File size: 3,676 Bytes
0ac2b72
 
e68b075
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0ac2b72
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e68b075
 
 
 
 
 
 
 
0ac2b72
 
 
 
 
 
 
 
677eb18
 
 
 
 
 
 
d8d0e35
 
 
b8d1272
 
 
bb30a6a
 
d8d0e35
bb30a6a
d8d0e35
 
 
 
 
 
 
 
 
bb30a6a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d8d0e35
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
---
dataset_info:
- config_name: cleaned
  features:
  - name: input
    dtype: string
  - name: output
    list: string
  - name: metadata
    dtype: string
  splits:
  - name: train
    num_bytes: 175281573
    num_examples: 1256
  - name: validation
    num_bytes: 23257908
    num_examples: 166
  - name: test
    num_bytes: 12708144
    num_examples: 86
  download_size: 128398306
  dataset_size: 211247625
- config_name: default
  features:
  - name: input
    dtype: string
  - name: output
    list: string
  - name: metadata
    dtype: string
  splits:
  - name: train
    num_bytes: 266540793
    num_examples: 1256
  - name: validation
    num_bytes: 35881749
    num_examples: 166
  - name: test
    num_bytes: 19669178
    num_examples: 86
  download_size: 142100992
  dataset_size: 322091720
configs:
- config_name: cleaned
  data_files:
  - split: train
    path: cleaned/train-*
  - split: validation
    path: cleaned/validation-*
  - split: test
    path: cleaned/test-*
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
license: odc-by
task_categories:
- text-classification
language:
- en
size_categories:
- 1K<n<10K
tags:
- long-context
- movie
- film
- screenplay
- narrative
---

# MulD: Movie Character Type Classification

this is the Movie Character Types task from [MuLD](https://arxiv.org/abs/2202.07362):

- Task: Classify characters as Hero/Protagonist or Villain/Antagonist
- Data: Movie scripts matched with Wikipedia plot summaries
- Method: Amazon Turk annotation based on plot summaries
- Average length: ~45,000 tokens
- Challenge: Character role understanding from full narrative context

```
@inproceedings{hudson-al-moubayed-2022-muld,
    title = "{M}u{LD}: The Multitask Long Document Benchmark",
    author = "Hudson, George  and
      Al Moubayed, Noura",
    editor = "Calzolari, Nicoletta  and
      B{\'e}chet, Fr{\'e}d{\'e}ric  and
      Blache, Philippe  and
      Choukri, Khalid  and
      Cieri, Christopher  and
      Declerck, Thierry  and
      Goggi, Sara  and
      Isahara, Hitoshi  and
      Maegaard, Bente  and
      Mariani, Joseph  and
      Mazo, H{\'e}l{\`e}ne  and
      Odijk, Jan  and
      Piperidis, Stelios",
    booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
    month = jun,
    year = "2022",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://aclanthology.org/2022.lrec-1.392/",
    pages = "3675--3685",
    abstract = "The impressive progress in NLP techniques has been driven by the development of multi-task benchmarks such as GLUE and SuperGLUE. While these benchmarks focus on tasks for one or two input sentences, there has been exciting work in designing efficient techniques for processing much longer inputs. In this paper, we present MuLD: a new long document benchmark consisting of only documents over 10,000 tokens. By modifying existing NLP tasks, we create a diverse benchmark which requires models to successfully model long-term dependencies in the text. We evaluate how existing models perform, and find that our benchmark is much more challenging than their `short document' equivalents. Furthermore, by evaluating both regular and efficient transformers, we show that models with increased context length are better able to solve the tasks presented, suggesting that future improvements in these models are vital for solving similar long document problems. We release the data and code for baselines to encourage further research on efficient NLP models."
}
```