HITgram: A Platform for Experimenting with n-gram Language Models
محفوظ في:
| الحاوية / القاعدة: | arXiv.org (Dec 14, 2024), p. n/a |
|---|---|
| المؤلف الرئيسي: | |
| مؤلفون آخرون: | , , , , |
| منشور في: |
Cornell University Library, arXiv.org
|
| الموضوعات: | |
| الوصول للمادة أونلاين: | Citation/Abstract Full text outside of ProQuest |
| الوسوم: |
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| مستخلص: | Large language models (LLMs) are powerful but resource intensive, limiting accessibility. HITgram addresses this gap by offering a lightweight platform for n-gram model experimentation, ideal for resource-constrained environments. It supports unigrams to 4-grams and incorporates features like context sensitive weighting, Laplace smoothing, and dynamic corpus management to e-hance prediction accuracy, even for unseen word sequences. Experiments demonstrate HITgram's efficiency, achieving 50,000 tokens/second and generating 2-grams from a 320MB corpus in 62 seconds. HITgram scales efficiently, constructing 4-grams from a 1GB file in under 298 seconds on an 8 GB RAM system. Planned enhancements include multilingual support, advanced smoothing, parallel processing, and model saving, further broadening its utility. |
|---|---|
| تدمد: | 2331-8422 |
| المصدر: | Engineering Database |