Self-driving laboratories (SDLs) accelerate discovery and improve reproducibility, but the lack of a unified language for robot-executable synthesis operations limits their scalability. This study addresses this gap by identifying and standardizing operation verbs, revealing that differences between organic and inorganic synthesis lie primarily in verb usage frequency. A total of 166 distinct operation verbs were identified, encompassing most synthesis operations. NL2SOU model achieves 86.52% (verb-argument)/86.65% (argument-value) accuracy, improving to 91.49%/91.23% by removing ambiguous ‘yield’ errors. It further analyzed verb-argument associations using Normalized Pointwise Mutual Information (NPMI). Results revealed significant variability in verb-argument relationships, highlighting the need for precise argument specifications. To construct robot-executable synthesis units, this study proposes a strategy based on the semantic and procedural similarities of verbs. A total of 26 operational units were standardized from 47 verbs, offering a more structured representation than χDL. Future work will focus on expanding the coverage of operational units, refining the NL2SOU model, and improving its applicability to autonomous synthesis systems.