Update README.md

README.md

@@ -17,13 +17,33 @@ configs:
 - config_name: summaries
   data_files:
   - split: authors
-    path: "data/summaries/parquet/authors_summary.parquet"
+    path: "summaries/parquet/ol_dump_authors_2024-04-30_summary.parquet"
   - split: works
-    path: "data/summaries/parquet/works_summary.parquet"
+    path: "summaries/parquet/ol_dump_works_2024-04-30_summary.parquet"
   - split: editions
-    path: "data/summaries/parquet/editions_summary.parquet"
+    path: "summaries/parquet/ol_dump_editions_2024-04-30_summary.parquet"
 ---
 
 # OpenLibrary Dump (2024-04-30)
 
-This repo contains the OpenLibrary dump converted to tabular format using DuckDB for easier querying
+This dataset contains the [OpenLibrary dump](https://openlibrary.org/developers/dumps) of April 2024 converted to Parquet and DuckDB for easier querying.
+
+## Formats
+
+### Original GZIP dumps
+
+The original GZIP dumps are available at [data/dumps](https://huggingface.co/datasets/storytracer/openlibrary_dump_2024-04-30/tree/main/data/dumps). The dumps are gzipped TSV files with the original OL JSON record contained in the fifth column of the TSV.
+
+### DuckDB
+
+The authors, works and editions dumps were imported as tables into [data/duckdb/ol_dump_2024-04-30.duckdb](https://huggingface.co/datasets/storytracer/openlibrary_dump_2024-04-30/blob/main/data/duckdb/ol_dump_2024-04-30.duckdb) using the script [ol_duckdb.sh](https://huggingface.co/datasets/storytracer/openlibrary_dump_2024-04-30/blob/main/ol_duckdb.sh). The scripts extract the JSON record from the fifth column and pipes it to DuckDB to import using the flags ```union_by_name=true``` and ```ignore_errors=true``` to account for the inconsistent JSON structure of the dumps.
+
+### Parquet
+
+The authors, works and editions tables were exported as Parquet from DuckDB to [data/parquet](https://huggingface.co/datasets/storytracer/openlibrary_dump_2024-04-30/tree/main/data/parquet). These Parquet files are contained in the default ```data``` config for this dataset.
+
+### Table Summaries
+
+To give users an easy overview of the fields contained in the dump, the DuckDB tables have been summarized using the ```SUMMARIZE``` function as Markdown and Parquet files at [summaries](https://huggingface.co/datasets/storytracer/openlibrary_dump_2024-04-30/tree/main/data/summaries), excluding the summary columns ```min```, ```max```, ```avg``` for easier viewing. You can also explore the table summaries using the dataset viewer by selecting the ```summaries``` config.
+
+The dumps fields are supposed to be consistent with the schema referenced [in the documentation](https://openlibrary.org/developers/dumps#:~:text=Format%20of%20JSON%20records). However, the summaries show that the dumps are not consistent with the documentation: the dumps contain some undocumented fields and some of the fields almost exclusively contain ```null``` values.