Industry

How Leading Customer Research Platforms Leverage ASR, NLP, and NLU Tools

Let's examine the top three ways that ASR, NLP, and NLU APIs serve as the standout tools comprising the best Customer Research Platforms today.

How Leading Customer Research Platforms Leverage ASR, NLP, and NLU Tools

Table of contents

Today’s connected world means companies always have an abundance of customer feedback and opinions to tap into. But how do companies effectively sift through all of this data to turn it into intelligent, actionable insights?

Customer Research Platforms, also referred to as Voice of Customer Analysis Platforms and UX Analysis Platforms, may be the answer.

Using the power of Artificial Intelligence (AI) and Machine Learning (ML), Customer Research Platforms analyze customer qualitative and quantitative audio, video, and text-based feedback to find commonalities, trends, insights, and more.

One of the most significant sets of these AI- and ML-backed tools include cutting-edge Automatic Speech Recognition (ASR), Natural Language Processing (NLP), and Natural Language Understanding (NLU) research applications.

This article examines how these ASR, NLP, and NLU tools can be leveraged to help Customer Research Platforms:

  • Transcribe asynchronous and live voice and video feedback to make review and analysis more efficient.
  • Generate key themes and highlights to speed up research analysis.
  • Produce digestible insights that can be easily categorized, tagged, and searched.

First, we’ll look at exactly what a Customer Research Platform is before diving into the top three ways that ASR, NLP, and NLU APIs serve as the standout tools comprising the best Customer Research Platforms today.

What are Customer Research Platforms?

At their core, Customer Research Platforms serve as huge customer knowledge databases and command centers.

Customer Research Platforms aggregate video, audio, and text-based customer feedback, housing the data in one collective and searchable platform. Then, the platform uses AI- and human-based analysis tools to identify patterns and themes in the responses and research to generate insights that can be turned into specific, results-driven actions.

For example, one Customer Research Platform has helped inform healthcare marketing by comparing respondents’ descriptions of pharmaceuticals to brand messaging.

Other platforms use thematic clustering to analyze qualitative data from open-ended survey questions to generate action items for companies.

Because much of the customer feedback consists of voice and video responses, Customer Research Platforms are investing in ASR and NLU/NLP tools to optimize these services.

Speech-to-Text and Audio Intelligence for Customer Research Platforms

Thanks to significant advances in AI, ML, and Deep Learning (DL) research, today’s Automatic Speech Recognition, or ASR systems are the most accurate, affordable, and accessible they’ve ever been. Top Speech-to-Text APIs, for example, can transcribe audio and video stream data, like those used by Customer Research Platforms, at near-perfect accuracy for both asynchronous and real-time transcription.

Then, NLP/NLU tools can help Customer Research Platforms quickly build high ROI features and applications on top of the transcribed survey data. For example, these tools, referred to as Audio Intelligence APIs, can automatically generate key highlights from survey responses, analyze customer sentiment, categorize and tag conversations, and more.

Let’s dive deeper into the three biggest impacts Speech-to-Text and Audio Intelligence technology can have on Customer Research Platforms:

1. Create High Utility Transcripts from Survey Responses

Using ASR technology to transcribe asynchronous and livestream voice and video feedback makes the customer data review and analysis process significantly more efficient. Transcription accuracy, of course, is extremely important, but not the only factor to consider. Instead, the best Speech-to-Text APIs prioritize transcript readability as well.  

What does transcript readability mean? Many APIs will output a transcription that, while accurate, lacks basic punctuation and casing, paragraph structure, and speaker labels, making it difficult to read.

See this transcription below as an example:

But how did you guys first meet and how do you guys know each other? I  
actually met her not too long ago. I met her, I think last year in  
December, during pre season, we were both practicing at Carson a lot. 
And then we kind of met through other players. And then I saw her a few 
her last few torments this year, and we would just practice together 
sometimes, and she's really, really nice. I obviously already knew who 
she was because she was so good. Right. So. And I looked up to and I met 
her. I already knew who she was, but that was cool for me. And then I 
watch her play her last few events, and then I'm actually doing an 
exhibition for her charity next month. I think super cool. Yeah. I'm 
excited to be a part of that. Yeah. Well, we'll definitely highly
promote that. Vania and I are both together on the Diversity and 
Inclusion committee for the USDA, so I'm sure she'll tell me all about 
that. And we're really excited to have you as a part of that tournament. 
So thank you so much. And you have had an exciting year so far. My 
goodness. Within your first WTI 1000 doubles tournament, the Italian 
Open.Congrats to that. That's huge. Thank you.

To fix this problem, top Speech-to-Text APIs offer Automatic Casing and Punctuation Models and Paragraph Detection features that automatically format transcription texts with appropriate paragraph structure, sentence structure, capitalization, and punctuation, making them much more readable. Some also offer Speaker Diarization APIs that automatically detect and label multiple speakers in an audio or video stream.

