Thailand

To be included into chatbot systems, Large language models (LLMs) must be aligned with human conversational conventions. However, being trained mainly on web-scraped data gives existing LLMs a voice closer to informational text than actual human speech. In this paper, we examine the effect of decoding methods on the alignment between LLM-generated and human conversations, including Beam Search, Top K Sampling, and Nucleus Sampling. We present new measures of alignment in substance, style, and psychometric orientation, and experiment with two conversation datasets. Our results provide subtle insights: better alignment is attributed to fewer beams in Beam Search and lower values of P in Nucleus Sampling. We also find that task-oriented and open-ended datasets perform differently in terms of alignment, indicating the significance of taking into account the context of the interaction.

ACL 2024

Impact of Decoding Methods on Human Alignment of Conversational LLMs

llms

alignment

workshop paper

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

While large language models (LLMs) show promise for various tasks, their performance in compound aspect-based sentiment analysis (ABSA) tasks lags behind fine-tuned models. However, the potential of LLMs fine-tuned for ABSA remains unexplored. This paper examines the capabilities of open-source LLMs fine-tuned for ABSA, focusing on LLaMA-based models. We evaluate the performance across four tasks and eight English datasets, finding that the fine-tuned Orca 2 model surpasses state-of-the-art results in all tasks. However, all models struggle in zero-shot and few-shot scenarios compared to fully fine-tuned ones. Additionally, we conduct error analysis to identify challenges faced by fine-tuned models.

LLaMA-Based Models for Aspect-Based Sentiment Analysis

Cross-lingual emotion detection allows us to analyze global trends, public opinion, and so- cial phenomena at scale. We participated in the Explainability of Cross-lingual Emotion De- tection (EXALT) shared task, achieving an F1- score of 0.6046 on the evaluation set for the emotion detection sub-task. Our system outper- formed the baseline by more than 0.16 F1-score absolute, and ranked second amongst compet- ing systems. We conducted experiments using fine-tuning, zero-shot learning, and few-shot learning for Large Language Model (LLM)- based models as well as embedding-based BiL- STM and KNN for non-LLM-based techniques. Additionally, we introduced two novel meth- ods: the Multi-Iteration Agentic Workflow and the Multi-Binary-Classifier Agentic Workflow. We found that LLM-based approaches provided good performance on multilingual emotion de- tection. Furthermore, ensembles combining all our experimented models yielded higher F1- scores than any single approach alone.

TEII: Think, Explain, Interact and Iterate with Large Language Models toSolve Cross-lingual Emotion Detection"

Cross-lingual emotion detection faces challenges such as imbalanced label distribution, data scarcity, cultural and linguistic differences, figurative language, and the opaqueness of pre-trained language models. This paper presents our approach to the EXALT shared task at WASSA 2024, focusing on emotion transferability across languages and trigger word identification. We employ data augmentation techniques, including back-translation and synonym replacement, to address data scarcity and imbalance issues in the emotion detection sub-task. For the emotion trigger identification sub-task, we utilize token and label mapping to capture emotional information at the subword level. Our system achieves competitive performance, ranking 13th, 1st, and 2nd in the Emotion Detection, Binary Trigger Word Detection, and Numerical Trigger Word Detection tasks.

Enhancing Cross-Lingual Emotion Detection with Data Augmentation and
Token-Label Mapping

This study describes the model design of the NYCU-NLP system for the EXALT shared task at the WASSA 2024 workshop. We instruction-tune several large language models and then assemble various model combinations as our main system architecture for cross-lingual emotion and trigger detection in tweets. Experimental results showed that our best performing submission is an assembly of the Starling (7B) and Llama 3 (8B) models. Our submission was ranked sixth of 17 participating systems for the emotion detection subtask, and fifth of 7 systems for the binary trigger detection subtask.

NYCU-NLP at EXALT 2024: Assembling Large Language Models for
Cross-Lingual Emotion and Trigger Detection

This paper describes our task 1 submission for the WASSA 2024 shared task on Explainability for Cross-lingual Emotion in Tweets. Our task is to predict the correct emotion label (Anger, Sadness, Fear, Joy, Love, and Neutral) for a dataset of English, Dutch, French, Spanish, and Russian tweets, while training exclusively on English emotion labeled data, to reveal what kind of emotion detection information is trans- ferable cross-language (Maladry et al., 2024). To that end, we used an ensemble of mod- els with a GPT-4 decider. Our ensemble con- sisted of a few-shot GPT-4 prompt system and a TwHIN-BERT system fine-tuned on the EX- ALT and additional English data. We ranked 8th place under the name WU_TLAXE with an F1 Macro score of 0.573 on the test set. We also experimented with an English-only TwHIN-BERT model by translating the other languages into English for inference, which proved to be worse than the other models.

