Deep Learning Framework for Transcript Search Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional solutions for searching text transcripts of customer support interactions are ineffective due to the presence of acronyms, shorthand notation, varying word choices, lack of grammatical structure, and punctuation, making it difficult to identify specific customer experiences.
Innovation Solution
A deep learning framework is employed to process text transcripts, utilizing a phrase model for key phrase extraction, a topic model for topic classification, and a sentiment model to compute sentiment scores. These models work together to identify suboptimal customer experiences by analyzing sentiment and session duration, and returning transcripts with total scores exceeding a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional search solutions are used to search text transcripts, then the system is simple and easy to operate, but the search effectiveness deteriorates due to acronyms, shorthand notation, and varying word choices
Solution Approach 1:
The patent introduces an intermediary processing layer consisting of multiple models (phrase model, topic model, sentiment model) that mediate between the raw transcript data and the search query. This intermediary layer standardizes the unstructured transcript data into structured representations, enabling conventional search algorithms to effectively process the data despite the presence of acronyms, shorthand, and varying terminology.
Solution Approach 2:
The patent transforms the parameters of the text data by extracting key phrases, identifying topics, and computing sentiment scores. These transformations change the representation of the transcript from raw text to structured features that can be effectively searched and compared, thereby improving search effectiveness without requiring fundamental changes to the search system architecture.
2Measurement precision
If transcripts are processed to handle acronyms and shorthand notation, then search accuracy improves, but processing time increases
Solution Approach 1:
The patent performs preliminary processing of the transcript data by extracting key phrases, identifying topics, and computing sentiment scores before the actual search operation. This preliminary action prepares the data in advance, transforming it into a structured format that can be quickly searched and compared, thereby reducing the time required during the search process itself while maintaining high accuracy.
Solution Approach 2:
The patent segments the transcript processing into distinct functional components: phrase extraction, topic identification, and sentiment analysis. Each component processes specific aspects of the text independently, allowing for efficient parallel processing and reducing the overall time required to analyze the transcript while maintaining comprehensive search accuracy.
3Measurement precision
If deep learning models are applied to analyze customer experiences, then identification accuracy improves, but computational resources required increase
Solution Approach 1:
The patent divides the complex task of customer experience analysis into separate deep learning models, each responsible for a specific function: phrase model for key phrase extraction, topic model for topic identification, and sentiment model for sentiment analysis. This segmentation allows each model to be optimized for its specific task, improving overall identification accuracy while enabling more efficient resource allocation compared to a single monolithic model.
Solution Approach 2:
The patent applies deep learning models selectively rather than universally. Instead of processing every aspect of every transcript with full computational intensity, the system uses different levels of processing depth appropriate to each analysis stage and data characteristic, thereby achieving high identification accuracy for critical aspects while reducing overall computational resource consumption.
Data Source
AI summary
Learning frameworks for processing text transcripts may include receiving, by an application, a query comprising a topic. A topic model may determine a plurality of subtopics based on the topic. The application may receive, from a database based on the topic and plurality of subtopics, a plurality of text transcripts. A sentiment model may compute, for each text transcript, a respective sentiment score based on a text of the respective text transcript. The application may determine, for each text transcript, a duration of a communication session associated with the respective text transcript. The application may compute, for each text transcript, a total score based on the sentiment score and the duration of the respective text transcript. The application may return, as responsive to the query, a subset of the plurality of text transcripts having a total score that exceeds a threshold.