Contextual comparative analysis with Unique to This Filter
Identifying patterns and important concepts in unstructured text is key to understanding what people are really saying. Although there are valuable takeaways in entire text datasets, the insights you really need are those most relevant to the problem you’re trying to solve right now. These critical insights can usually be found within segments of your dataset, whether it’s in a specific region, for a particular set of buyers, or even for a new product you’ve just brought to market.
People’s responses to questions are heavily context-based, so it’s absolutely necessary for your text analytics application to give you the ability to understand key groupings and parse feelings distinct to those groups. This added depth makes all the difference in your ability to respond with personalized outreach, offerings, and services versus a one-size-fits-all approach. You need analytics that surface the most important insights without human, biased intervention . That’s why we want to highlight one of Luminoso Daylight’s most powerful features: Unique to this Filter.
Why filters are key to understanding open-ended text
Unique to this Filter is a capability in Luminoso Daylight’s analytics features. Using the Luminoso approach to natural language science, Unique to this Filter enhances your ability to combine open-ended text with important metadata, or closed-end data, to understand what issues are most relevant to the subset you’re examining. When you apply a filter to a dataset in Daylight to create a subset and select Unique to this Filter for the visualization, Daylight immediately updates to show concepts that are uniquely relevant to this group.
Unique to this Filter allows you to dig deeper into your dataset once it’s been analyzed by our leading approach to understanding language. This capability goes beyond surface-level insights, applying our science to identify which concepts are especially prevalent in subsets of your dataset. Prevalence – our word for those concepts that occur more often in your dataset than they generally would in the language of analysis as a whole – gives you the advantage of truly knowing what’s important in your analyses. For example, prevalence would make rarer concepts like “pandemic” stand out over words and concepts that are more common such as “computer” or “virus”.
Unique to this Filter works on datasets across any industry or domain to identify concepts that are more relevant to a subset than to a project as a whole. So regardless of the dataset you’re examining – from pharmaceutical notes, to online product reviews, to annual employee satisfaction surveys – there’s no need for training or coding. And without having to tell the application what to look for, you eliminate the human biases that muddy results.
How to use Unique to Filter in your analyses
There are as many ways to use Unique to this Filter as there are datasets. For starters, you could use it for:
Identifying demographic trends
Competitive analysis and comparing brands
Discovering concepts inside a group of high-scoring reviews
Comparing between gender or age groups
Finding differences between cities, regions, or countries
In some cases, you might be analyzing a dataset that covers unfamiliar content, making it harder to identify trends or concepts that are interesting and relevant. Whether you’re interested in age groups, gender, location, or any other filter, Luminoso automatically determines if there are concepts unique to the chosen segment. With AI helping you separate the signal from the noise, finding out what specific groups say about certain topics is easy.
Here’s an image from Daylight’s Volume feature, showing data from an anonymous employee survey at a major hospitality company. Employees answered the survey with short text responses and semi-structured data sources, displayed here as metadata filters like CEO approval, Date, and Star Rating. For this example, we’ll look at survey answers grouped under the “pros” and “cons” sections of the “Text Field” option:
No filter is selected and Unique to this Filter is not applied. The concepts you see are the top concepts by Volume in this project. Since this is an employee survey, these concepts make sense: “work,” “customers,” “compensation,” “management” and “benefits” occur at the highest volume, meaning they were discussed the most throughout the entire dataset. Looking at these concepts, we might be able to start identifying broad trends, but digging deeper through filtering will produce more insightful results.
In this next image, the “cons” filter is selected, and Unique to this Filter is applied. Notice how the top concepts by volume have changed drastically from the first image. Daylight has identified the concepts that came up most frequently within the response under the “cons” filter. Concepts like “low,” “low pay,” “difficult,” and “stressful” float to the top – painting a very different picture than the one we saw from the entire dataset. Simply applying Unique to this Filter immediately exposed illuminating and actionable trends.
Lastly, here’s the same dataset again, but with “pros” selected as the filter for Unique to this Filter instead. The concepts in this filter are radically different from both the overall volume view, and from the “cons” Unique to this Filter concepts. Similar to the last set, the “pros” filter, with Unique to this Filter applied, provides instant insight into concepts that are especially relevant to the “pros” category. Concepts like “Great benefits,” “good benefits,” and “Great perks” show things that employees appreciate.
Keep in mind that this example only shows Unique to this Filter in Volume. Using the same data in Drivers or Sentiment might reveal new information – that the company’s corporate values are associated with high review scores, or the perks that employees think about most positively – valuable information for recruiting materials and outreach campaigns for new talent.
This is just one example of the many uses for Unique to this Filter. Often, it’s tempting to focus on only overall trends in a project. But when combined with Daylight’s analysis features, Unique to this Filter offers an opportunity to find true, insightful differentiators– especially for analysts or business users that are new to the data they’re examining.
Besides being faster than a manual approach, Daylight’s AI removes human bias as it identifies powerful relationships in unstructured text sources. Since we as humans tend to look for the patterns we expect, it takes practice to remove your own biases when analyzing data … especially when you’re familiar with it. When you apply our AI to analyze text, these biases disappear, making it easy to keep your analysis impartial. With Unique to this Filter, understanding trends in filter groups and market segments is fast, easy, and powerful – and helps you discover way more than the things you were looking for.