A Supervised Approach to Predict the Hierarchical Structure of Conversation Threads for Comments

Table 3

Example of calculation of interrogative metrics for the example shown in Figure 18.

Evaluation metric	Calculation
Accuracyedge	Replyrelations={1→𝑅,2→𝑅,7→𝑅,9→𝑅,10→𝑅,11→𝑅,3→1,4→3,5→2,6→5,8→5}; Detectedrelations=={1→𝑅,2→𝑅,9→𝑅,10→𝑅,11→𝑅,3→1,4→1,5→2,6→5,7→2,8→7}; 811≈0.73
Accpath	|(1→𝑅,2→𝑅,9→𝑅,10→𝑅,11→𝑅,3→1→𝑅,5→2→𝑅,6→5→2→𝑅)|=8Numberofcomments≈110.73
𝑃𝑅-path	(1→𝑅,2→𝑅,9→𝑅,10→𝑅,11→𝑅,3→1→𝑅,5→2→𝑅,6→5→2→𝑅)+(4→1→𝑅,7→2→𝑅,8→7→2→𝑅)Numberofcomments; =(1+1+1+1+1+1+1+1)+(1+1/2+2/3)=1110.1711≈0.92
𝑅𝑅-path	(1→𝑅,2→𝑅,9→𝑅,10→𝑅,11→𝑅,3→1→𝑅,5→2→𝑅,6→5→2→𝑅)+(4→3→1→𝑅,7→𝑅,8→5→2→𝑅)Numberofcomments; =(1+1+1+1+1+1+1+1)+(2/3+1+2/3)=1110.3411≈0.94
𝐹1-score𝑅-path	2∗0.92∗0.940.92+0.94≈0.93