HITgram: A Platform for Experimenting with n-gram Language Models

TL;DR

HITgram is a lightweight, efficient platform for experimenting with n-gram language models, enabling resource-constrained environments to perform language modeling tasks with high speed and flexibility.

Contribution

It introduces a novel, resource-efficient platform supporting n-gram models with advanced features like context weighting and smoothing, optimized for speed and scalability.

Findings

Achieves 50,000 tokens/second in experiments

Constructs 2-grams from 320MB corpus in 62 seconds

Scales to 4-grams from 1GB in under 298 seconds

Abstract

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.