TeknoAssistant : a domain specific tech mining approach for technical problem-solving support

• Published in: Scientometrics 2022-09, Vol.127 (9), p.5459-5473
• Main Authors: Garechana, Gaizka, Río-Belver, Rosa, Zarrabeitia, Enara, Alvarez-Meaza, Izaskun
• Format: Article
• Language: English
• Subjects: Bayesian analysis; Clustering; Computer Science; Data mining; Indicators; Information Storage and Retrieval; Library Science; Natural language processing; Network analysis; Problem solving; Scientific papers; Social networks; Social organization; Technicians
• ISSN: 0138-9130; 1588-2861
• DOI: 10.1007/s11192-022-04280-2

This paper presents TeknoAssistant, a domain-specific tech mining method for building a problem–solution conceptual network aimed at helping technicians from a particular field to find alternative tools and pathways to implement when confronted with a problem. We evaluate our approach using Natural Language Processing field, and propose a 2-g text mining process adapted for analyzing scientific publications. We rely on a combination of custom indicators with Stanford OpenIE SAO extractor to build a Bernoulli Naïve Bayes classifier which is trained by using domain-specific vocabulary provided by the TeknoAssistant user. The 2-g contained in the abstracts of a scientific publication dataset are classified in either “problem”, “solution” or “none” categories, and a problem–solution network is built, based on the co-occurrence of problems and solutions in the abstracts. We propose a combination of clustering technique, visualization and Social Network Analysis indicators for guiding a hypothetical user in a domain-specific problem solving process.