A Voice Calculator Based On Speech Error Correction: Literature Review By: Imran Ali & Hizbullah

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 13

A Voice Calculator based on Speech

Error Correction

Literature Review by : Imran Ali & Hizbullah


Abstract
 Speech Recognition Post-
processing
 Automatic Correction
 Natural Language
Processing
 Compiler Design
Introduction
 Problem statement
Work flow Diagram
System Structure and Resources
• voice recognition and correction module
• expression recognition module
• expression calculation module
Processing of Speech Information

Error types of Speech Recognition


Build the Semantic Framework

The semantic framework describes user-supplied data, which is


a set of semantic features, with each one representing specific
semantic information and corresponding to a context-free
grammar rule. The system extracts specific information from
the recognized expression and fills it in the semantic features
with the application of the grammar rules.
Build the Semantic Framework
For example, applying the following grammar rules to the incorrect recognized text, "three thousand two
hand and eighty-four"
Correction of feature word recognition

For errors like tone or dialect


system uses a unique localization correction method, namely,
"Pinyin correction mechanism"
Recognition and Calculation of Expressions

Recognition of Expression
• Exchange semantic model
• Bracket Semantic Model
• potential Ignore semantic model
Recognition and Calculation of Expressions

Analysis and calculation of expressions


For the recognized expression, the system
analyzes it as the conventional compiler.
Here, we use recursive descent parsing
algorithm to get the parsing tree of the
expression, and the calculation of the
expression use the synthesized attributes
of the grammatical structures in the
parsing tree.
Conclusion

 For the shortcomings of conventional handheld calculator, we


develop a voice calculator using
automatic error correction techniques which supports voice input
and output. Currently, the system
is tested by experiments and has reached the desired goals. And
through user error correction, its
correction library can be further enriched. The system is fully
applicable to any user.
References

You might also like