--- configs: - config_name: default data_files: - split: train path: "html_extraction_task_train_dataset.jsonl" - split: test path: "html_extraction_task_test_dataset.jsonl" - split: unseen_test path: "html_extraction_task_unseen_test_dataset.jsonl" --- # HTML to JSON Information Extraction Dataset ## Description The `html_to_json_information_extraction` dataset is a collection of over 7300 HTML snippets and their extracted information in JSON. These HTML have been sourced (scraped) from about 25 companies' career pages. The dataset contains three splits - `train`, `test`, `unseen_test`. This dataset has been built to fine tune SLMs & LLMs for the information extraction task. #### `train` split This split contains over 5700 pair of HTML-JSON data points which have been sourced (scraped) from 22 companies' career pages. #### `test` split This split contains over 1300 pair of HTML-JSON data points which have been sourced (scraped) from the same 22 companies' career pages as the `train` split. #### `unseen_test` split This split contains over 300 pair of HTML-JSON data points, and, these have been sourced (scraped) from 3 companies' career pages. These 3 companies are not present in the list of 22 companies from which `test` & `train` splits have been built. ## Sample Here is a sample HTML and its label - extracted information in JSON. #### HTML ``` true"> Frontend Engineer
New York City
Fullstack Engineer
New York