Browse Prior Art Database

Presenting Predicted Emotional Impact upon Pending Inputs in Online Chat

IP.com Disclosure Number: IPCOM000247910D
Publication Date: 2016-Oct-11
Document File: 4 page(s) / 210K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is an online chat interface and underlying system that presents predicted emotional impact upon a pending text input. The interface provides the sender a brief opportunity to review the potential impact the text has on the receiver’s emotions.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 4

Presenting Predicted Emotional Impact upon Pending Inputs in Online Chat

Online chats are widely used for both work- and personal-purpose remote communication. Compared to face-to-face communication, online chats do not usually convey facial expression, nuances, accents, and intonations. Often, the absence of such rich context leads the receiver to develop an unintended emotion (e.g., anger). Sometimes, the sender is careless with the text, causing unintended emotions for the receiver. Such unintended emotions can cause obstacles for successful online communication, lead to a serious misunderstanding, and even jeopardize work or personal relationships.

So far, the best practice to avoid problems is for the sender to always read over the text before sending it. Customer representatives are trained to not use particular phrases that might negatively affect the customer's emotion. However, it is not a well-followed practice, especially in busy communication. People often make mistakes, which require much effort

to restore a good relationship.

In addition, some people have clinical conditions that cause difficulties in predicting others' emotions and being sympathetic. As online chats allow very limited channels to convey emotions, these people experience elevated difficulties in predicting the impact on the receiver's emotion.

The novel solution is an online chat interface and underlying system that presents predicted emotional impact upon a pending text input. The interface provides the sender a brief opportunity to review the potential impact the text has on the receiver's emotions.

1


Page 02 of 4

Figure 1: Example interface with predictions for the reader's emotional response

The system is implemented into two major stages: (1) model training and (2) real-time emotional impact estimation.

Figure 2: Model Training architecture
(TFIDF: Term Frequency Inverse Document Frequency NLC: Natural Language Classifier)

2


Page 03 of 4

As shown in Figure 2, the system feeds a given corpus of chat logs to (1) the emotion classifier and (2) the feature vectorizer. The syste...