Improving Factual Accuracy of Neural Table-to-Text Output by Addressing Input Problems in ToTTo (2024.naacl-long)
Copied to clipboard
| Challenge: | Neural Table-to-Text models produce hallucinated outputs that are factually incorrect or unrelated to the input data. |
| Approach: | They manually annotated 1,837 texts generated by multiple Neural Table-to-Text models in the politics domain of the ToTTo dataset. |
| Outcome: | The proposed model reduces factual errors by 52% to 76% . the proposed model also struggles with tabular inputs that are structured in a non-standard way, especially when the input lacks distinct row and column values or the column headers are not correctly mapped to corresponding values. |
Similar Papers
ToTTo: A Controlled Table-To-Text Generation Dataset (2020.emnlp-main)
Copied to clipboard
Ankur Parikh, Xuezhi Wang, Sebastian Gehrmann, Manaal Faruqui, Bhuwan Dhingra, Diyi Yang, Dipanjan Das
| Challenge: | Existing methods for data-to-text generation often hallucinate phrases not supported by the Wikipedia table. |
| Approach: | They propose a controlled task where annotators directly revise existing Wikipedia sentences to generate a one-sentence description. |
| Outcome: | The proposed task produces a one-sentence description from a Wikipedia table and highlighted cells. |
Revisiting Pathologies of Neural Models under Input Reduction (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have shown that modern neural models tend to be miscalibrated. |
| Approach: | They examine why models produce high-confidence predictions on inputs that appear nonsensical to humans . previous work suggested that models fail to assign low probabilities due to model overconfidence . |
| Outcome: | The proposed methods can be extended to reduce the number of examples but with the cost of miscalibration. |
Controlled Hallucinations: Learning to Generate Faithfully from Noisy Data (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Neural text generation (data- or text-to-text) demonstrates remarkable performance when training data is abundant which for many applications is not the case. |
| Approach: | They propose a technique to treat hallucinations as a controllable aspect of the generated text without dismissing any input and without modifying the model architecture. |
| Outcome: | The proposed technique can be used on a WikiBio dataset and in a human evaluation. |
Plan-then-Generate: Controlled Data-to-Text Generation via Planning (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies focus on producing results that are close to the references, i.e. what to generate and in what order (the output structure) cannot be explicitly controlled by the users. |
| Approach: | They propose a Plan-then-Generate framework to improve the controllability of neural data-to-text models. |
| Outcome: | The proposed model can control both the intra-sentence and inter-sentent structure of the generated output. |
Evaluating Factuality in Text Simplification (2022.acl-long)
Copied to clipboard
| Challenge: | Automated simplification models aim to make input texts more readable without altering their meaning. |
| Approach: | They propose a taxonomy of errors that are used to analyze simplification models . they propose to use simplification methods to make input texts more readable . |
| Outcome: | The proposed models introduce errors that are not captured by existing evaluation metrics. |
On the Origin of Hallucinations in Conversational Models: Is it the Datasets or the Models? (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing knowledge-grounded conversational benchmarks produce factually invalid statements, a phenomenon commonly called hallucination. |
| Approach: | They conduct a human study on knowledge-grounded conversational benchmarks and state-of-the-art models. |
| Outcome: | The findings raise important questions on the quality of existing datasets and models. |
What’s Going On in Neural Constituency Parsers? An Analysis (N18-1)
Copied to clipboard
| Challenge: | a number of differences have emerged between classical and modern constituency parsing approaches . structural components like grammars and feature-rich lexicons are becoming less central . recurrent neural networks have gained traction as a powerful and general purpose tool for representation . |
| Approach: | They propose a model that implicitly learns to encode much of the same information as grammars and lexicons in the past. |
| Outcome: | The proposed model outperforms state-of-the-art models under similar conditions. |
Searching for Search Errors in Neural Morphological Inflection (2021.eacl-main)
Copied to clipboard
| Challenge: | Neural sequence-to-sequence models are the predominant choice for language generation tasks. |
| Approach: | They find that on word-level tasks, the empty string is often the global optimum . they suggest that the poor calibration of many neural models may stem from characteristics of a specific subset of tasks rather than general ill-suitedness of such models for language generation. |
| Outcome: | The results suggest that the poor calibration of many neural models may stem from characteristics of a specific subset of tasks rather than general ill-suitedness of such models for language generation. |
A Simple Recipe towards Reducing Hallucination in Neural Surface Realisation (P19-1)
Copied to clipboard
| Challenge: | Recent neural language generation systems often hallucinate contents when trained on loosely corresponding pairs of the input structure and text. |
| Approach: | They propose to integrate a language understanding module for data refinement with self-training iterations to induce strong equivalence between the input data and the paired text. |
| Outcome: | Experiments on the E2E challenge dataset show that the proposed framework reduces relative unaligned noise by 50% compared with the current state-of-the-art ensemble generator. |
Improving User Controlled Table-To-Text Generation Robustness (2023.findings-eacl)
Copied to clipboard
| Challenge: | In experiments, models perform well on test sets coming from the same distribution as the train data but their performance drops when evaluated on realistic noisy user inputs. |
| Approach: | They propose a user controlled table-to-text generation task where users explore the content in a table by selecting cells and reading a natural language description thereof. |
| Outcome: | The proposed model gains 4.85 BLEU points on user noisy test cases and 1.4 on clean test cases. |