Salvato in:
| Autori principali: | , , , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2511.23119 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917313519812608 |
|---|---|
| author | Liu, Mengjie Peng, Jiahui Ning, Wenchang Chu, Pei Qiu, Jiantao Ma, Ren Zhu, He Min, Rui Lu, Lindong Hou, Linfeng Liu, Kaiwen Qu, Yuan Li, Zhenxiang Xu, Chao Tu, Zhongying Zhang, Wentao He, Conghui |
| author_facet | Liu, Mengjie Peng, Jiahui Ning, Wenchang Chu, Pei Qiu, Jiantao Ma, Ren Zhu, He Min, Rui Lu, Lindong Hou, Linfeng Liu, Kaiwen Qu, Yuan Li, Zhenxiang Xu, Chao Tu, Zhongying Zhang, Wentao He, Conghui |
| contents | High-quality main content extraction from web pages is a critical prerequisite for constructing large-scale training corpora. While traditional heuristic extractors are efficient, they lack the semantic reasoning required to handle the structural heterogeneity of the modern web. Conversely, well-pretrained generative Large Language Models (LLMs) offer superior document comprehension but are prohibited by excessive computational costs, limited context windows, and hallucination risks when applied at web scale. We present \textbf{Dripper}, a lightweight framework that resolves these bottlenecks through four contributions: (1) We reformulate extraction as a \textbf{constrained sequence labeling} task using SLMs (Small Language Models). This paradigm eliminates generative hallucinations and achieves exceptional efficiency, reaching a throughput of 3.08 pages per second on a single A100 GPU. (2) We construct \textbf{WebMainBench}, a rigorous benchmark of 7,809 human-annotated pages covering 5,434 unique domains and multiple languages. Evaluations show our Dripper-0.6B model \textbf{outperforms} heuristics like Trafilatura and rivals massive models like DeepSeek-V3.2(685B), GPT-5 and Gemini-2.5-Pro, offering an optimal efficiency-accuracy trade-off. (3) We demonstrate infrastructural value by \textbf{pre-training a 1B model} on a Dripper-curated corpus (63B tokens). This model significantly outperforms baselines in downstream tasks, proving the critical role of extraction quality and the effectiveness of our framework. (4) We \textbf{open-source} the Dripper-0.6B weights and codebase to facilitate the construction of high-quality datasets. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_23119 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Dripper: Token-Efficient Main HTML Extraction with a Lightweight LM Liu, Mengjie Peng, Jiahui Ning, Wenchang Chu, Pei Qiu, Jiantao Ma, Ren Zhu, He Min, Rui Lu, Lindong Hou, Linfeng Liu, Kaiwen Qu, Yuan Li, Zhenxiang Xu, Chao Tu, Zhongying Zhang, Wentao He, Conghui Computation and Language High-quality main content extraction from web pages is a critical prerequisite for constructing large-scale training corpora. While traditional heuristic extractors are efficient, they lack the semantic reasoning required to handle the structural heterogeneity of the modern web. Conversely, well-pretrained generative Large Language Models (LLMs) offer superior document comprehension but are prohibited by excessive computational costs, limited context windows, and hallucination risks when applied at web scale. We present \textbf{Dripper}, a lightweight framework that resolves these bottlenecks through four contributions: (1) We reformulate extraction as a \textbf{constrained sequence labeling} task using SLMs (Small Language Models). This paradigm eliminates generative hallucinations and achieves exceptional efficiency, reaching a throughput of 3.08 pages per second on a single A100 GPU. (2) We construct \textbf{WebMainBench}, a rigorous benchmark of 7,809 human-annotated pages covering 5,434 unique domains and multiple languages. Evaluations show our Dripper-0.6B model \textbf{outperforms} heuristics like Trafilatura and rivals massive models like DeepSeek-V3.2(685B), GPT-5 and Gemini-2.5-Pro, offering an optimal efficiency-accuracy trade-off. (3) We demonstrate infrastructural value by \textbf{pre-training a 1B model} on a Dripper-curated corpus (63B tokens). This model significantly outperforms baselines in downstream tasks, proving the critical role of extraction quality and the effectiveness of our framework. (4) We \textbf{open-source} the Dripper-0.6B weights and codebase to facilitate the construction of high-quality datasets. |
| title | Dripper: Token-Efficient Main HTML Extraction with a Lightweight LM |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2511.23119 |