Many Faces of Feature Importance: Comparing Built-in and Post-hoc Feature Importance in Text Classification (D19-1)
Copied to clipboard
| Challenge: | Feature importance is commonly used to explain machine predictions . however, the consistency of feature importance via different methods remains understudied . |
| Approach: | They compare feature importance from built-in mechanisms and post-hoc methods that approximate model behavior to find similarities between models. |
| Outcome: | The proposed methods show that features from traditional models are more similar with each other than with deep learning models. |
Similar Papers
Assessing Word Importance Using Models Trained for Semantic Tasks (2023.findings-acl)
Copied to clipboard
| Challenge: | Many NLP tasks require to automatically identify the most significant words in a text. |
| Approach: | They propose to use attribution methods to explain the predictions of two NLP tasks to derive word significance from models trained to solve semantic tasks. |
| Outcome: | The proposed method is robust to the initial task and is able to identify important words in sentences without explicit word importance labeling in training. |
Relative Importance in Sentence Processing (2021.acl-short)
Copied to clipboard
| Challenge: | In natural language processing, the relative importance of words is usually interpreted with respect to a specific task. |
| Approach: | They compare the relative importance of words in English language processing by humans and neural language models by using saliency methods. |
| Outcome: | The proposed method could be used to interpret neural language models. |
Is Attention Interpretable? (P19-1)
Copied to clipboard
| Challenge: | Attention mechanisms have recently boosted performance on a range of NLP tasks. |
| Approach: | They propose to manipulate attention weights in text classification models and analyze the resulting differences in their predictions. |
| Outcome: | The proposed approach improves models' predictions by using gradient-based rankings of attention weights. |
What Matters to an LLM? Behavioral and Computational Evidences from Summarization (2026.findings-eacl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly entrusted with the management of information. |
| Approach: | They combine behavioral and computational analyses to find out what LLMs prioritize . they generate length-controlled summaries and derive empirical importance distributions . |
| Outcome: | The proposed model converges on consistent importance patterns and clusters more by family than by size. |
Modeling Content Importance for Summarization with Pre-trained Language Models (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on content importance do not consider semantics and context when evaluating importance. |
| Approach: | They apply information theory to pre-trained language models to define the concept of importance from the perspective of information amount. |
| Outcome: | Experiments on CNN/Daily Mail and New York Times show that the proposed model can model the importance of content better than previous methods based on F1 and ROUGE scores. |
“Will You Find These Shortcuts?” A Protocol for Evaluating the Faithfulness of Input Salience Methods for Text Classification (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing work on faithfulness evaluation is not conclusive and does not provide a clear answer as to how different methods are to be compared. |
| Approach: | They propose a protocol for faithfulness evaluation that makes use of partially synthetic data to obtain ground truth for feature importance ranking. |
| Outcome: | The proposed method is based on partially synthetic data and is compared with lexical shortcuts on a range of datasets and LSTM models. |
Necessity and Sufficiency for Explaining Text Classifiers: A Case Study in Hate Speech Detection (2022.naacl-main)
Copied to clipboard
| Challenge: | XAI features usually provide a single importance score for each token, but feature attribution methods provide two complementary and theoretically-grounded scores for each utterance. |
| Approach: | They propose a feature attribution method that generates explicit perturbations of the input text, allowing the importance scores themselves to be explainable. |
| Outcome: | The proposed method explain the predictions of hate speech detection models on a set of curated examples from a test suite. |
On Importance Sampling-Based Evaluation of Latent Language Models (2020.acl-main)
Copied to clipboard
| Challenge: | Existing approaches to evaluate language models using latent structures are intractable as they require marginalizing over the latent space. |
| Approach: | They propose to use importance sampling to evaluate latent language models . they elucidate subtle differences in how importance sampling is applied . |
| Outcome: | The proposed model performs better on tasks requiring structure and interpretability. |
On the Importance of Delexicalization for Fact Verification (D19-1)
Copied to clipboard
| Challenge: | Neural networks (NNs) perform state-of-the-art (SOA) performance in many complex tasks. |
| Approach: | They investigate the importance that a model assigns to various aspects of data . they experiment with two strategies of masking to mitigate this dependence on lexicalized information . |
| Outcome: | The proposed model improves on the in-domain dataset by 10% compared to the fully lexicalized model. |
Rethinking Attribute Representation and Injection for Sentiment Classification (D19-1)
Copied to clipboard
| Challenge: | Existing models that use text attributes to improve sentiment classification use text as a categorical feature. |
| Approach: | They propose to represent attributes as chunk-wise importance weight matrices and consider four locations to inject attributes. |
| Outcome: | The proposed method outperforms the state-of-the-art and outperformed previous models. |