Transcription, Speech Analytics, AI

Call Transcription

Convert Speech into searchable text

Call Transcription (Speech-2-Text) converts a recorded telephone call into writing and stores it in searchable text format. CreaLog Transcription does this automatically and can provide the text in near real time. Helpful for converting voicemail-2-text or protocols of talks that have been recorded by our Mobile and Fixnet Recording or Enterprise Communication Recording.

This high performance transcription functionality is also a prerequisite for methodical call analysis to provide customer service management using an IVR with important findings on issues such as customer satisfaction and service quality. Once converted into text, a telephone conversation is easier to analyze, either manually or automatically. By contrast, it takes large amounts of time and processing power to search whole voice recordings for any specific subject. Supervisors can search for issues more easily, the high relevance of which is only recognized in hindsight – or was not even known beforehand.

Download our Speech (Voice) Analytics Data Sheet for more information.

Application domains for Transcription and Speech Analytics

To significantly increase customer and employee satisfaction and boost productivity, companies rely on analyzing customer conversations, evaluating the quality of dialogs and discussing them with service or contact center employees.

Modern speech technologies allow to automatically record, transcribe and evaluate telephone conversations. This way, Speech or Customer Analytics immediately delivers relevant information about the customer's concerns (intent), the dialog flow and the content of the conversation.

Separate audio tracks facilitate downstream analysis

In the first step, a high-quality transcript of the telephone conversation is created in real time and saved as text (speech-to-text). Both parties are recorded separately in order to simplify the subsequent automatic call analysis. This also allows further analysis of particularly interesting or problematic conversations.

The next step is an in-depth and AI-supported analysis (voice analytics/speech analytics/customer analytics) of the texts. This provides information about the structure, content, participants' conversational shares, and the flow of the conversations.

A Highlight — data protection by default

Anonymize critical data in records

Why are anonymization and data protection so important?

Technological advances in the automated transcription of telephone conversations have brought increased attention to the issue of data protection and data sovereignty: What exactly happens to the telephone calls in text form and the personal content?

The employee's view

When it comes to call recording and transcription, employees and their representatives are often rightly critical and skeptical. Much of the concern can be addressed with our technologies for securely anonymizing your own employees. Here are a few examples:

  • Employee names and other personal data are automatically recognized and immediately noised for anonymization in the call recording or replaced in the transcription by placeholders such as asterisks.
  • Individual digits or strings of digits (telephone numbers/extensions) can also be filtered and then anonymized or deleted from the recordings.
  • In smaller service units, colleagues can often be easily recognized by their voice. To rule this out, the voice can be distorted in the call recordings by changing the pitch and speed so that it is no longer possible to draw conclusions about the individual employee.

The caller's view

GDPR-compliant data processing is also enormously important for callers and increases customer confidence in the company. Here are some examples of which personal customer data can be identified and anonymized directly in the recording or transcription process:

  • Names, phone numbers, birth dates and addresses,
  • Social security numbers and health information,
  • Account balances or delinquencies,
  • passwords, PINs, and many other sensitive and private data.

Opportunities and benefits of Speech and Customer Analytics at a glance

Precise recognition of intentions

  • What exactly do customers say, what are their concerns and how do they express their intents?
  • How does the frequency of intents change over time over hours, days or weeks?
  • Are thresholds being exceeded on a particular issue that provide supervisors with important insights?

Identification of automation potentials

  • What is the percentage of routine calls and issues, and can these concerns be automated on a case-by-case basis by ServiceBots?
  • How can I best balance the mix of human and automated customer service?

Sustainable training, live coaching and further qualification

  • Are the conversational proportions between customer and agent right?
  • Is what I have learned being applied in customer conversations?
  • What can be learned from best and worst case conversations?
  • Where is coaching or follow-up training necessary?

Quality control and assurance

  • Did the conversation proceed according to company compliance guidelines?
  • How satisfied are customers with the conversation - were there any expressions of displeasure or did the customer thank you?
  • How can I minimize frustration for agents and customers alike?

Operational control - valuable insights for the management

What do customers think of your company? Are they satisfied with the goods or services or are there frequent complaints?

Transcription and analysis provide management with answers to these questions - for example:

  • Are advertising and marketing campaigns successful?
  • Do customers still have questions or need clarification? And what exactly are they about?
  • Are there certain main topics?
  • How do the main topics change over time (day, week, month)?
  • In what context are certain topics repeatedly addressed?
  • How do customer queries change with regard to the top topics? Can positive or negative trends be identified here?

This list can be expanded and extended almost at will, depending on the industry and company.

But the decisive factor is that the findings from the strategic analysis help to optimize the relevant business processes.

Live view and thematic radar

The CreaLog Cockpit provides initial results of the call analysis in the live view. Special phrase spotting is used to find words or phrases mentioned by the customer in the recorded dialogs and thus determine the most frequent reasons for calling.

The topic radar provides the ten most frequently given answers by callers to the question "What can we do for you?" and also visualizes these as hot topic statistics over the course of the day. It enables current trends to be quickly identified as deviations from the average as well as changes compared to previous periods.

Alarm when threshold values are exceeded

The success of current campaigns is immediately recognized if deviations from the comparative value are more than 10 and 15 percent.

Acceptance thresholds can be defined for selected and critical terms. If these are exceeded, an alarm is triggered by visualization on the desktop of the supervisor or service manager. If an alarm occurs, a context analysis can be used to immediately start determining the possible causes (root cause analysis).

Deep insights into call reasons and customer concerns

Part of the preparations for a system is building the domain-specific terminology for the new system. This is the only way to ensure that the recognition is optimally adapted to the customer's needs for the specific application purpose

To this end, the following questions, among others, should be answered:

  • How exactly does the customer contacting customer service formulate his concerns?
  • How does he say it, which terms or dialects occur frequently?
  • What tonalities or expressions are typical in the calls?

In the first step, the CreaLog Transcription Engine records hundreds of customer conversations, automatically converts them into text, and analyzes them using speech analytics.

Analysis of speaker characteristics and dialog structure

In addition to the actual content of a call, the structure of the dialog and characteristics of the speaker can also provide important insights.

Crealog speaker analysis therefore determines:

  • Age and gender,
  • emotions (mood analysis),
  • as well as the language of the speaker

The analysis provides information on how speech and silence are distributed within a call dialog:

  • Are there gaps in the conversation for a longer period of time?
  • Does one speaker interrupt the other?
  • Do the speaker speak at the same time?

Answering these questions also provides relevant information about the quality of customer service in the contact center, potential for improvement in process flows, for example to increase the first call resolution rate (FCR), and a possible need for training or education of individual employees.

For further information ...

... and a personal consultation

please contact us here!


General requests

+49 89 324656-0

Service Hotline (24/7)

+49 89 324656-10

Please make sure you have your credentials at hand.


Click here to subscribe to our quarterly newsletter!

Frankfurter Ring 211
80807 Munich

Please use our 
Google Maps Entry 
for localization.
Park comfortably behind our building on the customer parking lots marked with CreaLog.