| Challenge: | Existing methods for community membership detection using only text features are not effective. |
| Approach: | They propose an unsupervised task for learning user embeddings using text features . they propose a proxy task for embeddables that uses two embeddibles from the same account . |
| Outcome: | The proposed model is more effective than unsupervised representations of user embeddings with 16 different communities. |
Similar Papers
Interpretable Text Embeddings and Text Similarity Explanation: A Survey (2025.emnlp-main)
Copied to clipboard
| Challenge: | Text embeddings are a fundamental component in many NLP tasks, but their interpretation and explanation remain challenging. |
| Approach: | They propose a framework for interpretable text embeddings and text similarity explanation . they characterize the main ideas, approaches, and trade-offs and discuss lessons learned . |
| Outcome: | The proposed methods are compared with existing models and compare them with existing ones. |
Self-Discriminative Learning for Unsupervised Document Embedding (N19-1)
Copied to clipboard
| Challenge: | Existing methods for document embedding learning do not consider inter-document relationships. |
| Approach: | They propose to exploit the inter-document information and directly model the relations of documents in embedding space with a discriminative network and a novel objective. |
| Outcome: | The proposed method has errors that are 5 to 13% lower than state-of-the-art models and is even more pronounced in scarce label setting. |
Representing Social Media Users for Sarcasm Detection (D18-1)
Copied to clipboard
| Challenge: | Existing annotated corpus of Reddit comments is limited by available annotation methods. |
| Approach: | They propose a Bayesian approach that directly represents authors’ propensities to be sarcastic and a dense embedding approach that can learn interactions between the author and the text. |
| Outcome: | The proposed approach performs better in homogeneous contexts, whereas the dense embeddings prove valuable in more diverse contexts. |
Entity or Relation Embeddings? An Analysis of Encoding Strategies for Relation Extraction (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to relation extraction use concatenating embeddings of head and tail entities . however, such representations capture the types of the entities involved, leading to false positives and confusion between relations involving entities of the same type. |
| Approach: | They propose a model which combines [MASK] embeddings with entity embedds to learn relation embeddations. |
| Outcome: | The proposed model outperforms the state-of-the-art on several benchmarks . it uses a self-supervised pre-training strategy which further improves the results. |
Using RL to Identify Divisive Perspectives Improves LLMs Abilities to Identify Communities on Social Media (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Experimental results show improvements on Reddit and Twitter data . |
| Approach: | They propose to take advantage of Large Language Models (LLMs) to better identify user communities. |
| Outcome: | The proposed model improves on Reddit and Twitter data and tasks of community detection, bot detection, and news media profiling. |
One-Model-Connects-All: A Unified Graph Pre-Training Model for Online Community Modeling (2023.findings-emnlp)
Copied to clipboard
Ruoxue Ma, Jiarong Xu, Xinnong Zhang, Haozhe Zhang, Zuyu Zhao, Qi Zhang, Xuanjing Huang, Zhongyu Wei
| Challenge: | Existing studies on integrating online community to solve social problems have not fully utilized these three components and the relationship among them. |
| Approach: | They propose a framework that simultaneously considers communities, users, and texts and can easily connect with a variety of downstream tasks related to social media. |
| Outcome: | The proposed model can be used to perform violation detection, sentiment analysis, and community recommendation across multiple tasks. |
A supervised approach to taxonomy extraction using word embeddings (L18-1)
Copied to clipboard
| Challenge: | a recent evaluation of a method for organizing texts into a hierarchy showed that it did not outperform a baseline. |
| Approach: | They propose a method that uses supervised learning to combine multiple features with a support vector machine classifier including the baseline features. |
| Outcome: | The proposed method outperforms the baseline method and provides stronger method for identifying taxonomic relations than previous methods. |
LLMs Enable Bag-of-Texts Representations for Short-Text Clustering (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for short text clustering require labeling and no embeddings optimization. |
| Approach: | They propose a training-free method for unsupervised short text clustering that relies less on careful selection of embedders than other methods. |
| Outcome: | The proposed method achieves comparable or superior results to state-of-the-art methods, but without embeddings optimization or prior knowledge of clusters or labels. |
PeerDA: Data Augmentation via Modeling Peer Relation for Span Identification Tasks (2023.acl-long)
Copied to clipboard
| Challenge: | Experimental results on ten datasets across seven domains demonstrate the effectiveness of PeerDA. |
| Approach: | They propose a new approach which uses span pairs with the PR relation as the augmentation data for training. |
| Outcome: | The proposed approach achieves state-of-the-art results on ten datasets across seven domains. |
Know Who Your Friends Are: Understanding Social Connections from Unstructured Text (N18-5)
Copied to clipboard
Léa Deleris, Francesca Bonin, Elizabeth Daly, Stéphane Deparis, Yufang Hou, Charles Jochim, Yassine Lassoued, Killian Levacher
| Challenge: | Having an understanding of interpersonal relationships is helpful in many contexts. |
| Approach: | They propose a system that extracts qualitative and quantitative information from texts and aggregates it to provide a condensed view of relationships. |
| Outcome: | The proposed system extracts qualitative and quantitative information elements about interactions and aggregates those to provide a condensed view of relationships. |