[9p-W321-1~14] 23.1 Joint Session N "Informatics"

[9p-W321-4] Automatic Extraction of Solvent-names on Polymer's Solubility from Academic Articles

〇Hiroyuki Oka¹, Atsushi Yoshizawa¹, Hiroyuki Shindo^2,3, Yuji Matsumoto^2,3, Masashi Ishii¹ (1.NIMS, 2.NAIST, 3.RIKEN AIP)

Keywords:polymer, solvent-name, automatic extraction

To automatically extract solvent-names on polymer's solubility from academic articles, relation extraction between polymer- and solvent-names in text was examined by a rule-based method. In text, there are some words that directly mean solubility (for example, ”dissolve”, ”soluble”) and that express experimental methods that polymers have to be dissolved in solvents (for example, ”spin-coat”, ”viscosity”). The rule-based relation extraction utilizing these resulted in the extraction accuracy (F-score) of ca. 0.7. To evaluate the accuracy, the human-labor annotated data to polymer-names, solvent-names, and the relation between them in text was used.