PathosAI

PathosAI is an in-house solution used for Text and Sentiment Analysis. Using PathosAI you can analyze your multi-line comment responses and understand the customers’ opinion and sentiment behind every keyword in the response. The tool performs following tasks:

• Categorize by theme and topic
• Implicit and Explicit Sentiment detection - at an overall level as well as by topic and theme

It is designed to help our clients gain insights into their customers’ feedback, understand what is working and what is not working, and where they need to prioritize actions. These insights can help them improve specific features of their products, services and quality of interactions with their customers across touchpoints.

Customers have a general tendency to openly share what they truly feel when asked to respond to open-ended questions. Therefore, our ability to analyze this data, identify implicit sentiments and emotions, and do it through a contextually aware AI model is a critical value addition to our clients.

Another key element of the product is our ability to perform analysis based on the industry sector. What is important to customers when dealing with one product/service, may not be as important to them when dealing with a different industry. For e.g. safety would be most important for airline customers, but not so much for customers of a fashion retailer.

We have built the dashboard answering the ‘so what’ question as we go through the dashboard from top to bottom.

PathosAI supports multilingual text analysis. Currently we support English, Spanish, and German. Other languages will be added shortly.

• Go to: Login >> Survey >> Analytics.
• Click on the Text Analysis icon and select PathosAI from the drop-down menu.

Setting up sentiment industry to categorize extracted topics into domain-based themes.

To set up sentiment industry:

Click on the pencil icon, select industry from the dropdown and click on update industry.

The industry update can occur a maximum of three times.

PathosAI automatically selects the first open-ended question in the survey sequence as the data analysis source. You have the flexibility to choose an alternative question from the question source drop-down menu, which exclusively displays open-ended questions.

• Click on Segment
• A pop-up appears where you can create and apply filters.

The "Breakdown" feature serves as a secondary filter, enabling a more detailed examination of the displayed data. By choosing a breakdown question, you can explore various facets of the responses and conduct a more comprehensive analysis in conjunction with the selected source open-ended question.

PathosAI dashboard displays

• Theme and Topic filter
• Overall sentiment score
• Sentiment score by theme
• Response sentiment distribution
• Details

Theme is specific to the industry while topics are sub-categories under the selected theme.

The overall sentiment score is presented on a scale from 1 to 5. The sentiment bar is distributed into 6 sentiments Very Negative, Negative, Mixed, Positive, Very Positive and Neutral Sentiments.

The PathosAI model analyzes customer expressions to unearth the ‘intensity’ of sentiment. Intensity of sentiment expressed by customers show how important the issue is to them as well as the extent to which it is likely to affect their future decisions. We measure and report sentiments on the following range: very positive-----positive—-neutral----mixed----negative-----very negative The categorization between the Very Positive to Positive and Very Negative to Negative is done based on the language used by the customer, number of issues or positive areas, willingness to recommend, intention to spend, likelihood to switch, level of satisfaction. Based on this, the model predicts the correct sentiment. Very Positive sentiments refer to superior experience and potentially higher levels of loyalty and intention to spend, while Very Negative sentiment means higher levels of negativity, which could mean higher retention and revenue at risk. This categorization will help our clients prioritize the most critical areas to focus on first.

Within these charts, there are multiple themes available, yet the top 10 themes are exhibited alongside their associated sentiment scores. When the breakdown feature is enabled, it will provide sentiment scores categorized by themes for the chosen responses. For instance, if you choose responses related to "How often do you travel via airline ?" The chart will exhibit sentiment scores grouped by themes for the selected responses. When you click on any of the bars in the charts, it will prompt a pop-up window to appear, presenting all comments and details regarding the current selection. When you click on any of the bars in the charts, it will prompt a pop-up window to appear, presenting all comments and details regarding the current selection.

It presents the overall response count per theme. If you opt for positive sentiment, it will showcase the total number of positive responses for each theme, broken down by the selected answer options.

This provides a comprehensive view of all themes and topics, including their response counts, scores, and sentiment distribution.

Topics are grouped under the theme.