Kanjoya’s natural language processing (NLP) delivers insight into human behavior and motivation

From surveys to reviews to CRM data, Kanjoya analyzes vast sets of unstructured data, uncovers meaning, and delivers insights that improve business results.

Emotion Recognition and EPS

Kanjoya’s proprietary technology goes beyond simple positive and negative sentiment to decipher emotions, human factors, and nuance in language.

Audience Categorization

We integrate emotion analytics with structured demographic information (e.g., location, age, and gender) to provide rich, high-signal guidance that improves business results.

Intelligent Baselines

We benchmark your data against the statistical profiles for different data sources, enabling you to identify ‘hot spots’, unusual activity, and key trends in your business.

Multi-dimensional Data

Built for massive scale, Kanjoya’s NLP technology automatically ingests vast sets of unstructured data, from Facebook and Twitter to surveys, search trends, and forum posts.

Kanjoya NLP technology is based on a decade of innovation and co-developed with leading linguistics researchers from Stanford University.

Moritz Sudhof, Senior NLP Engineer and Algorithmic Lead at Kanjoya, manages the development of Kanjoya’s sentiment, emotion, and experience modeling technology.

His team, in collaboration with researchers at Stanford University, is leveraging the rich corpus of emotion data from Experience Project to model human expression in language.

Listen to Moritz present at the Symposium this year to learn more about the world beyond sentiment analysis: Context-aware emotion modeling for customer insights.

Perception Whitepapers

Learn more about our technology

Sentiment Analysis vs. Emotion Analytics

Sentiment Analysis vs. Emotion Analytics Whitepaper

Perception Diversity Whitepaper

Frequently Asked Questions
1. Is Perception different than sentiment analysis?

Sentiment Analysis relies on a simplistic classification system that only allows information to be designated as positive, negative, or neutral. These simple classifiers are often insufficient for anything but the most basic phrases. In contrast, Kanjoya Perception utilizes advanced natural language processing to identify a full range of emotions in unstructured text. Perception is able to handle nuance within polarity, conflicting polarity, and promoter/detractor detection and allows customers to understand and predict human behavior with a much higher degree of accuracy than sentiment analysis products.

2. What types of data does Perception analyze?

Perception can handle any type of unstructured text as well as data integrated from multiple structured and unstructured sources. Common data types for Perception customers include open-ended survey responses; reviews from sites like Yelp, Trip Advisor, and mobile app stores; social sites, such as Facebook and Twitter; enterprise networks like Yammer; and text from help-desk systems.

3. What is the setup time for a new Perception system? How much training is required?

Perception includes a consumer-based UI to encourage a broad range of users across the enterprise. Training requires less than an hour, and the setup of new data streams generally requires less then 2 minutes.

4. How much data can Perception process?

Kanjoya Perception is built to handle vast amounts of data and can scale to handle the biggest enterprise deployments. Dozens of Fortune 500 companies already rely on Kanjoya to support their marketing, customer experience, and HR analytics requirements.

5. How much data is needed to have meaningful results?

Perception customers get the best results from datasets ranging from 500 rows to as high as five million social media messages.

6. Can Perception handle existing data or does it only work with new data once the system is deployed?
Perception can handle existing data, such as unstructured text from surveys or reviews, as well as new data. Perception handles data from almost any source and can even collect data for customers through Perception’s complete survey solution.
7. How long does it take to calculate results?
Results are calculated almost instantaneously. Perception often delivers results to customers in seconds or minutes for data volumes that used to take weeks or months to analyze.
8. Does Perception include spam filters?
Spam is context-dependent. What’s considered spam by one user (e.g., re-tweets of promotions)may be highly relevant to another customer. As a result, Perception handles spam by providing customers a way to filter across various dimensions, such as promoters and detractors in a customer experience context.
9. What is an EPS score? What is a good/bad score?
EPS stands for Emotional Promoter Score and behaves similarly to a Net Promoter Score (NPS).Perception customers can easily generate EPS by analyzing unstructured text to detect promoter and detractor statements. Unlike NPS, EPS does not require a special survey. It’s a basic feature of the Perception solution. Perception calculates EPS with the following formula:  (%Promoters -% Detractors). Similar to NPS, good and bad scores typically vary across industries. A rule of thumb for a good EPS is to compare the EPS to an industry’s baseline NPS.  
10. Is there a relationship between emotions and promoters/detractors?
Emotion models and promoter/detractor models are distinct models. Emotion models identify the emotions present in a statement, regardless of where the emotions are directed. Promoter/Detractor models identify whether the speaker is promoting or detracting from a particular company or product, for example. The following statement helps to clarify the distinction: “I had a horrible day, but the people at the Ritz made it better.” The speaker is upset(emotion) but is still making a promoter statement for Ritz. One of the reasons that Kanjoya’s promoter/detractor model is effective is because it is able to handle emotions as an input. Most analytical systems do not incorporate emotion, and as a result, provide less meaningful results.
11. How does Perception handle both unstructured and structured data?
When integrating structured and unstructured data, the structured data is used as filters that can be applied to the unstructured data. An example would be an employee pulse survey. Open-ended text responses (unstructured text) can be filtered or segmented with metadata describing the employees’ position in the company. In this case, metadata could include department, tenure, office location, and position.
12. What are the technology requirements to run Kanjoya Perception?
There is no installation or backend requirements in order to use Kanjoya Perception. An updated version of Firefox, Chrome, and Internet Explorer (IE 9 and above) with internet access is all that is needed.
13. Does Perception work with other platforms or products that are common in the enterprise, such as Medallia, enterprise social networks like Yammer, and in-house surveys?
Customers can import any CSV or Excel file using the Perception API. Perception can also import Medallia and other internal data sources through the Perception API. On the social side,customers can incorporate Twitter, Facebook, Twitter demographics, blogs and message boards,and app store reviews. Perception also includes integrations with Yammer and email and can distribute its own surveys and collect responses through our complete survey solution.
14. How is Perception different than other analytics and sentiment analysis solutions?
Perception relies upon an industry-leading natural language processing engine that is more accurate and better able to handle qualitative and emotional factors than its competitors. In addition, Perception not only handles structured and unstructured data, it can even perform the
data collection through Kanjoya’s complete survey solution.