DEV Community

Pulsebit News Sentiment API
Pulsebit News Sentiment API

Posted on

How to Detect Business Sentiment Anomalies with the Pulsebit API (Python)

#python #api #datascience #nlp

How to Detect Business Sentiment Anomalies with the Pulsebit API (Python)

We recently encountered an intriguing anomaly: a 24-hour momentum spike of +0.136 in the business sentiment data. This spike stands out, especially when we consider the context of the surrounding news landscape. It raised questions about how our existing pipelines are handling multilingual data and the dominance of certain entities, particularly in the business sector.

![DATA UNAVAILABLE: lag_hours — verify /dataset/daily_dataset
[DATA UNAVAILABLE: lag_hours — verify /dataset/daily_dataset is returning sentiment_by_lang data for topic: business]

Your model might have missed this significant shift by several hours, potentially losing out on timely insights. This is especially critical when we consider that the dominant language in the articles is likely a mix of regional business vernacular, which can skew sentiment analysis if not adequately managed. For instance, if your pipeline only processes English, you may overlook critical developments from non-English speaking regions that could influence market sentiment.

To catch such anomalies, we can utilize our API for real-time sentiment analysis. Below is a concise Python snippet designed to identify this specific spike in business sentiment:

import requests

![Left: Python GET /news_semantic call for 'business'. Right: ](https://pub-c3309ec893c24fb9ae292f229e1688a6.r2.dev/figures/g3_code_output_split_1773183737033.png)
*Left: Python GET /news_semantic call for 'business'. Right: returned JSON response structure (clusters: 3). Source: Pulsebit /news_semantic.*


# Setting up our parameters
topic = 'business'
momentum = +0.136
score = +0.000
confidence = 0.000

# Geographic origin filter (example: filtering for English-speaking countries)
geo_filter = {
    "language": "en",
    "country": "US"
}

# Check for articles relevant to business
response = requests.get('https://api.pulsebit.lojenterprise.com/news_recent', params={
    'topic': topic,
    'geo_filter': geo_filter
})

data = response.json()

# Checking the momentum
if data['momentum_24h'] > momentum:
    print("Anomaly Detected: ", data)
Enter fullscreen mode Exit fullscreen mode

Keep in mind that geographic filtering is contingent upon having language and country data available in your dataset. In this example, we assumed that the relevant articles are primarily in English, allowing us to filter for insights directly from the U.S.

![DATA UNAVAILABLE: countries — verify /news_recent is return
[DATA UNAVAILABLE: countries — verify /news_recent is returning country/region values for topic: business]

Next, we can run the cluster reason string through our sentiment scoring endpoint to analyze the narrative framing itself:

# Analyzing the cluster narrative
narrative = "Clustered by shared themes: telkommetra, business, portfolio, admedika, group."

sentiment_response = requests.post('https://api.pulsebit.lojenterprise.com/sentiment', json={
    'text': narrative
})

sentiment_data = sentiment_response.json()
print("Meta-Sentiment Score: ", sentiment_data)
Enter fullscreen mode Exit fullscreen mode

This step ensures that we assess how the narrative surrounding this spike is being framed, providing us with a deeper understanding of underlying sentiment trends.

Now, let's explore three specific builds that can leverage this anomaly pattern:

  1. Geographic Anomaly Tracker: Build a signal that continuously monitors the momentum of business sentiment in real-time for non-English articles, specifically targeting regions in Africa. Set a threshold where momentum spikes over +0.1 trigger alerts.

  2. Meta-Sentiment Analyzer: Create an endpoint that routinely analyzes narratives related to key business clusters (like TelkomMetra and AdMedika) using the meta-sentiment loop. If the sentiment score falls below a certain threshold (e.g., +0.050), escalate it for further analysis.

  3. Gap Identification Tool: Develop a utility that identifies forming gaps in business sentiment between mainstream news and emerging narratives. This could utilize the forming themes such as "business" and "Africa" to highlight discrepancies, providing actionable insights for your trading strategies.

For further exploration, you can check out our documentation at pulsebit.lojenterprise.com/docs. The best part? You can copy-paste the code snippets provided above and run them in under 10 minutes to start detecting sentiment anomalies in your projects.

Top comments (0)