Leveraging NLP to Answer Subjective Questions

In today’s data-rich landscape, natural language processing (NLP) is redefining the boundaries of insight in business intelligence, helping organizations tackle questions that aren’t just quantitative but subjective, contextual, and sometimes ambiguous. By moving beyond strictly numeric metrics, companies can explore what lies beneath the surface of financial performance, customer sentiment, and market behaviors. In our recent webinar, “Leveraging NLP to Answer Subjective Questions,” experts from S&P Global delved into how NLP can bring value to subjective data inquiries. Here’s a recap of the conversation and some key insights on what’s possible today—and where this technology is heading.

Subjective questions often require nuanced responses that go beyond “yes” or “no” answers. In a business context, this could be questions like: “Is our brand meeting customer expectations?” or “How are investor sentiment trends evolving across sectors?” These questions, while abstract, are deeply meaningful to an organization’s strategic planning and alignment with market needs.

For companies today, subjective questions can validate the relevance of their mission, guide future directions, and even reveal hidden opportunities. As one of our speakers noted, asking subjective questions is a “rubber meets the road” moment for business intelligence. By doing so, companies can start to see a bigger picture and identify areas where previously overlooked insights may exist.

The importance of having diverse data

The quality of insights from NLP depends largely on the data feeding into it. As discussed in the webinar, data from a single source often doesn’t tell the full story. By combining multiple datasets—such as transcripts, reports, and external publications—organizations can deepen their understanding. For instance, S&P Global’s data ecosystem includes varied data points such as geographic and industry-specific tags, enabling analysts to slice and dice data according to different dimensions, such as sector or region.

One of our experts shared an example: by tagging information by entity, speaker, and industry in Q&A sections of corporate transcripts, we’re able to measure the emphasis placed on certain topics in investor calls. The goal is to capture and quantify discussions that may not directly show up in financial metrics but can impact market perception and company valuation. For example, how often and in what contexts do executives mention “sustainability” or “risk mitigation”? Such tagging allows us to turn unstructured data into measurable insights, helping companies map the weight of these subjective elements over time.

Insights at Scale and the Dangers of Too Much Data

While it’s tempting to collect data across every dimension possible, organizations must be cautious of data overload. The scope of NLP’s data intake can become unwieldy without a clear purpose. As our speakers suggested, businesses must balance depth with relevance. A broad dataset expands the “sandbox,” but there’s a risk of diverting focus onto data that isn’t truly valuable. By starting with core datasets, businesses can address their primary questions before expanding to secondary metrics, adding layers of complexity only where meaningful insights can result.

For example, S&P Global recently analyzed conversations around Environmental, Social, and Governance (ESG) factors by segmenting mentions across different geographies. This allowed analysts to pinpoint which regions and industries are actively addressing ESG, often revealing unexpected trends in countries or sectors that are typically less visible in global discussions. But this focused approach is key—too many metrics, and the clarity of insight is at risk.

Key Steps to Enable Subjective Questioning with NLP

1. Define Clear Objectives: Successful NLP analysis begins with a well-defined question. What exactly are you trying to uncover? Start with a clear question and stick with it throughout the process. While NLP can reveal multiple layers, sticking to the core question ensures the insights remain actionable.

2. Use Tagging for Granularity: One valuable method for analyzing subjective questions is tagging the data early in the process. In our approach, tagging data by location, sector, and speaker identity allowed us to filter through vast quantities of text without losing sight of core insights.

3. Incorporate Time-Based Analysis: Timing matters, especially in questions about sentiment and market perception. Observing changes in language around particular events can be illuminating. For instance, mapping shifts in corporate language about climate responsibility before and after the Paris Agreement revealed patterns in how companies communicate their commitments—and whether those commitments align with actions taken in later periods.

4. Monitor Changes Across Dimensions: A single dataset may suggest a trend, but cross-referencing it with complementary dimensions (e.g., revenue, geographical region, sector) allows for a more comprehensive view. For example, a CEO’s emphasis on sustainability in a Q&A section may correlate with the company’s increased ESG focus, but it’s the cross-analysis of revenue and alpha generation data that can validate this.

NLP as a Tool for Accountability

Using NLP, businesses can look beyond corporate messaging to gauge real accountability, helping to answer: Are companies following through on commitments made in previous years? This secondary level of analysis compares earlier communications with recent outcomes, creating a record that can demonstrate consistency—or a lack thereof—over time.

For example, if a company consistently promotes a sustainable message, NLP can help assess whether there’s a matching trend in actions taken by analyzing external data such as product announcements, geographic expansion, or shifts in regulatory compliance. This approach was demonstrated in our recent study on employee caregiving responsibilities, which examined how such commitments evolve over time.

NLP’s Evolution: From Text Mining to Insight Discovery

So, what’s next for NLP in answering subjective questions? As one speaker noted, models are getting “smarter and bigger,” but it’s more than just growth. Emerging trends point toward an “orchestrator model,” where multiple NLP models collaborate to complete a task. This framework could mean that one model might handle sentiment analysis, while another assesses the entity's financial language, each contributing unique insights to produce a fuller picture.

Moreover, human expertise remains essential, particularly when understanding subjective questions. While NLP can tag, count, and segment data, determining what’s meaningful still requires domain knowledge. The technology, while advanced, isn’t yet ready to replace this nuanced judgment but can certainly augment it, freeing analysts to focus on high-level questions and strategic recommendations.

NLP brings a disciplined yet flexible approach to answering subjective questions by balancing quantitative analysis with human oversight. If done thoughtfully, NLP allows businesses to better understand not only “what” is happening but “why” and “what it means” for the future.

Curious to learn more about the latest NLP techniques for answering subjective questions? Watch the full webinar, ‘AI in Action: Leveraging NLP to Answer Subjective Questions’, for a deeper dive into practical examples and techniques.