Scenario-Aware Audio Transcription With Personalized AI Summaries

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing speech transcription products lack the ability to provide highly personalized and accurate summaries tailored to specific scenarios, occupying user storage space and failing to meet diverse user needs.

Innovation Solution

A method and apparatus for training a language model through supervised fine-tuning using scenario-specific samples and prompts, integrated with a cloud storage solution to transcribe and summarize audio within a storage service application, ensuring personalized and accurate summaries without local storage occupation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If existing speech transcription products use general-purpose models, then they can serve multiple scenarios, but they cannot provide highly personalized and accurate summaries tailored to specific scenarios

Engineering Contradiction:
Improvesummarization accuracyVSAvoidscenario adaptability
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent applies local quality by creating scenario-specific language models tailored to different transcription needs (meetings, lectures, interviews, etc.). Each scenario receives a customized model trained on relevant data, ensuring high accuracy for that specific context while maintaining overall system versatility through the modular architecture that supports multiple specialized models.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the general transcription system into scenario-specific sub-models. Instead of using one general-purpose model for all scenarios, the system divides functionality into multiple specialized language models, each optimized for particular scenarios like meetings, lectures, or interviews, thereby achieving both personalization and accuracy.

Inventive Principle:
Principle #1Segmentation

2Speed

If audio transcription and summarization are performed locally, then processing speed is fast, but user storage space is occupied

Engineering Contradiction:
Improvetranscription speedVSAvoidstorage space
Core Design Contradiction:
SpeedVSVolume of stationary object

Solution Approach 1:

The patent introduces cloud-based processing as an intermediary between the user device and the transcription service. The client application uploads audio files to cloud servers for transcription and summarization processing, then retrieves results. This mediator approach enables fast processing through powerful cloud infrastructure without occupying user device storage space.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transitions the processing location from the local device dimension to the cloud dimension. By moving computation and storage to a different spatial dimension (cloud infrastructure), the system achieves high-speed processing capabilities while keeping user devices lightweight and storage-free.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Volume of stationary object

If a cloud storage solution is used, then user storage pressure is reduced, but integration of full-link closed loop from audio production to summarization is required

Engineering Contradiction:
Improvestorage pressureVSAvoidsystem integration complexity
Core Design Contradiction:
Volume of stationary objectVSDevice complexity

Solution Approach 1:

The patent merges multiple previously separate functions (audio recording, cloud storage, speech-to-text transcription, and AI summarization) into a single integrated cloud-based system. By combining these functions into one unified service, the system reduces user storage pressure while managing integration complexity through a cohesive architecture where all components work together seamlessly.

Inventive Principle:
Principle #5Merging (Combining)

4Measurement precision

If scenario-specific language models are trained, then summarization accuracy improves, but training data processing time increases

Engineering Contradiction:
Improvesummarization accuracyVSAvoidmodel training time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-training scenario-specific language models during system setup or offline periods using relevant training data for each scenario. This advance preparation ensures high summarization accuracy is ready when needed, without causing delays during actual transcription operations. The models are trained beforehand and stored for rapid deployment.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4664450A1Method and apparatus for transcribing audio
Publication Date: 2025.12.17 BEIJING BAIDU NETCOM SCI & TECH CO LTD
  • EP4664450A1 patent drawingFigure 1
  • EP4664450A1 patent drawingFigure 2
  • EP4664450A1 patent drawingFigure 3(a)~3(d)

AI summary

The present disclosure provides a method and apparatus for transcribing audio, relates to the field of artificial intelligence technology, in particular to the fields of natural language processing and deep learning technology. A specific embodiment of the method includes: receiving audio information uploaded through a scenario entry of a storage service application installed on a client; determining, based on the scenario entry, a scenario type of the audio information; performing speech recognition on the audio information to obtain text information corresponding to the audio information; and inputting the text information and a prompt corresponding to the scenario type into a language model to obtain summary information, where the language model is obtained by performing supervised fine-tuning on a pre-trained model using samples corresponding to various scenario types, and the prompts corresponding to the various scenario types are obtained by tuning initial prompts corresponding to the various scenario types using the language model. An output effect of the language model of this embodiment is more in line with needs of users in various scenario types.