allenai/System3_DREAM_FLUTE_social_norm_FigLang2022

Model description

This is the T5-3B model for System 3 DREAM-FLUTE (social norm), as described in our paper Just-DREAM-about-it: Figurative Language Understanding with DREAM-FLUTE, FigLang workshop @ EMNLP 2022 (Arxiv link: https://arxiv.org/abs/2210.16407)

Systems 3: DREAM-FLUTE - Providing DREAM’s different dimensions as input context

We adapt DREAM’s scene elaborations (Gu et al., 2022) for the figurative language understanding NLI task by using the DREAM model to generate elaborations for the premise and hypothesis separately. This allows us to investigate if similarities or differences in the scene elaborations for the premise and hypothesis will provide useful signals for entailment/contradiction label prediction and improving explanation quality. The input-output format is:

Input <Premise> <Premise-elaboration-from-DREAM> <Hypothesis> <Hypothesis-elaboration-from-DREAM>
Output <Label> <Explanation>

where the scene elaboration dimensions from DREAM are: consequence, emotion, motivation, and social norm. We also consider a system incorporating all these dimensions as additional context.

In this model, DREAM-FLUTE (social norm), we use elaborations along the "social norm" dimension. For more details on DREAM, please refer to DREAM: Improving Situational QA by First Elaborating the Situation, NAACL 2022 (Arxiv link: https://arxiv.org/abs/2112.08656, ACL Anthology link: https://aclanthology.org/2022.naacl-main.82/).

How to use this model?

We provide a quick example of how you can try out DREAM-FLUTE (social norm) in our paper with just a few lines of code:

>>> from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
>>> model = AutoModelForSeq2SeqLM.from_pretrained("allenai/System3_DREAM_FLUTE_social_norm_FigLang2022")

>>> tokenizer = AutoTokenizer.from_pretrained("t5-3b")
>>> input_string = "Premise: Sure ,he snorted just to make me feel even better about the already great situation. [Premise - social norm] It's good to make people feel better about a situation. Hypothesis: Sure, he snorted, just rub it in. [Hypothesis - social norm] It's rude to rub something in someone's face when they don't want to. Is there a contradiction or entailment between the premise and hypothesis?"
>>> input_ids = tokenizer.encode(input_string, return_tensors="pt")
>>> output = model.generate(input_ids, max_length=200)
>>> tokenizer.batch_decode(output, skip_special_tokens=True)
['Answer : Contradiction. Explanation : To rub it in means to make someone feel bad about themselves, but in this sentence he is making the speaker feel better about the already great situation.']

More details about DREAM-FLUTE ...

For more details about DREAM-FLUTE, please refer to our:

• 📄Paper: https://arxiv.org/abs/2210.16407
• 💻GitHub Repo: https://github.com/allenai/dream/

This model is part of our DREAM-series of works. This is a line of research where we make use of scene elaboration for building a "mental model" of situation given in text. Check out our GitHub Repo for more!

More details about this model ...

Training and evaluation data

We use the FLUTE dataset for the FigLang2022SharedTask (https://huggingface.co/datasets/ColumbiaNLP/FLUTE) for training this model. ∼7500 samples are provided as the training set. We used a 80-20 split to create our own training (6027 samples) and validation (1507 samples) partitions on which we build our models. For details on how we make use of the training data provided in the FigLang2022 shared task, please refer to https://github.com/allenai/dream/blob/main/FigLang2022SharedTask/Process_Data_Train_Dev_split.ipynb.

Model details

This model is a fine-tuned version of t5-3b.

It achieves the following results on the evaluation set:

• Loss: 0.7615
• Rouge1: 57.6814
• Rouge2: 37.489
• Rougel: 51.4698
• Rougelsum: 51.4842
• Gen Len: 40.8553

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 2
• total_eval_batch_size: 2
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
0.9974	0.33	1000	0.8955	39.7614	27.3853	38.1015	38.0975	19.0
0.8983	0.66	2000	0.8300	41.4031	29.3769	39.5984	39.5886	19.0
0.8764	1.0	3000	0.7855	41.2619	29.5054	39.5859	39.5748	18.9980
0.5603	1.33	4000	0.8033	41.0015	29.8488	39.5492	39.522	18.9980
0.5619	1.66	5000	0.7869	41.3655	30.0581	39.7462	39.7231	19.0
0.5389	1.99	6000	0.7615	41.3902	30.2049	39.8797	39.8779	19.0
0.3325	2.32	7000	0.8776	41.1737	30.3441	39.6744	39.652	18.9954
0.3509	2.65	8000	0.8501	41.2653	30.5342	39.7315	39.7252	18.9907
0.3499	2.99	9000	0.8546	41.6401	31.0585	40.2659	40.2487	18.9907

Framework versions

• Transformers 4.22.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1