
Idea Mining
Sentence Similarity
BERT
Machine Learning
Neural Networks
- Language
- ISSN
- 2770-7962
Product reviews from online communities are good sources to innovative ideas. Making use of this unstructured data from these reviews is a complicated process. The major task of identifying good ideas is ensuring the correct reviews are picked up for consideration. We propose a method of mining ideas by exploring online reviews data with a particular interest in discovering helpful suggestions in innovation. We use a classifier model to categorize whether the reviews from an online community provide a suggestion. The non-suggestive texts were discarded from the further analysis. The texts are then processed further by a sentence similarity comparison module that uses the Bidirectional Encoder Representations from Transformers (BERT) model. We use the sentence similarity techniques to filter common suggestions that are attractive to many reviewers. We describe how to use BERT based model for Natural Language Tasks. We conclude that mining reviews can help extract ideas from online communities. This is a generic implementation and can be adopted in any domain. We also suggest a human intervention to assess the context of the generated ideas. This method can help researchers and professionals discover ideas buried in unstructured online reviews.