Papers by Ho-Jin Choi
Enhancing Arguments Recognition for Financial Mathematical Reasoning over Hybrid Data (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for question answering on textual data are difficult to train and pose a misrecognition problem. |
| Approach: | They propose an approach to train a reasoning program generator to improve argument recognition by aggregating arguments and loss argument set. |
| Outcome: | The proposed method improves the probabilities of proper arguments in a reasoning program generation so that arguments comprising the ground truth have higher weights. |
DialogCC: An Automated Pipeline for Creating High-Quality Multi-Modal Dialogue Dataset (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing multi-modal dialogue datasets that focus on image-based dialogues have low quality and limited diversity of images per dialogue. |
| Approach: | They propose to construct a multi-modal dialogue dataset that guarantees both dialogue quality and image diversity without requiring minimum human effort. |
| Outcome: | The proposed dataset outperforms existing datasets in terms of quality and diversity in human evaluation. |
Does GPT-3 Generate Empathetic Dialogues? A Novel In-Context Example Selection Method and Automatic Evaluation Metric for Empathetic Dialogue Generation (2022.coling-1)
Copied to clipboard
| Challenge: | Empathy is a multi-dimensional concept consisting of cognitive and affective aspects. |
| Approach: | They propose two new in-context example selection methods that utilize emotion and situational information. |
| Outcome: | The proposed method is effective in measuring the degree of human empathy. |
Korean TimeBank Including Relative Temporal Information (L18-1)
Copied to clipboard
| Challenge: | Temporal information extraction is one of the important research fields in natural language processing. |
| Approach: | They propose a concept of relative temporal information and supplement a Korean annotation language to represent new relative expressions and extend an annotated dataset through the revised language. |
| Outcome: | The proposed language can be used to represent relative temporal information and extend an annotated dataset, Korean TimeBank, through the revised language. |
Pneg: Prompt-based Negative Response Generation for Dialogue Response Selection Task (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for synthesizing adversarial negative responses are limited by their scalability and cost. |
| Approach: | They propose a method for generating adversarial negative responses using a large-scale language model. |
| Outcome: | The proposed method outperforms other methods on dialogue selection tasks. |
Stark: Social Long-Term Multi-Modal Conversation with Persona Commonsense Knowledge (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies focus on image-sharing behavior in singular sessions, leading to limited long-term social interaction. |
| Approach: | They propose a large-scale long-term multi-modal dialogue dataset that generates long-time multi-modity dialogue distilled from ChatGPT and proposed image aligner. |
| Outcome: | The proposed framework generates long-term multi-modal dialogue from ChatGPT and image aligner. |
Large Language Models can Share Images, Too! (2024.findings-acl)
Copied to clipboard
| Challenge: | Using a zero-shot prompting, large language models can be used to share images in a multi-tasking environment. |
| Approach: | They introduce a dataset that includes enriched annotations and a framework to evaluate LLMs. |
| Outcome: | The proposed framework unlocks image-sharing capability of LLMs in zero-shot prompting, with ChatGPT achieving the best performance. |
Korean-Specific Emotion Annotation Procedure Using N-Gram-Based Distant Supervision and Korean-Specific-Feature-Based Distant Supervision (2020.lrec-1)
Copied to clipboard
| Challenge: | Existing methods to annotate unlabeled data with emotions are expensive and time-consuming. |
| Approach: | They propose an annotation procedure that leverages Korean emotion lexicons and Korean-specific emotion features to annotate unlabeled data. |
| Outcome: | The proposed procedure compares with the KTEA dataset and a large-scale emotion-labeled dataset. |
Constructing Multi-Modal Dialogue Dataset by Replacing Text with Semantically Relevant Images (2021.acl-short)
Copied to clipboard
| Challenge: | Existing training methods for multi-modal dialogue systems rely on image captioning or visual question answering datasets that are irrelevant to the dialogue context. |
| Approach: | They propose to create a 45k multi-modal dialogue dataset with minimal human intervention . they use text dialogue datasets, image-mixed dialogues and contextual-similarity filtering . |
| Outcome: | The proposed dataset can be used as training data for multi-modal dialogue systems . human evaluations show that the model can be effectively used . |