nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Eikon Data API: Desktop Session Independence, News Discrepancies, and Sentiment Pipeline Methodology

SOM_IITB

LSEG Data Library / Eikon Data API: Desktop Session Independence, News Discrepancies, and Sentiment Pipeline Methodology
Hello Developer Community,
I am developing an automated news and sentiment extraction workflow and need clarification regarding session dependencies, API-to-terminal news alignment, and the underlying mechanics of the news sentiment engine.
1. Headless / Server-Side Authentication (Decoupling from Desktop Application)
My current workflow relies on local desktop session proxying. If the LSEG Workspace/Eikon desktop application is closed, API requests fail immediately, and running an active web session in a browser does not bridge the connection.

How can we authenticate and execute these API calls independently in a headless or cloud environment (e.g., AWS/Docker) without an active local desktop client?
What is the recommended transition path (e.g., using the LSEG Data Library for Python via Platform/RDP session authentication), and what specific credentials/entitlements are required for headless execution?

2. News Article Count Discrepancy (API vs. Workspace Terminal)
When querying the API for today's news stories for a specific instrument/company code, the API returns a significantly smaller result set than the Workspace terminal (e.g., 1 article returned via API versus 4 displayed in Workspace under the same entity search for the same date window).

Is this discrepancy caused by default source filtering, entitlement constraints on specific news wires via API, or indexing/caching latency between the desktop app and backend API endpoints?
What parameters or syntax must be passed to the API to guarantee parity with the standard Workspace News monitor?

3. News Sentiment Pipeline & Data Interpretation

Underlying Architecture: What model architecture powers the sentiment scoring engine (e.g., fine-tuned transformer/FinBERT variants, rule-based proprietary NLP, or another architecture)?
Missing Sentiment Data: Several returned articles contain no sentiment values. Is there an asynchronous processing delay post-ingestion before sentiment tags populate, or are third-party/non-English wires systematically excluded from sentiment analysis?
Neutral vs. Unclassified Logic: How can consumers programmatically distinguish between a genuine neutral score and an unclassified/unscored article?
Score vs. Confidence Metric: In several responses, the returned sentiment value and confidence metric display identical numbers. What is the explicit technical distinction and scale for each field?

Any code samples, API parameter recommendations, or documentation references would be greatly appreciated.

Find more posts tagged with

api

news

refinitiv-dataplatform-eikon

refinitiv-data-platform

#content

#technology

Accepted answers

All comments

Jirapongse

@SOM_IITB

Thank you for reaching out to us.

To retrieve data without running the Workspace application, you need to use the LSEG Data Library with a platform session. With a platform session, the library retrieves data directly from the LSEG Data Platform, which is available in the cloud. To obtain an LSEG Data Platform account, please contact your LSEG account team or sales representative directly.
As far as I know, this may be related to news licensing restrictions. Certain news content, such as Dow Jones News, may be filtered out or unavailable through the API due to licensing limitations. Could you please share the news query you are using to retrieve the data? I can then test it on my machine and verify whether the issue is related to licensing restrictions or the query itself.
Do you mean the news sentiment displayed in the News Monitor app? If so, please contact the Workspace Support team directly through LSEG Support for clarification. If not, could you please clarify the source of the sentiment data? This will help us better understand the issue and provide more accurate guidance.

SOM_IITB

Thank you for following up and providing additional context. Below are the details regarding each point:
1. Headless / Platform Session & App Key Entitlements
I do have access to the LSEG Data Platform; however, when attempting to generate new App Keys via the portal to enable EDP API access, the portal consistently throws a server error. I have contacted Workspace Support, but the issue remains unresolved on their side.
Because of this blocker, I am forced to fall back on an older key created in 2022. That legacy key only supports local desktop proxying and lacks EDP API entitlements, which prevents running headless platform sessions in cloud environments. If there is an alternate internal route or permission profile to renew key generation for EDP access, please let me know.

2. News Count Discrepancy (API vs. Workspace Terminal)
I have attached screenshots showing the disparity for the same instrument (APSE.NS / Adani Ports):

Workspace Entity News Monitor: Displays dozens of announcements and wire items for the selected day (including BSE regulatory filings, DJN, BUSSTA, FINEXP, etc.).
API Result: Returns only 3 articles across the entire day for APSE.NS (and this behavior is consistent across other tickers as well).

3. News Sentiment Pipeline & Fields (Score vs. Confidence)
Workspace Support recommended asking here on the Developer Community, as the underlying NLP implementation and data definitions fall under developer analytics.
When retrieving sentiment metadata via the API, the output exposes two numeric fields alongside the classification: Score and Confidence (e.g., Score: -0.32, Confidence: 32% or Score: +0.64, Confidence: 64%).
Could you clarify:

The distinction between Score and Confidence: In several instances (as shown in the second screenshot), the absolute value of the score directly mirrors the confidence metric (e.g., -0.32 to 32%, +0.64 to 64%). Is Score a signed polarity indicator (-1.0 to +1.0) while Confidence simply scales the polarity probability, or does Confidence represent entity-mapping certainty?
Methodology & Modeling: What NLP methodology powers this scoring (e.g., Machine Readable News / MarketPsych NLP models, a fine-tuned transformer/FinBERT architecture, or a dictionary/lexicon-based rule system)?
Neutral vs. Unscored: How should unanalyzed news items (such as tabular exchange filings) be distinguished programmatically from stories with an evaluated neutral sentiment (0.0 score with 50% confidence)?

Jirapongse

@SOM_IITB

You can use the App Key Generator to generate application keys.

The App Display Name must be unique.

Can you share the App Display Name that you are using?

2. I ran the following code and compared the news headlines with the news headlines from the news monitor app. They looks similar.

df = ld.news.get_headlines(
    query="R:APSE.NS",
    count = 50
)

Please share the code that you are using and the output.

3. Which API are you using to retrieve those sentiments? Please also share the code.

Jirapongse

@SOM_IITB

Are you using a university account?

If so, I'm not sure whether that account has access to the LSEG Data Platform. You may want to verify that the university account is licensed or authorized to use the service.

SOM_IITB

I'm using Eikon Data API as my older api key has this configured. I guess with EDP i might get proper full data as you are getting.

SOM_IITB

I'm not entirely sure about account type, but I'm currently encountering a server error whenever I try to generate new API keys. In fact, even when I attempt to generate the same Eikon Data API key that I created previously, it still throws a server error. Even if my account lacks EDP database access, I should at least be able to generate a key for Eikon, given that I already have access via my older API key.

Jirapongse

@SOM_IITB

For any issues related to the App Key Generator, please contact the Workspace Support team directly through LSEG Support.

Additionally, please provide a screenshot of the App Key Generator showing the error when it occurred. This information will help the support team investigate the issue more effectively.

EXPLORE OUR SITES