Natural Language Processing for Virtual Reference Analysis

Objective – Chat transcript analysis can illuminate user needs by identifying common question topics, but traditional hand coding methods for topic are time-consuming and poorly suited to large datasets. The research team explored the viability of automatic natural language processing (NLP) strategies perform rapid on a dataset transcripts from consortial chat service. Methods developed toolchain data analysis, which incorporated targeted searching query terms using regular expressions Python spaCy library analysis. Processed was exported Tableau visualization. Results were compared hand-coded test accuracy conclusions. processed provided insights about volume chats originating each participating library, proportion answered operator groups percentage different staff types. also captured top referring URLs service, course codes file extensions mentioned, hits. Natural revealed that most topics related citation, subscription databases, finding full-text articles, aligns with types identified in transcripts. Conclusion Compared coding, NLP approaches have benefits be analyzed time frame required they come trade-off accuracy, such as false Therefore, computational should used supplement methods. As becomes more accurate, these may widen avenues insight into virtual reference patron needs.