Learning Complementary Action Modeling from Automotive Maintenance Instructions

The article introduces Complementary Action Modeling (CAM), a task focused on identifying or generating procedural counterparts in automotive maintenance instructions by modifying only the action phrase while preserving the rest of the sentence context.

• The study defines CAM as distinguishing complementarity from surface similarity and controlling generation at the action-phrase level.
• Evaluation methods include retrieval, overlap-based metrics, and human evaluation to assess relational correctness.
• Experiments utilize a German automotive maintenance dataset with candidate matching and controlled Seq2Seq generation.

The authors conclude that complementary maintenance instructions are best modeled as procedural associations grounded in subtle lexical cues rather than as ordinary sentence similarity or synonym-based paraphrasing.