Skip to main content

Table 2 Results table

From: Is it possible to describe television series from online comments?

Series ID Concatenated \({\mathcal {C}^{d}}\) text TKW-AF TKW-MSC TextRank
  Text length ROUGE-1 Text length ROUGE-1 Text length ROUGE-1 Text length ROUGE-1
1 48800 0.6642 15444 (31.65%) 0.5087 (76.59%) 4202 (8.61%) 0.3750 (56.46%) 9545 (19.56%) 0.4201 (63.24%)
19 98412 0.7800 52210 (53.05%) 0.7152 (91.70%) 13186 (13.40%) 0.5657 (72.53%) 19912 (20.23%) 0.5705 (73.14%)
30 99388 0.7821 37417 (37.65%) 0.6893 (88.13%) 6877 (6.92%) 0.4578 (58.53%) 19580 (19.70%) 0.5683 (72.67%)
205 32171 0.6834 11446 (35.58%) 0.5620 (82.24%) 3270 (10.16%) 0.4011 (58.69%) 6297 (19.57%) 0.4433 (64.86%)
226 36146 0.7500 14074 (38.94%) 0.6496 (86.61%) 2470 (6.83%) 0.3675 (49.00%) 7692 (21.28%) 0.5406 (72.08%)
356 59593 0.7852 27185 (45.62%) 0.6984 (88.95%) 6194 (10.39%) 0.4552 (57.97%) 13529 (22.70%) 0.5661 (72.10%)
457 56062 0.8452 15958 (28.46%) 0.7619 (90.14%) 4809 (8.58%) 0.6111 (72.30%) 11044 (19.70%) 0.6786 (80.28%)
777 43624 0.6724 15165 (34.76%) 0.5666 (84.26%) 3547 (8.13%) 0.4061 (60.41%) 9234 (21.17%) 0.4642 (69.04%)
790 50300 0.7566 21334 (42.41%) 0.6675 (88.22%) 6855 (13.63%) 0.5084 (67.20%) 9908 (19.70%) 0.5229 (69.11%)
820 114110 0.8073 50503 (44.26%) 0.7205 (89.25%) 11416 (10.00%) 0.5088 (63.02%) 22596 (19.80%) 0.5645 (69.92%)
877 67725 0.7962 34445 (50.86%) 0.7208 (90.52%) 7960 (11.75%) 0.5491 (68.96%) 13406 (19.79%) 0.5849 (73.46%)
934 82331 0.7458 28013 (34.02%) 0.6325 (84.81%) 6091 (7.40%) 0.4305 (57.73%) 17586 (21.36%) 0.5527 (74.11%)
13599 99245 0.7856 27932 (28.14%) 0.6689 (85.15%) 6200 (6.25%) 0.4106 (52.27%) 21323 (21.49%) 0.5674 (72.23%)
Average 68300.54 0.7580 27009.69 (39.55%) 0.6586 (86.89%) 6390.54 (9.36%) 0.4651 (61.36%) 13973.23 (20.46%) 0.5418 (71.48%)
  1. The table shows for each series: the length (number of words) of the concatenated texts of all comments \(c \in \mathcal {C}^{d}\), the ROUGE-1 score of the this concatenated text in respect to the human-written summary Rd, and the lengths and ROUGE-1 scores for the summaries generated with methods TKW-AF, TKW-MSC, and TextRank, with how much of the respective values from the full text of \(\mathcal {C}^{d}\) they represent indicated in parenthesis