TL;DR
This paper introduces a novel Chinese character input system that leverages linguistic affinities and a "friendly interface" to significantly reduce ambiguity and increase typing speed. By incorporating machine intelligence to handle common character pairings, the system aims to make Chinese character input as efficient and fatigue-free as possible for professional typists.
What Problem Does This Paper Address?
Inputting Chinese characters into computers has historically been a challenge due to the sheer number of unique characters and the ambiguity inherent in phonetic or shape-based encoding schemes. Traditional methods often require extensive memorization and manual disambiguation, leading to slow typing speeds and operator fatigue, especially for professional typists who spend hours at the keyboard daily. This paper addresses the need for a Chinese character input system that is easy to learn, fast (targeting 50 characters/minute), and minimizes fatigue, particularly for users with a junior middle school education.
Methods
The core of the Cornell system is an adaptation of Wang Yun-yu's four-corner numbering system for Chinese character input. This system encodes characters based on their peripheral shapes, assigning them to ten basic stroke shape classes, which are then mapped to single-digit values. The four-corner method reads these shapes from the corners of a character to generate a numerical identifier.
The Cornell team introduces two key refinements:
- Elimination of Null Zeros: Identifiers are shortened by removing leading "null zeros," resulting in variable-length codes (1-4 digits). This variation increases the number of identifier categories, reducing ambiguity.
- Linguistic Affinity and Machine Intelligence: The system capitalizes on the fact that about 60% of Chinese characters appear in two-syllable compounds or phrase words. By storing a dictionary of these common pairings, the machine can automatically retrieve the correct character when a sequence of two identifiers is entered, eliminating the need for manual disambiguation in many cases. This "machine intelligence" significantly reduces ambiguity, which is a major bottleneck in other systems.
The keyboard layout itself is also designed with human factors in mind, placing frequently used shapes on the middle row and associating keys with shape recognition rather than rote memorization.
Datasets
Not discussed in the paper. The paper mentions a set of 8,000 characters and the use of a stored dictionary of pairings, but does not specify the source or nature of these datasets.
Key Insights and Ideas
- Linguistic Structure as an Input Aid: The most significant insight is the successful application of linguistic structure (character pairings) to overcome the inherent ambiguity of character encoding. This moves beyond purely shape-based or phonetic methods.
- "Friendly Interface" for Reduced Cognitive Load: By designing a keyboard layout based on shape association and incorporating machine intelligence for disambiguation, the system aims to reduce the cognitive burden on the operator, making it easier to learn and use.
- Variable-Length Codes Enhance Efficiency: Eliminating null zeros and allowing variable-length identifiers not only reduces the number of keystrokes but also increases the distinctiveness of codes, further aiding disambiguation.
- Human Factors are Crucial for Professional Use: The paper emphasizes that for professional typists, ease of learning, speed, and fatigue reduction are paramount, driving the design choices for the keyboard and system.
Key Results
- Ease of Learning: Chinese test subjects required only about half an hour of instruction to learn the shape keyboard and disambiguation methods.
- Speed: After approximately 80 hours of practice, typists achieved speeds of over 40 characters/minute with very low error rates (approaching 0).
- Motivation: Test subjects maintained high motivation and willingness to work long hours over a year.
- Ambiguity Reduction: The system eliminates manual disambiguation in nineteen out of twenty cases by leveraging linguistic affinities.
- Projected Speeds: The authors anticipate mean typing speeds of 50-60 characters/minute with further development and training.
Limitations and Future Work
Limitations:
- The paper acknowledges that manual disambiguation is still required when ambiguous pairs are encountered, though this is less frequent due to the system's intelligence.
- The current system relies on a specific keyboard layout, and further exploration of alternative input methods like chord keyboards is suggested.
Future Work:
- Implementation of simplified character sets within the machine.
- Isolation of specialized vocabulary into specific sets for faster access.
- Continued field testing with Chinese operators.
- Enabling touch-tone telephones to act as remote terminals for the system.
- Developing various output applications such as printing, typesetting, and direct telegraph transmission.
- Studying the implementation of a chord keyboard.
Summary updated