STT, TA, FAQs - Solutions for Customer Service
Generating and Executing STT/TA/FAQs using SQL
Introducing the core technologies of ThanoSQL's VoC (Voice of Customer) analysis service: STT, TA, and FAQs.
You can view examples of VoC analysis using LLM-based natural language processing at the link below.
– VoC Analysis App
– Interactive Analysis Case – 01 Customer Service Interaction Analysis
Like the above case, even users without IT knowledge of queries or databases can easily and quickly analyze data using ThanoSQL's interactive analysis service in natural language.
However, in this post, let us explore how to analyze VoC using the Query Manager option in ThansoSQL's workspace. Users can intuitively analyze data using the Query Manager instead of natural language-based interactive analysis using relational database schemas and SQL queries.
The image below indicates that ThanoSQL includes STT, TA, and FAQ Gen modules for the VoC analysis. In customer service, call records of customers and customer service representatives get uploaded to ThanoSQL in voice file form, and the file will be transformed into a text file if one runs STT. The text file can be use for various analysis in the TA module. Also, repeated questions and answers can be collected, and FAQs can be autonomously created.
Before researching each STT/TA/FAQ, let us introduce ThanoSQL's Query Manager.
ThanoSQL's Query Manager is the tool that adds the targeting data in the ThanoSQL database, runs SQL queries, and analyzes its results simultaneously.
More detailed explanations on Query Manager can be found in the link below.
– Query Manager
Returning to the main topic, let us examine ThanoSQL's STT/TA/FAQs.
1. STT (Speech To Text)
STT, or speech-to-text, literally converts spoken audio into text. It is challenging to analyze the human voice to analyze customer service information; the voice must be transformed into text first. The image below is the data flow and format of ThanoSQL's STT.
In customer service, every voice recording is stored in cloud storage, like Amazon S3 or the organization's local storage. Upload mp3 or WAV format audio files from the storage to ThanoSQL STT's database. Then, they will be transformed into text files through Open AI or Whisper-like AI models. The transferred files are in CSV or XLS format.
2. TA (Text Analysis)
TA is a function that can summarize context using AI models on transformed text, analyze customer emotions by extracting keywords and determining negativity or positivity from the text, or categorize it based on the topic.
The below image is the initial starting screen of ThanoSQL Query Manager.
Press the + button next to my_data to add text data for the analysis.
Upload text file (excel or CSV format) created as STT's output on ThanoSQL.
This image shows the process of inserting the content of the uploaded text file into a table in the ThanoSQL database.
Since data is inserted in the database's table, the data can be utilized freely via API in other locations.
The above image is the screen that enters query statements, sorting out each conversational topic. The Mistral 7B model is being used in the example.
The screen displays the 12 categories that query results on running. The execution result also gets stored as a table in the database, enabling the data to be called using API from outside and utilized for analysis.
3. FAQ Generation
It is the function of generating FAQs by collecting responses to repeated or frequently asked questions based on the analysis of the whole context of the text transformed in STT.
The above image displays typing queries and making FAQs based on the call context. The GPT-4o model has been used in the example.
The FAQ screen has been formed, resulting in query execution. Similarly, the generated FAQ is stored in a database table, allowing the data to be analyzed via API access.
Distinctness of ThanoSQL
Suppose an AI model or algorithm other than pre-built in ThanoSQL is needed. In that case, you can easily integrate external models and verify the results.
The above case is checking the result after applying ML needed for ThanoSQL by registering the recent model of Hugging Face on ThanoSQL's FAQ result or STT result table, storing out and making queries based on the positivity and negativity in context, and then counting the result after placing in the table.
Expected Benefit - Utilization in AICC, CS Industry
Lastly, the table below organizes how AI-based STT and TA technology are used and their effects.
As shown in the table above, utilizing AI-based STT/TA/FAQ can significantly enhance customer service efficiency and service quality. Customer service representatives can avoid simple and repetitive work and focus on more complex task-solving. Customers can receive fast and precise services.
Furthermore, organizations can provide a better customer experience by providing custom services and understanding customers' requirements earlier through data analysis. In addition, not only can the customer service representative's education and management be enhanced, but the whole service task can also become efficient. SmartMind's AI technology will bring innovative changes to the future of customer services, providing a better environment for both customers and customer service representatives.
EN 
KR