How to use AI Function

If you can create your own functions, you can freely perform complex calculations that aren’t possible with existing tools like Exploratory. Furthermore, if you can build your own models, you can create complex functions for prediction, classification, scoring, and more.

Until now, these capabilities were exclusive to those with programming skills, but Exploratory v14 has broken down this barrier with the introduction of “AI Functions”!

With these “AI Functions,” anyone can easily create their own “functions” simply by using natural language.

But that’s not all. You can create functions tailored to your requirements using the world’s most advanced AI models, built on vast global datasets and cutting-edge AI algorithms.

AI Functions make the following tasks simple:

Generate company names, industries, and company size information from customer data
Automatically create optimized campaign emails for each customer based on purchase history
Categorize, perform sentiment analysis, and automatically label free-text survey responses

This means we’re entering an era where anyone can enhance the value of their data using “their own AI functions.”

Basic Usage

Select “Create AI Function” from the column header menu.

This will display the AI Function dialog.

You can create and execute new prompts, and if you find a good prompt, you can save it as a template.

In this note, we’ll introduce two use cases for AI Functions.

Sentiment Score

One common type of question in surveys is “free-text responses.”

For example, in the survey data below, there’s a column for free-text responses such as “Reasons for recommendation.”

Many Exploratory customers have asked, “Is it possible to calculate sentiment scores from free-text responses?”

This is where “AI Functions” come in handy.

Execute the following prompt with AI Function:

For each provided sentence, calculate a polarity score in the range of -1.0 to +1.0.

Scoring Criteria:

Extremely Positive (+0.8 ~ +1.0)
- Expressions showing strong satisfaction or praise
- Situations where problems are completely resolved
- Mentions of outstanding performance or quality

Positive (+0.4 ~ +0.7)
- Mentions of clear advantages or strengths
- Reports of good experiences or results
- Results exceeding expectations

Mildly Positive (+0.1 ~ +0.3)
- Small improvements or advantages
- Fulfillment of basic functions
- General sense of satisfaction

Neutral (-0.1 ~ +0.1)
- Statement of facts
- Explanations without emotion
- Simple situation descriptions

Mildly Negative (-0.3 ~ -0.1)
- Small inconveniences or challenges
- Minor issues
- Points with room for improvement

Negative (-0.7 ~ -0.4)
- Clear problems or dissatisfaction
- Disappointing results
- Lack of important functions

Extremely Negative (-1.0 ~ -0.8)
- Serious problems or malfunctions
- Strong dissatisfaction or negative emotions
- Major obstacles or hindrances

Factors to Consider in Assessment:
- Strength of expressions (degree adverbs like "very," "extremely," etc.)
- Degree of problem resolution (complete resolution, partial resolution, etc.)
- Relationship between expectations and results (above expectations, as expected, below expectations)
- Specificity and importance of advantages/disadvantages
- Overall context and intent

By executing this, we can obtain sentiment scores determined by AI for each row.

Looking at these sentiment scores, we can see that as the score approaches negative one (-1), strongly negative content becomes more prominent.

Conversely, as the sentiment score approaches positive one (+1), we can confirm that strongly positive content is included.

Using AI Functions, we can easily generate sentiment scores for free-text responses and confirm that appropriate scores are assigned based on context.

Text Grouping

We’ll continue targeting the free-text response column such as “Reason for recommendation.”

When you have such text, you may want to label and group each piece of text.

Previously, it was standard to use the “Topic Model” in the Analytics View to group texts, but there were cases where we wanted to label using predefined groups.

With AI Functions, “text grouping” can also be achieved.

In the prompt, specify the following. It’s possible to differentiate between “good support” and “Bad support” at the labeling stage.

Please assign labels to categorize the provided text into the following groups:

Good Support
Bad Support
High Product Rating
Low Product Rating
High Rating for Features
Low Rating for Features
Comparison with Competitor Services
Implementation-Related Content
Pricing-Related Content
Other

By executing this, we can create a text group column.

We can confirm that the texts are grouped according to the predefined labels.

The following shows sample rows from each group, confirming that the quality of support is also well differentiated.

Summary

The “AI Functions” added in Exploratory v14 enable users without programming skills to perform complex calculations and data analysis using AI. This feature facilitates information generation from customer data, automatic creation of campaign emails based on purchase history, categorization of survey free responses, sentiment analysis, and labeling.

Please try the AI Function feature that allows you to easily enhance your data value using AI.

Want to Experience It Now?

Please try these new features like AI Functions for yourself!

If you haven’t used Exploratory yet, please take advantage of our 30-day free trial!