WU_TLAXE at WASSA 2024 Explainability for Cross-Lingual Emotion in
Tweets Shared Task 1: Emotion through Translation using TwHIN-BERT
and GPT

This presentation explains our submitted systems for WASSA 2024 Shared Task 2: Cross-Lingual Emotion Detection. We implemented a BERT-based classifier and an in-context learning-based system. Our best-performing model, using English Chain of Thought prompts with trigger words, reached 3rd overall with an F1 score of 0.6015. Further analysis on the scalability and transferability of the monolingual English dataset on cross-lingual tasks demonstrates the importance of data quality over quantity. We also found that augmented multilingual data does not necessarily perform better than English monolingual data in cross-lingual tasks.

Effectiveness of Scalable Monolingual Data and Trigger Words Prompting
on Cross-Lingual Emotion Detection Task

This paper presents our system built for the WASSA-2024 Cross-lingual Emotion Detection Shared Task. The task consists of two subtasks: first, to assess an emotion label from six possible classes for a given tweet in one of five languages, and second, to predict words triggering the detected emotions in binary and numerical formats. Our proposed approach revolves around fine-tuning quantized large language models, specifically Orca 2, with low-rank adapters (LoRA) and multilingual Transformer-based models, such as XLM-R and mT5. We enhance performance through machine translation for both subtasks and trigger word switching for the second subtask. The system achieves excellent performance, ranking 1st in numerical trigger words detection, 3rd in binary trigger words detection, and 7th in emotion detection.

UWB at WASSA-2024 Shared Task 2: Cross-lingual Emotion Detection

Emotion detection from text is a crucial task in understanding natural language with wide-ranging applications. Existing approaches for multilingual emotion detection from text face challenges with data scarcity across many languages and a lack of interpretability. We propose a novel method that leverages both monolingual and multilingual pre-trained language models to improve performance and interpretability. Our approach involves 1) training a high-performing English monolingual model in parallel with a multilingual model and 2) using knowledge distillation to transfer the emotion detection capabilities from the monolingual teacher to the multilingual student model. Experiments on a multilingual dataset demonstrate significant performance gains for refined multilingual models like XLM-RoBERTa and E5 after distillation. Furthermore, our approach enhances interpretability by enabling better identification of emotion-trigger words. Our work presents a promising direction for building accurate, robust and explainable multilingual emotion detection systems.

Knowledge Distillation from Monolingual to Multilingual Models for
Intelligent and Interpretable Multilingual Emotion Detection

This paper addresses the shared task of multilingual emotion detection in tweets, presented at the Workshop on Computational Approaches to Subjectivity, Sentiment, and Social Media Analysis (WASSA) co-located with the ACL 2024 conference. The task involves predicting emotions from six classes in tweets from five different languages using only English for model training. Our approach focuses on addressing class imbalance through data augmentation, hierarchical classification, and the application of focal loss and weighted cross-entropy loss functions. These methods enhance our transformer-based model’s ability to transfer emotion detection capabilities across languages, resulting in improved performance despite the constraints of limited computational resources.

PCICUNAM at WASSA 2024: Cross-lingual Emotion Detection Task with
Hierarchical Classification and Weighted Loss Functions

This paper describes the system developed by the HITSZ-HLT team for WASSA-2024 Shared Task 2, which addresses two closely linked sub-tasks: Cross-lingual Emotion Detection and Binary Trigger Word Detection in tweets. The main goal of Shared Task 2 is to simultaneously identify the emotions expressed and detect the trigger words across multiple languages. To achieve this, we introduce a Language-agnostic Multi Task Learning (LaMTL) framework that integrates emotion prediction and emotion trigger word detection tasks. By fostering synergistic interactions between task-specific and task-agnostic representations, the LaMTL aims to mutually enhance emotional cues, ultimately improving the performance of both tasks. Additionally, we leverage large-scale language models to translate the training dataset into multiple languages, thereby fostering the formation of language-agnostic representations within the model, significantly enhancing the model's ability to transfer and perform well across multilingual data. Experimental results demonstrate the effectiveness of our framework across both tasks, with a particular highlight on its success in achieving second place in sub-task 2.

Premium content

Impact of Decoding Methods on Human Alignment of Conversational LLMs

Next from ACL 2024

LLaMA-Based Models for Aspect-Based Sentiment Analysis

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES