[Paper Review] EmoWrite: A Sentiment Analysis-Based Thought to Text Conversion -- A Validation Study

Imran Raza, Syed Asad Hussain+5 more|arXiv (Cornell University)|Mar 3, 2021

Advanced Text Analysis Techniques35 references4 citations

TL;DR

EmoWrite proposes a novel brain-computer interface (BCI) that converts silent speech into text using a sentiment analysis-integrated recurrent neural network (RNN), dynamic circular keyboard, and real-time emotional state detection. It achieves 6.6 WPM and 90.36% accuracy with an 87.55 bits/min Information Transfer Rate for commands, demonstrating significant improvements in usability and emotional context integration for individuals with motor disabilities.

ABSTRACT

Objective- The objective of this study is to introduce EmoWrite, a novel brain-computer interface (BCI) system aimed at addressing the limitations of existing BCI-based systems. Specifically, the objective includes improving typing speed, accuracy, user convenience, emotional state capturing, and sentiment analysis within the context of BCI technology. Method- The method involves the development and implementation of EmoWrite, utilizing a user-centric Recurrent Neural Network (RNN) for thought-to-text conversion. The system incorporates visual feedback and introduces a dynamic keyboard with a contextually adaptive character appearance. Comprehensive evaluation and comparison against existing approaches are conducted, considering various metrics such as accuracy, typing speed, sentiment analysis, emotional state capturing, and user interface latency. The data required for this experiment was obtained from a total of 72 volunteers (40 male and 32 female) aged between 18 and 40 Results- EmoWrite achieves notable results, including a typing speed of 6.6 Words Per Minute (WPM) and 31.9 Characters Per Minute (CPM) with a high accuracy rate of 90.36%. It excels in capturing emotional states, with an Information Transfer Rate (ITR) of 87.55 bits/min for commands and 72.52 bits/min for letters, surpassing other systems. Additionally, it offers an intuitive user interface with low latency of 2.685 seconds. Conclusion- The introduction of EmoWrite represents a significant stride towards enhancing BCI usability and emotional integration. The findings suggest that EmoWrite holds promising potential for revolutionizing communication aids for individuals with motor disabilities.

Motivation & Objective

• To address the limitations of existing BCI systems in typing speed, accuracy, user convenience, and emotional state integration.
• To develop a user-centric thought-to-text conversion system that captures emotional states and sentiment during communication.
• To improve usability and reduce latency in BCI-based communication for individuals with paralysis or motor disabilities.
• To validate the effectiveness of a dynamic, contextually adaptive virtual keyboard with sentiment-enhanced word prediction.
• To establish a new benchmark in BCI performance through integration of emotional context and predictive text.

Proposed method

• The system employs a user-specific Recurrent Neural Network (RNN) to decode EEG signals into text, trained on individual typing and emotional patterns.
• A dynamic circular keyboard displays only relevant characters based on context and predicted sentiment, minimizing navigation time.
• Emotional state is captured via EEG signals and integrated into the RNN model to influence word prediction and character selection.
• The system uses visual feedback and real-time sentiment analysis to adapt the interface and improve user engagement.
• Information Transfer Rate (ITR) is calculated for both command and letter-level inputs to evaluate system efficiency.
• A comprehensive evaluation framework compares EmoWrite against existing BCI systems using metrics including accuracy, WPM, CPM, latency, and ITR.

Experimental results

Research questions

• RQ1 Can a sentiment-aware BCI system improve typing speed and accuracy compared to conventional thought-to-text interfaces?
• RQ2 To what extent does integrating emotional state detection enhance the predictive performance and user experience in BCI-based text entry?
• RQ3 How does the dynamic, contextually adaptive keyboard layout affect user latency and typing efficiency?
• RQ4 What is the Information Transfer Rate (ITR) of the proposed system for both command and letter-level inputs?
• RQ5 How does individualized RNN training based on EEG and emotional patterns affect system accuracy and personalization?

Key findings

• EmoWrite achieved a typing speed of 6.6 Words Per Minute (WPM) and 31.9 Characters Per Minute (CPM) with 90.36% accuracy in thought-to-text conversion.
• The system demonstrated an Information Transfer Rate (ITR) of 87.55 bits/min for commands and 72.52 bits/min for letters, indicating high communication efficiency.
• User interface latency was measured at 2.685 seconds, reflecting low system responsiveness delay.
• The integration of sentiment analysis significantly improved word prediction accuracy and contextual relevance, especially for emotionally charged content.
• The dynamic circular keyboard reduced navigation time and contributed to faster character selection compared to static layouts.
• The system showed strong personalization potential, with performance improving after user-specific RNN fine-tuning and emotional pattern learning.