Term Context Sentences
Term-Context Sentences Sentence dataset, standalone. Text only, kept apart from the sentence banks. How every dataset fits together: the dataset index. Training data for danielrosehill/term-context-sentence-generator-granite-4.2-3b-v1: given a term, a one-line context and a count n, write n natural English sentences that each use the term once, in the meaning the context describes. The sentences are short (6 to 18 words) and are made to be read aloud when recording… See the full description on the dataset page: https://huggingface.co/datasets/danielrosehill/term-context-sentences.
About This Project
This project provides a structured dataset containing over 10,000 synthetic sentences designed to improve automatic speech recognition (ASR) performance. Each entry pairs a specific term—including romanized Hebrew and Yiddish words, Israeli locations, and brand names—with a one-line context to ensure the generated sentences reflect the intended meaning of the vocabulary.
The dataset is formatted for machine learning workflows and is compatible with libraries such as pandas, polars, and mlcroissant. It serves as the training foundation for the danielrosehill/term-context-sentence-generator-granite-4.2-3b-v1 model, making it a valuable resource for developers working on speech-to-text systems that require high accuracy for niche, multilingual, or code-switched terminology.