sentiment parameter with a value ranging from -1.00 to 1.00, indicating how that chunk text might impact the market and whether it will have a positive or negative effect.
Sentiment Model
The Bigdata.com Sentiment Model is an advanced language model designed to quantify the market impact of financial and business news with precision and consistency. Using BERT-like transformer-based architecture, it captures contextual cues to evaluate how news content may influence market perception and investor behavior. The model produces a continuous sentiment impact score, ranging from -1, for strongly negative to 1, for strongly positive. This allows users to identify both the direction and intensity of the expected market reactions. The scoring is derived from a large-scale analysis of financially relevant headlines and articles, particularly those related to equities and corporate events. Unlike general-purpose sentiment models, the Bigdata.com Sentiment Model has been specifically trained on financial language, capturing the meaning behind complex expressions or market-specific phrasing. The training data spans from 2014 to 2023.Examples
| Chunk text | Sentiment score |
|---|---|
| Piper Sandler raised the firm’s price target on Microsoft to $475 from $435 and keeps an Overweight rating on the shares. The firm notes that better-than-feared results and acceleration in Azure contributed to a 7% after-hours relief rally in the shares. Azure growth surprisingly accelerated to 35% cc on a combination of non-AI outperformance and continued AI momentum. Piper says its forecast revisions on downstream impact to policy and tariff changes last week appear too aggressive. | 0.76 |
| Corporate earnings contributed to market optimism. Amazon (NASDAQ:AMZN) and Microsoft (NASDAQ:MSFT) exceeded Wall Street estimates, with Microsoft posting its best weekly performance in years. Despite these gains, concerns linger about the long-term economic impact of tariffs, as businesses face uncertainty and potential cost increases. | 0.4 |
| General Motors (GM) reported Q2 adjusted earnings Tuesday of $2.53 per diluted share, down from $3.06 a year earlier but above the FactSet consensus analyst estimate of $2.34. Second-quarter revenue was $47.12 billion, down from $47.97 billion a year earlier but above the FactSet consensus of $45.84 billion. For 2025, the automaker said it expects adjusted EPS of $8.25 to $10. Analysts expect $9.31. General Motors said it suffered a $1.1 billion impact from tariffs in Q2 and expects a worse impact in Q3. The company said it is taking steps to reduce at least 30% of the gross tariff impact in 2025 by making adjustments to manufacturing, targeting cost-reduction initiatives and consistent pricing. General Motors shares were down nearly 7%. | -0.7 |
| Treasury Secretary Scott Bessent said Tuesday that he will meet with his counterparts from China on Monday and Tuesday in Stockholm, according to media reports.\nIn company news, Microsoft (MSFT) said Tuesday it has observed two Chinese nation-state actors and another China-based threat actor exploiting vulnerabilities to attack on-premises SharePoint servers. The two Chinese nation-state actors were identified by Microsoft as Linen Typhoon and Violet Typhoon, while the other China-based actor was tracked as Storm-2603. The attacks, flagged by Microsoft on Friday, exploited a spoofing vulnerability and a remote code vulnerability, the company said. Microsoft shares were down 0.4% around midday. | -0.59 |
| Bessent says substantial probability of cuts expected after disappointing jobs report\nTreasury Secretary Scott Bessent said Thursday that financial markets are pricing in a high likelihood of the Federal Reserve cutting interest rates before the end of the year amid concerns about tariffs pushing inflation higher. | -0.59 |
| 5,800 IRS telework requests remain stuck in the queue \nBacklog is unintended consequence of Trump’s RTO orders\nThe US Treasury Department faces potentially steep penalties over a backlog of remote work requests sparked by President Donald Trump’s return-to-office push, internal records show. | -0.71 |
| Scott Bessent calls Democrat governor’s plan to withhold federal taxes ‘extremely reckless’ amid transgender athlete dispute\nTreasury Secretary Scott Bessent has accused California’s Gov. Gavin Newsom of threatening to commit tax evasion after he threatened to stop paying the “over $80 BILLION” in taxes the state pays to the federal government. | -0.69 |
| Yellen’s Commitment to the Global Energy Transition and Decades of Economic Leadership Will Add Value to Angeleno Group’s Future Investment Activities and Existing Portfolio\nAngeleno Group, the pioneering Los Angeles-based investment firm providing venture capital and growth equity to clean energy and climate solutions companies, announced today that former U.S. Treasury Secretary and Federal Reserve Chair Janet Yellen has joined the firm’s Board of Advisors. In this senior advisory role, Secretary Yellen will offer strategic guidance to the firm, providing her considerable economic and policy expertise to benefit investors and entrepreneurs committed to accelerating the global energy transition. | 0.71 |
Questions
If the text chunk mentions multiple companies, is the sentiment related to both of them?
If the text chunk mentions multiple companies, is the sentiment related to both of them?
What if the text chunks talk about a person or organization? What does the sentiment tell us about?
What if the text chunks talk about a person or organization? What does the sentiment tell us about?
The model was trained mostly to understand economically relevant events when generating sentiment scores, with a focus on equities, but, as for most transformer models pretrained with a large corpus of text, its language understanding goes well beyond that. This means that while there is no specific training for events around people, for example, it will understand the text and apply sentiment analysis based on the wording and context.
Does the sentiment logic take into account the entity provided in the filters to focus on the impact for those entities? Or it does not influence the sentiment value.
Does the sentiment logic take into account the entity provided in the filters to focus on the impact for those entities? Or it does not influence the sentiment value.
The sentiment model calculates and persists the chunk sentiment when Bigdata processes documents; therefore, its value is static and not conditioned by any search filter in the request.