Call Tokenization Engine for Audio Transcription Sentence Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech-to-text processing of telephone call audio files lacks reliable sentence tokenization due to the absence of punctuation marks and dependency on speaker pace, leading to inaccurate transcriptions.
Innovation Solution
An improved call tokenization engine is implemented to identify candidate expressions and expression pairs indicative of turn-taking and context changes, inserting punctuation markers to accurately tokenize sentences in transcriptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If punctuation marks are inferred from pauses or silence in audio, then sentence boundaries can be identified, but the accuracy deteriorates due to dependency on speaker pace and removal of pauses
Solution Approach 1:
The patent introduces an intermediary processing layer between speech-to-text transcription and final text output. This layer analyzes linguistic patterns, context, and semantic structures to insert punctuation marks without relying on audio pauses. The intermediary process uses rules-based algorithms to detect sentence boundaries through linguistic cues such as question marks, exclamation points, and contextual phrase patterns, thereby mediating between raw transcription and properly punctuated text.
2Ease of manufacture
If audio files have only one side of conversation available, then transcription can be performed, but sentence tokenization becomes inaccurate due to missing pause information
Solution Approach 1:
The system employs self-service mechanisms where the transcription process itself generates metadata about speaker turns and conversation structure. By analyzing the transcription text for patterns indicative of question-answer pairs, statement completions, and contextual cues, the system autonomously determines sentence boundaries without requiring additional audio input from both conversation participants. The algorithm serves itself by extracting structural information from the unilateral transcription data.
Data Source
AI summary
A system and method for tokenizing sentences in transcriptions of electronic communication audio files includes one or more electronic data storage systems coupled in electronic communication with an audio receiver. A collection of audio files is stored in the electronic data storage system(s). The audio receiver is configured to receive one or more audio files from the electronic data storage systems. The system also includes a speech-to-text processing engine configured to transcribe the one or more audio files received by the audio receiver to produce input transcriptions. In addition, the system includes a call tokenization engine that stores a list of candidate expressions and expression pairs that are indicative of turn taking and change of context in phone conversations. The call tokenization engine also receives input transcriptions of the one or more audio files from the speech-to-text processing engine, scans each input transcription for the one or more candidate expressions and expression pairs, inserts a punctuation marker adjacent a detected expression when the detected expression is one of the one or more candidate expressions and expression pairs, and outputs a tokenized transcription when the call tokenization engine reaches an end of the input transcription.