<![CDATA[Functional requirement identification is a critical stage in information system development that determines system alignment with user needs. Interviewing is a dominant technique for eliciting requirements, but interview results in the form of long, unstructured text require significant analysis time and are prone to subjectivity if done manually. This research aims to apply the TextRank algorithm to automatically summarize interview results so that important information can be obtained more concisely. The research was conducted through text preprocessing stages, sentence similarity calculation (word overlap), graph construction, and sentence ranking using the PageRank algorithm. Important sentences were selected based on threshold variations (k = 3, 5, 7, 9). Three methods were compared: standard TextRank, TextRank with TF-IDF weighting, and TextRank with stemming preprocessing. Evaluation used precision, recall, and F1-score metrics against ground truth. Results show that standard TextRank with threshold 7 provided the most balanced performance (F1-score 71.05%), being more stable than TF-IDF based methods for interview data. This algorithm proved effective in improving the efficiency of the functional requirement identification process.]]>
Copyrights © 2026
