doc_id stringlengths 16 16 ⌀ | vector list |
|---|---|
a676d1cd8b94d163 | [
0.043212890625,
-0.006763458251953125,
-0.038909912109375,
-0.011871337890625,
0.0078887939453125,
0.014404296875,
-0.0213623046875,
0.035308837890625,
0.01290130615234375,
-0.029327392578125,
0.0161285400390625,
0.03271484375,
-0.054443359375,
-0.0631103515625,
0.031982421875,
0.03031... |
8d3d7adc3464f6f3 | [
0.0217437744140625,
0.00428009033203125,
-0.004161834716796875,
0.021453857421875,
0.03033447265625,
0.0222320556640625,
-0.02630615234375,
0.0257720947265625,
0.0132904052734375,
-0.00328826904296875,
-0.014892578125,
-0.041290283203125,
-0.01165771484375,
-0.0255126953125,
0.0087585449... |
6219119513fe4a32 | [
-0.035003662109375,
-0.0149688720703125,
0.045867919921875,
-0.00885009765625,
-0.0194854736328125,
-0.006175994873046875,
0.02337646484375,
0.0223541259765625,
0.005615234375,
-0.012725830078125,
0.05975341796875,
0.0007338523864746094,
-0.01029205322265625,
-0.0010309219360351562,
0.09... |
c77e8e28d10aa168 | [
-0.029388427734375,
0.0185089111328125,
0.09375,
0.020904541015625,
-0.00849151611328125,
-0.0016355514526367188,
-0.037933349609375,
0.03656005859375,
0.028717041015625,
0.0148773193359375,
0.0185699462890625,
-0.01181793212890625,
-0.036956787109375,
-0.0206451416015625,
0.006172180175... |
d50e1ad6f21d48d1 | [
-0.0015344619750976562,
0.087890625,
0.0308074951171875,
0.02142333984375,
-0.0141754150390625,
-0.006931304931640625,
0.017578125,
0.01465606689453125,
-0.017425537109375,
0.0145416259765625,
0.05010986328125,
-0.029388427734375,
-0.037750244140625,
0.04425048828125,
0.0158843994140625,... |
d8b8357e5817aa0a | [
0.0198516845703125,
-0.03448486328125,
0.02874755859375,
-0.01297760009765625,
-0.06329345703125,
0.046478271484375,
-0.0245208740234375,
0.0156402587890625,
0.03509521484375,
-0.03375244140625,
0.041900634765625,
-0.037200927734375,
0.004985809326171875,
-0.06658935546875,
0.03384399414... |
f86100a033c753b3 | [0.050567626953125,0.00171661376953125,-0.01030731201171875,-0.0029773712158203125,0.021347045898437(...TRUNCATED) |
3a7364e4a66ffcbb | [0.0095977783203125,0.042449951171875,0.018341064453125,0.0252532958984375,-0.009246826171875,-0.000(...TRUNCATED) |
5d3725e0b2fccd51 | [0.024139404296875,0.01293182373046875,0.0186004638671875,0.06561279296875,0.006298065185546875,0.00(...TRUNCATED) |
2080c36ef2c9b504 | [-0.01739501953125,0.003509521484375,-0.0091400146484375,0.0238037109375,0.04034423828125,0.00236892(...TRUNCATED) |
End of preview. Expand in Data Studio
newsmood-data
Data artifacts produced by the newsmood project for evaluating news-embedding trading signals on US equities.
Contents
| File | Size | Description |
|---|---|---|
embeddings.parquet |
~1.0 GB | 359,962 article embeddings as 1536-d float32 vectors from OpenAI text-embedding-3-small. Columns: doc_id (SHA-1[:16] of canonical text), vector. |
news.parquet |
~213 MB | Canonical news DataFrame from newsmood.ingest.ingest_searches. 359,948 rows. Columns: doc_id, ts, source, query, title, description, text_to_embed, link, source_id, category, country, language, key. |
raw_searches.tar.gz |
~329 MB | Raw JSON snapshots from newsdata.io, yahoo_finance, and yahoo_finance_headlines. Use only if you need fields beyond what news.parquet carries; the canonical DataFrame is the recommended entry point. |
ohlcv/*.parquet |
~280 KB | Per-ticker daily OHLCV from yfinance for 14 SPY-universe tickers (AAPL, AMD, AMZN, BAC, COIN, GOOGL, JPM, META, MSFT, MSTR, NFLX, NVDA, SPY, TSLA). |
Time span
January 2025 – May 2026 (~17 months).
Usage
from newsmood.data import hf_embedding_cache, hf_news_dataframe
# Read-through cache: downloads once to ~/.cache/huggingface/, reuses thereafter
vectors = hf_embedding_cache() # MutableMapping[doc_id -> list[float]]
vec = vectors["a1b2c3d4e5f6abcd"]
# Canonical news DataFrame
df = hf_news_dataframe() # pandas.DataFrame
Or via plain huggingface_hub:
from huggingface_hub import hf_hub_download
import pandas as pd
path = hf_hub_download("thorwhalen/newsmood-data", "news.parquet", repo_type="dataset")
df = pd.read_parquet(path)
Reproducibility
These artifacts were generated by the newsmood pipeline at commit history pinned in
newsmood/misc/docs/PHASE_1_FINDINGS.md.
Re-running the full embedding step from scratch via text-embedding-3-small costs ≈ $0.20 USD and ~90 min wall time.
License
MIT for the structure/code; underlying news content remains the property of its respective publishers. The raw JSON tarball is included for reproducibility only — redistribute responsibly.
- Downloads last month
- 73