Input Method Using Historical Dialogue Context for Word Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Pinyin input methods rely solely on the last on-screen entry for word prediction, leading to one-sided or unreasonable candidate word queues, and fail to provide suggestions when the input text box is blank.
Innovation Solution
An input method that acquires historical dialogue information, extracts characteristic keywords, determines prediction information, and uses this information to generate a more reasonable on-screen candidate word queue by searching associated candidate word banks, regardless of the presence of text in the input text box.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If word prediction is based only on the last on-screen entry, then the prediction process is simple and fast, but the candidate word queue becomes one-sided or unreasonable
Solution Approach 1:
The system performs preliminary analysis on historical dialogue information before prediction, extracting keywords and scenario feature tags in advance. This preparation work enables the system to generate more accurate candidate word queues without significantly increasing prediction time, as the historical data is pre-processed and stored for quick retrieval during prediction.
Solution Approach 2:
Scenario feature tags serve as intermediaries between historical dialogue information and candidate word selection. The system extracts scenario feature tags from historical dialogues and uses them to filter and rank candidate words, providing a bridge that connects past context with current prediction needs while maintaining prediction efficiency.
2Ease of operation
If the input text box is blank, then there is no text to analyze, but the system should still provide word predictions
Solution Approach 1:
The system performs preliminary analysis on historical dialogue information before prediction, extracting keywords and scenario feature tags in advance. This preparation work enables the system to generate more accurate candidate word queues without significantly increasing prediction time, as the historical data is pre-processed and stored for quick retrieval during prediction.
Solution Approach 2:
When the input text box is blank, the system copies relevant information from historical dialogue contexts to generate predictions. It extracts keywords and scenario feature tags from past conversations and uses them as substitutes for missing current input, allowing the system to provide predictions even without direct user input in the current context.
3Measurement precision
If historical dialogue information is used for prediction, then prediction accuracy improves, but system complexity increases
Solution Approach 1:
The system extracts only the most essential elements from historical dialogue information - specifically keywords and scenario feature tags - rather than processing the entire historical dialogue. This extraction approach maintains prediction accuracy by focusing on critical contextual elements while significantly reducing system complexity by ignoring redundant information.
Solution Approach 2:
The system applies different processing qualities to different parts of historical dialogue information. Keywords receive one level of analysis while scenario feature tags receive another, with each element processed according to its specific importance and characteristics. This localized quality approach optimizes prediction accuracy for each element type without uniformly increasing system complexity across all processing.
Data Source
AI summary
The present disclosure provides an input method, an apparatus and an electronic device. The input method comprises: acquiring historical dialog information about a dialog window; extracting characteristic information in the historical dialog information, wherein the characteristic information comprises a keyword; determining prediction information, wherein the prediction information comprises the keyword; based on the prediction information, acquiring an on-screen candidate word queue in an input text box in the dialog window; and outputting the on-screen candidate word queue.