DocGraphLM: Documental Graph Language Model for Information Extraction
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913186924462080 |
|---|---|
| author | Wang, Dongsheng Ma, Zhiqiang Nourbakhsh, Armineh Gu, Kang Shah, Sameena |
| author_facet | Wang, Dongsheng Ma, Zhiqiang Nourbakhsh, Armineh Gu, Kang Shah, Sameena |
| contents | Advances in Visually Rich Document Understanding (VrDU) have enabled information extraction and question answering over documents with complex layouts. Two tropes of architectures have emerged -- transformer-based models inspired by LLMs, and Graph Neural Networks. In this paper, we introduce DocGraphLM, a novel framework that combines pre-trained language models with graph semantics. To achieve this, we propose 1) a joint encoder architecture to represent documents, and 2) a novel link prediction approach to reconstruct document graphs. DocGraphLM predicts both directions and distances between nodes using a convergent joint loss function that prioritizes neighborhood restoration and downweighs distant node detection. Our experiments on three SotA datasets show consistent improvement on IE and QA tasks with the adoption of graph features. Moreover, we report that adopting the graph features accelerates convergence in the learning process during training, despite being solely constructed through link prediction. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_02823 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DocGraphLM: Documental Graph Language Model for Information Extraction Wang, Dongsheng Ma, Zhiqiang Nourbakhsh, Armineh Gu, Kang Shah, Sameena Computation and Language Information Retrieval Advances in Visually Rich Document Understanding (VrDU) have enabled information extraction and question answering over documents with complex layouts. Two tropes of architectures have emerged -- transformer-based models inspired by LLMs, and Graph Neural Networks. In this paper, we introduce DocGraphLM, a novel framework that combines pre-trained language models with graph semantics. To achieve this, we propose 1) a joint encoder architecture to represent documents, and 2) a novel link prediction approach to reconstruct document graphs. DocGraphLM predicts both directions and distances between nodes using a convergent joint loss function that prioritizes neighborhood restoration and downweighs distant node detection. Our experiments on three SotA datasets show consistent improvement on IE and QA tasks with the adoption of graph features. Moreover, we report that adopting the graph features accelerates convergence in the learning process during training, despite being solely constructed through link prediction. |
| title | DocGraphLM: Documental Graph Language Model for Information Extraction |
| topic | Computation and Language Information Retrieval |
| url | https://arxiv.org/abs/2401.02823 |