Documentation: News & Sentiment Data API

Overview

The Nextmark News & Sentiment API provides access to a 15-year archive of global financial news, processed by our proprietary NLP engine to quantify market narratives. Unlike raw news feeds, this dataset is structured, scored, and mapped to financial entities for immediate quantitative application.

Key Capabilities:

Source: Aggregated from 50+ global newswires, regulatory filings, and trade publications.
Enrichment: NLP-tagged for Sentiment (-1 to +1), Relevance, and Event Type.
Precision: Point-in-Time (PIT) timestamps to prevent look-ahead bias in backtesting.
Granularity: Entity-level resolution (Ticker, FIGI) distinguishing between "Subject" vs "Mention."

‍

1. Access Methods

Method	Use Case	Format
REST API	Searching for specific historical events ("NVDA Supply Chain 2023").	JSON
Bulk Feed	Ingesting full 15-year history for backtesting and model training.	JSONL
WebSocket	Live. Real-time streaming of scored news for algorithmic trading.	JSON Stream

‍

2. REST API: News & Sentiment Endpoint

Endpoint: GET https://api.nextmark.ai/v1/news/sentiment

Request Parameters:

Parameter	Type	Required	Description
api_token	string	Yes	Your API Key.
ticker	string	No	Filter by primary ticker (e.g., TSLA).
start_date	string	No	Filter news after this timestamp (ISO 8601).
min_relevance	float	No	Filter noise. 0.0 to 1.0 (Recommend > 0.7 for main subjects).
event_type	string	No	Filter by category: Guidance, M&A, Executive_Change.

‍

Example Request (Python)

import requests

url = "https://api.nextmark.ai/v1/news/sentiment"
params = {
    "api_token": "YOUR_KEY",
    "ticker": "TSLA",
    "event_type": "Guidance_Update",
    "min_relevance": 0.8
}

response = requests.get(url, params=params)
data = response.json()

‍

Request Structure (JSON)

{
  "article_id": "nm_news_8821",
  "timestamp_utc": "2024-02-10T14:30:00Z",
  "source_rank": 1,
  "headline": "Tesla cuts guidance on Model Y deliveries citing supply chain bottlenecks.",
  "novelty_score": 0.95,
  "entities": [
    {
      "ticker": "TSLA",
      "name": "Tesla Inc",
      "figi": "BBG000N9MNX3",
      "sentiment_score": -0.78,
      "relevance_score": 1.0,
      "event_type": "Guidance_Cut"
    },
    {
      "ticker": "NIO",
      "name": "NIO Inc",
      "sentiment_score": -0.20,
      "relevance_score": 0.3,
      "event_type": "Sector_Read_Across"
    }
  ]
}

‍

3. Data Dictionary (Key Metrics)

Field	Type	Description
sentiment_score	Float	Alpha Signal. Continuous score from -1.0 (Negative) to +1.0 (Positive). Derived from context, not just keywords.
relevance_score	Float	Noise Filter. Indicates if the ticker is the main subject (1.0) or a footnote (0.1).
event_type	String	Context. Auto-classification of the news topic (e.g., Labor_Strike, Earnings_Beat, Litigation).
novelty_score	Float	Uniqueness. Measures how "new" the story is compared to the last 24 hours. High score = Breaking News.

‍

4. NLP Enrichment & Context Highlighting

Because news is unstructured, we apply Transformer-based LLMs to structure the data before delivery.

Feature	Function	Use Case
Contextual Sentiment	Disambiguates "Good" vs "Bad" based on financial context.	"Revenue fell (Negative) but beat estimates (Positive)." -> Score: +0.4
Entity Resolution	Maps text to unique IDs (FIGI/Ticker).	Distinguishes between "Apple" the company ($AAPL) and "Apple" the fruit.
Event Classification	Tags the topic of the news event.	"Filter feed for all CEO_Resignation events in the S&P 500."

‍

5. Bulk Data Access (JSONL)

Download the full 15-year archive for offline analysis and backtesting.

File Naming Convention: Nextmark_News_Sentiment_{Year}_Q{Quarter}.jsonl
Format: JSON Lines. Each line is a full article object with nested entity scores.

How to Download:

Navigate to Data Export.
Select News & Sentiment Package.
Choose Year (e.g., "2020", "2023").
Click Download ZIP.

‍

6. Real-time WebSocket Feed (Live Trading)

Exclusive to Nextmark. This feed pushes scored news events with sub-millisecond latency for algorithmic execution.

Sample Stream Output (Decoded):

"timestamp": "14:30:05.001", "ticker": "META", "event": "Share_Buyback", "sentiment": 0.85, "headline": "Meta announces $50B expansion of share repurchase program."

‍

Use Case:

"Trigger a Buy order if Sentiment > 0.8 AND Novelty > 0.9."
"Halt trading on Ticker X if Sentiment drops below -0.5."

‍

7. Error Handling

401 Unauthorized: Invalid API Key or Subscription Tier.
422 Unprocessable Entity: Invalid ticker symbol or date format.
429 Too Many Requests: Rate limit exceeded (Standard: 100 req/sec).

‍

8. Need Help?

Quantitative Support: email quant-support@nextmark.ai
Developer Support: email dev-support@nextmark.ai

Author Name

Team Nextmark