Update README.md
Browse files
README.md
CHANGED
@@ -27,20 +27,59 @@ configs:
|
|
27 |
path: data/test-*
|
28 |
---
|
29 |
|
30 |
-
Assessing DIScourse COherence in Italian TEXts (DISCOTEX)
|
|
|
31 |
Original Paper: https://sites.google.com/view/discotex/
|
32 |
|
33 |
Task presented at EVALITA-2023
|
34 |
|
35 |
The original task is about modelling discourse coherence for Italian texts.
|
36 |
|
37 |
-
We focalized only on the first sub-task: Last
|
38 |
|
39 |
-
To assess the capability of a Language Model to solve such kind of task we reframed the task as
|
40 |
|
41 |
The question will ask to the model given a short paragraph which target sentence is the correct between a list of four, the answers will be the starting letters of the relative target, and a fifth option that indicate that no one target is the correct continuation.
|
42 |
|
|
|
|
|
43 |
For each sample, if the sample has 1 as label, we set the relative target as gold answer and three other random targets (from other samples) as distractors. On the other way around, if the sample has 0 as label, we set the relative target and other three random targets (from other samples) as distractors, as the gold answer will be chosen the sentence: "nessuna delle precedenti".
|
44 |
|
45 |
-
|
46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
path: data/test-*
|
28 |
---
|
29 |
|
30 |
+
# Assessing DIScourse COherence in Italian TEXts (DISCOTEX)
|
31 |
+
|
32 |
Original Paper: https://sites.google.com/view/discotex/
|
33 |
|
34 |
Task presented at EVALITA-2023
|
35 |
|
36 |
The original task is about modelling discourse coherence for Italian texts.
|
37 |
|
38 |
+
We focalized only on the first sub-task: **Last Sentence Classification**: given a short paragraph, and an individual sentence (target), the model will be asked to classify whether the target follows or not the paragraph.
|
39 |
|
40 |
+
To assess the capability of a Language Model to solve such kind of task we reframed the task as **Multi-Choice QA**.
|
41 |
|
42 |
The question will ask to the model given a short paragraph which target sentence is the correct between a list of four, the answers will be the starting letters of the relative target, and a fifth option that indicate that no one target is the correct continuation.
|
43 |
|
44 |
+
## Distractors Generation
|
45 |
+
|
46 |
For each sample, if the sample has 1 as label, we set the relative target as gold answer and three other random targets (from other samples) as distractors. On the other way around, if the sample has 0 as label, we set the relative target and other three random targets (from other samples) as distractors, as the gold answer will be chosen the sentence: "nessuna delle precedenti".
|
47 |
|
48 |
+
## Example
|
49 |
+
|
50 |
+
Here you can see the structure of the single sample in the present dataset.
|
51 |
+
|
52 |
+
|
53 |
+
```json
|
54 |
+
{
|
55 |
+
"text": string, # text of the short paragraph
|
56 |
+
"choices": list, # list of possible answers, with the correct one plus 4 distractors
|
57 |
+
"label": int, # index of the correct anser in the choices
|
58 |
+
}
|
59 |
+
```
|
60 |
+
|
61 |
+
|
62 |
+
## Statistics
|
63 |
+
|
64 |
+
Training: 16000
|
65 |
+
Test: 1600
|
66 |
+
|
67 |
+
## Proposed Prompts
|
68 |
+
|
69 |
+
Here we will describe the prompt given to the model over which we will compute the perplexity score, as model's answer we will chose the prompt with lower perplexity.
|
70 |
+
Moreover, for each subtask, we define a description that is prepended to the prompts, needed by the model to understand the task.
|
71 |
+
|
72 |
+
### Behaviour
|
73 |
+
|
74 |
+
Description of the task: "Ti verranno poste delle domande, nelle quali è presente un paragrafo, e come possibili risposte varie frasi che possono essere o meno il continuo.\nIndica la frase che rappresenta la continuazione del paragrafo oppure 'nessuna delle precedenti', se nessuna delle continuazioni è corretta."
|
75 |
+
|
76 |
+
Prompt: "Paragrafo: '{{text}}'\nDomanda: Quali delle seguenti frasi presenta una continuazione del precedente paragrafo?\nA. '{{choices[0]}}'\nB. '{{choices[1]}}'\nC. '{{choices[2]}}'\nD. '{{choices[3]}}'\nE. {{choices[4]}}\nRisposta:"
|
77 |
+
|
78 |
+
## Some Results
|
79 |
+
|
80 |
+
| DISCOTEX | ACCURACY |
|
81 |
+
| :--------: | :----: |
|
82 |
+
| Mistral-7B | 0 |
|
83 |
+
| ZEFIRO | 0 |
|
84 |
+
| Llama-3 | 0 |
|
85 |
+
| ANITA | 0 |
|