Word-Segment Language Model for Verbatim Text Input Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text input systems autocorrect verbatim input that does not correspond to known words in the language model, leading to a frustrating user experience and limiting predictions to only known words, thus failing to offer verbatim input as a prediction candidate.
Innovation Solution
A system utilizing a word-segment language model that predicts and validates word-segments, allowing for the generation of predictions that include verbatim input by segmenting words into parts that can be combined to form valid words, and using a multi-language model to combine predictions from both word-segment and word language models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system autocorrects verbatim input by replacing it with alternative word suggestions, then the system can maintain prediction accuracy for known words, but the user experience becomes frustrating and the system loses the ability to offer verbatim input as a prediction candidate
Solution Approach 1:
The system dynamically adjusts its prediction behavior based on whether the verbatim input matches known words. When input matches known words, the system provides autocorrection suggestions; when input is unknown, the system preserves verbatim input as a prediction candidate, allowing flexible adaptation to different input scenarios
Solution Approach 2:
The system changes the parameter of prediction candidate selection based on the nature of input. For known words, it selects from alternative suggestions; for unknown words, it includes verbatim input in the candidate set, thereby adjusting system behavior to match input characteristics
2Reliability
If the system limits predictions to only known words from the language model vocabulary, then the system maintains reliable prediction accuracy, but it cannot offer verbatim input as a prediction candidate when that input is not a known word
Solution Approach 1:
The system segments the prediction candidate set into two categories: known words from the language model and verbatim unknown words. This segmentation allows the system to maintain reliability for known words while adding adaptability for unknown words by treating them as separate prediction candidates
Solution Approach 2:
The prediction system serves multiple functions: it continues to provide accurate predictions for known words while simultaneously supporting unknown verbatim input as valid prediction candidates, thereby achieving multi-functionality in handling both familiar and novel vocabulary
3Device complexity
If the system uses a static language model with fixed vocabulary, then the system structure remains simple and manageable, but the system cannot adapt to new words or expand beyond training data vocabulary
Solution Approach 1:
The system transitions from a static vocabulary approach to a dynamic one where the prediction candidate set is not fixed. Unknown verbatim words are dynamically added to the candidate set during runtime, allowing the system to adapt to new vocabulary without requiring complex retraining or vocabulary updates
Data Source
AI summary
Systems for inputting text into an electronic device are provided. The systems are configured to receive a sequence of characters input into the device. The systems comprise a means configured to generate from the sequence of characters a sequence of word-segments. The systems also comprise a text prediction engine comprising a language model having stored sequences of word-segments. The text prediction engine is configured to receive the sequence of word-segments. In a first embodiment, the text prediction engine is configured to determine whether each word-segment of the sequence of word-segments corresponds to a stored word-segment of the language model and output the sequence of word-segments as a candidate prediction when each of the word-segments of the sequence of word-segments corresponds to a stored word-segment of the language model, regardless of whether the sequence of word-segments corresponds to a stored sequence of word-segments. Various other systems and corresponding methods are provided. A system is provided comprising a word-segment language model comprising stored sequences of word segments and a candidate filter. The candidate filter is used to filter word predictions generated by the word-segment language model.