A screener over news attention answers a question price screeners can't: what is everyone suddenly writing about?
/v2/symbols computes it for every entity in one response: mentions (statements), the topics and distinct outlets
behind them, the sentiment split, and the change against the previous period.
Today's movers
period=d compares the last 24 hours with the 24 hours before; sort=change puts the biggest increases first:
You need a RapidAPI key to run the examples: get one here.
curl -G https://finance-pulse.p.rapidapi.com/v2/symbols \
--data-urlencode "period=d" --data-urlencode "sort=change" --data-urlencode "min_mentions=5" \
-H "X-RapidAPI-Key: YOUR_RAPIDAPI_KEY" -H "X-RapidAPI-Host: finance-pulse.p.rapidapi.com"
id kind mentions previous change sources score
IMF org 126 24 +102 63 -0.10
META equity 59 12 +47 40 +0.34
NVDA equity 55 16 +39 25 -0.02
MSFT equity 61 24 +37 42 +0.20
KOSPI index 44 10 +34 16 -0.98
BTC crypto 50 20 +30 16 -0.10
SPX index 30 8 +22 20 +0.20
SBUX equity 22 1 +21 16 -0.27
USD fx 42 21 +21 28 -0.52
STLA equity 27 9 +18 15 -0.30
FED org 35 18 +17 18 -0.20
CRUDE commodity 29 13 +16 20 -0.17
V equity 18 2 +16 13 +0.44
BRENT commodity 27 12 +15 18 -0.30
Read it like a quant, and it is already telling stories: a company's quarter, a bond sell-off, a central bank's decision. Every row also says what kind of thing it is, and that is what makes the list usable.
Filter it
A symbol is an entity: an equity or ETF under its ticker (MU, 0700.HK, SAN.PA), or an index, rate,
currency, crypto asset, commodity, organisation or private company under a short code (SPX, US10Y, USD, BTC,
GOLD_COMMODITY, FED, OPENAI). Each row carries its kind and name, and kind= filters on the server. Two rules turn
the movers into candidates:
import requests
API = "https://finance-pulse.p.rapidapi.com"
HEADERS = {"X-RapidAPI-Key": "YOUR_RAPIDAPI_KEY", "X-RapidAPI-Host": "finance-pulse.p.rapidapi.com"}
rows = requests.get(f"{API}/v2/symbols", headers=HEADERS,
params={"period": "d", "sort": "change", "min_mentions": 5,
"kind": "equity,etf"}).json()["data"] # tradable tickers only
candidates = [r for r in rows if r["sources"] >= 3] # one outlet repeating itself is not attention
for c in candidates[:8]:
print(f"{c['id']:<11} {c['change']:>+4} mentions {c['sources']:>3} outlets score {c['score']:+.2f} {c['name']}")
META +47 mentions 40 outlets score +0.34 Meta Platforms Inc.
NVDA +39 mentions 25 outlets score -0.02 NVIDIA Corporation
MSFT +37 mentions 42 outlets score +0.20 Microsoft Corporation
SBUX +21 mentions 16 outlets score -0.27 Starbucks Corporation
STLA +18 mentions 15 outlets score -0.30 Stellantis NV
V +16 mentions 13 outlets score +0.44 Visa Inc. Class A
005930.KS +14 mentions 15 outlets score -0.46 Samsung Electronics Co Ltd
MU +12 mentions 13 outlets score -0.33 Micron Technology Inc
The rates, indices and central banks are gone, and the outlet rule keeps out a name whose new mentions come from one or two outlets only.
The week's leaders
period=w and sort=mentions give the steady leaders rather than the jumps:
NVDA 442 FED 383 MSFT 320 BTC 286 OPENAI 283 IMF 272 RBI 254 TSM 246 GOOGL 213 AMZN 196 META 195 USD 194 005930.KS 175 SPACEX 175 SPX 174 AAPL 173 TSLA 163 UST 152 AMD 139 ANTHROPIC 135
Companies such as OPENAI and SPACEX are kind private: some of the week's most-discussed companies have no ticker,
which is why they are entities rather than missing.
Sectors
/v2/sectors returns the same statistics for the 11 GICS sectors, a cheap read of where the news flow is
moving:
sector mentions change score
InformationTechnology 1435 +701 +0.11
Financials 1346 +610 -0.07
Energy 792 +374 -0.11
ConsumerDiscretionary 753 +288 -0.15
HealthCare 232 +127 +0.19
RealEstate 231 +95 -0.04
Industrials 147 +64 +0.16
CommunicationServices 128 +46 +0.00
Utilities 81 +36 -0.14
Materials 115 +5 +0.01
ConsumerStaples 7 +1 +0.14
A sector that gains while another loses usually marks one story taking over the day's news.
Any interval, not only the last day or week
since and until replace period with an interval of your own, on both lists. The counts cover statements
published in [since, until), and change compares with the equally long interval before it (it is null when
that would start before your plan's history). Both are rounded to the hour, and the response echoes the interval used. Leave
until out for "up to now": the comparison is then with as many whole hours before as have passed since since:
curl -G "https://finance-pulse.p.rapidapi.com/v2/symbols" \
-d since=$(date -u -d '2 days ago' +%Y-%m-%dT00:00:00Z) -d until=$(date -u -d yesterday +%Y-%m-%dT00:00:00Z) \
-d min_mentions=1 -d limit=5000 \
-H "X-RapidAPI-Key: $RAPIDAPI_KEY" -H "X-RapidAPI-Host: finance-pulse.p.rapidapi.com"
One call per day, with min_mentions=1 and a limit above the day's symbols, gives a daily table of every symbol,
where /v2/series would need one call per symbol. The
response says "period": "custom" and echoes the interval. Like every filter on post time, it files a late article
under the day it was published, not the day it arrived.
One symbol's page
/v2/symbols/{code} returns everything about one entity in one call: who it is (kind, name, exchange,
country, aliases), its counts for the last 24 hours, 7 days and your plan's history under periods, the sectors of
its statements, its largest topics, the entities it shares topics with (related, empty for a symbol with few
mentions) and its newest statements with a cursor to continue. An alias answers as its code:
/v2/symbols/NVIDIA returns NVDA. It covers symbols with at least one statement within your plan's history.
Don't know the code? q= looks one up by name or code: /v2/symbols?q=micron finds MU.
Sentiment that is about this symbol only
A statement's sentiment belongs to the statement, and a statement can name several entities: "Nvidia fell as AMD
won the contract" is bearish for one and bullish for the other, but counts once, with one label, for both. So every
symbol row also carries mentions_single, bullish_single, bearish_single, neutral_single and score_single:
the same counts over the statements that name only this entity (countries named in passing are ignored). Most
statements do. When the two scores
disagree, trust the single one.
How much is covered
Coverage is concentrated on a small share of the symbols. /v2/meta says where it stands: symbols_d, symbols_w and symbols_window (symbols mentioned in each
period) and symbols_w_3_sources and symbols_w_10_mentions (this week's with three outlets, or ten mentions). A
score on a handful of mentions is not a signal; filter with min_mentions and look at sources.
Run this yourself: get an API key on RapidAPI ↗.