Here’s what a transcript with these features included looks like:

<Speaker A> But how did you guys first meet and how do you guys know each 
other?
<Speaker B> I actually met her not too long ago. I met her, I think last 
year in December, during pre season, we were both practicing at Carson a 
lot. And then we kind of met through other players. And then I saw her a 
few her last few torments this year, and we would just practice together 
sometimes, and she's really, really nice. I obviously already knew who
she was because she was so good.
<Speaker A> Right. So.
<Speaker B> And I looked up to and I met her. I already knew who she 
was, but that was cool for me. And then I watch her play her last few 
events, and then I'm actually doing an exhibition for her charity next 
month.
<Speaker A> I think super cool.
<Speaker B> Yeah. I'm excited to be a part of that.
<Speaker A> Yeah. Well, we'll definitely highly promote that. Vania and 
I are both together on the Diversity and Inclusion committee for the 
USDA. So I'm sure she'll tell me all about that. And we're really 
excited to have you as a part of that tournament. So thank you so much. 
And you have had an exciting year so far. My goodness. Within your first 
WTI 1000 doubles tournament, the Italian Open. Congrats to that. That's 
huge.
<Speaker B> Thank you.

See how much easier that is to read?

Also consider if companies using the platform will need to publish user research feedback, externally or internally, without surfacing exactly who gave the feedback. Or maybe the survey data could potentially contain private or sensitive information, like medical data or phone numbers or addresses.

If this is the case, PII Redaction APIs can automatically remove Personally Identifiable Information (PII) from a transcription text, replacing any desired redacted items with a # instead.

This table shows the types of PII that can commonly be redacted:

2. Generate Key Highlights and Analysis

In addition to providing highly accurate, readable transcripts, Customer Research Platforms must help users make sense of these responses. One component of this process is automatically generating highlights and key analysis from the transcription data. Or the platform may create a highlight reel of the most important and/or sensitive segments of customer data.

There are a few Audio Intelligence APIs that work together to achieve this sophisticated analysis. The first is an Auto Chapters, or Text Summarization, API. Text Summarization APIs create “summaries over time” on top of transcription data from the audio or video stream. This process works in two parts: (1) Segmenting the audio or video stream into logical chapters, or where the conversation naturally changes topic and (2) Producing a single sentence headline and multi-sentence summary for each of the determined chapters.

Now, end users can have a summary at a glance for each survey result, significantly cutting down the review process.

Second, Entity Detection, or Named Entity Recognition, APIs identify and classify important information in the transcription text. For example, San Francisco is an entity that is classified as a location.

With Entity Detection, Customer Research Platforms can help end users identify commonly recurring entities, such as company names, locations, occupations, etc., and compile them for further analysis.

Other common entities that can be detected and classified with an Entity Detection API include:

Topic Detection APIs can work in conjunction with Entity Detection APIs as well. Topic Detection identifies and labels topics in the transcription text, also helping end users identify patterns in customer responses. Topics are classified using the IAB Taxonomy, a compilation of 698 commonly identifiable topics:

Finally, Sentiment Analysis APIs provide strategic insights by accurately labeling speech segments in a transcription text as positive, negative, or neutral. Customer Research Platforms can use Sentiment Analysis to help end users track and analyze customer feedback about products, services, companies, and more–both in individual survey responses and across collected responses as a whole.

When used together, Audio Intelligence APIs like Auto Chapters, Entity Detection, Topic Detection, and Sentiment Analysis can help Customer Research Platforms provide true insight into customer behaviors and attitudes.

3. Categorize, Tag, and Search Responses

Customer Research Platforms also need to help end users categorize, tag, and search qualitative and quantitative research data. This could include providing searchable topics and categories, similar to the use of hashtags on Twitter, or even automatically trimming and editing model responses for each searchable tag.

These “auto” or “smart” tags can be powerful filters that help end users sift through the Customer Research Platform and locate the content that really matters to inform actions or goals.

The Audio Intelligence features discussed above–Auto Chapters, Entity Detection, and Topic Detection–work to make this intelligent feature a reality. Entity Detection and Topic Detection, for example, can identify and label the common themes in the survey data. The Customer Research Platform can then use this informed data to generate smart tags that represent key bits of information an end user might want to search for. Then, Auto Chapters can provide a contextually based summary of important segments of the survey response that can be attached to that tag.

Intelligent Customer Research Platforms

Customer Research Platforms must leverage State-of-the-Art AI, ML, and DL technologies to offer competitive services to end users. Speech-to-Text and Audio Intelligence APIs help Customer Research Platforms build out these intelligent features by creating high utility transcripts, generating key highlights and analysis, and facilitating categorizing, tagging, and searching of response content.

Because these ASR and NLP/NLU tools are powered by cutting-edge, evolving research fields, Customer Research Platforms can be confident that their services offered are future-proofed for years to come.