How to extract keywords from Intercom Support Call Transcripts using generative AI

Keyword Identification

Intercom

If you work in customer support, you know the value of listening to your customers. The conversations your support team has with customers are a gold mine of insights. However, manually combing through each support call transcript would be wildly impractical and inefficient. Fortunately, there is an easy and cost-effective way to identify important insights from your Intercom support call transcripts using generative AI. In this post, we’ll show you how to use generative AI to automatically extract keywords from Intercom support call transcripts.

What is Keyword Extraction?

Keyword extraction is a natural language processing (NLP) technique that involves identifying the most important or relevant words or phrases in a piece of text. You can use it to extract key information and themes from text has many applications, such as search engine optimization (SEO), content analysis, and topic modeling.

Keyword extraction can be performed manually, but it can also be automated using machine learning algorithms. These algorithms learn to recognize patterns and features in the text that are associated with important words or phrases, and can be trained on a labeled dataset of text.

You can use keyword extraction to analyze and summarize large amounts of text data to quickly identify the most important information and themes.

Example Use Cases

Use cases for extracting keywords from Intercom support call transcripts include:

Identifying common issues and trends
Improving response times
Personalizing customer interactions
Identifying areas for training and improvement

Teams that might find these use cases helpful include: customer support, customer success, product, marketing, and operations.

Accessing and Analyzing Data

To access and analyze Intercom support call transcripts, you’ll need to use the Intercom API. The API allows you to programmatically retrieve transcripts and other data from Intercom. You can then use generative AI to automatically extract keywords from the transcripts.

Here's an example of how to use the Intercom API:

curl https://api.intercom.io/conversations?per_page=50 \
-X GET \
-H 'Authorization: Bearer YOUR_ACCESS_TOKEN'

Once you have your data, you can use generative AI to analyze and extract keywords. One popular tool for generative AI is Google Cloud Natural Language API. With this tool, you can easily analyze text and extract keywords. It also provides sentiment analysis and entity recognition, which can provide even more insights into your data.

Identifying Preliminary Keywords

Before you start analyzing your data with generative AI, it can be helpful (but not necessary) to identify common keywords that you may want to extract from your support call transcripts. This can help you get started with your analysis and ensure that you’re focusing on the most important information.

To identify preliminary keywords, you can manually review a sample of your support call transcripts and identify the most common words or phrases. Alternatively, you can use a tool like Google AdWords Keyword Planner or Ubersuggest to identify keywords that are relevant to your industry or product.

Using Google Cloud Natural Language API

Here's an example of how to use Google Cloud Natural Language API:

curl "https://language.googleapis.com/v1/documents:analyzeEntities?key=YOUR_API_KEY" \
-X POST \
-H "Content-Type: application/json" \
--data-binary @request.json

With this tool, you can easily extract keywords from your Intercom support call transcripts. You can also perform sentiment analysis and entity recognition to gain even more insights into your data.

By following these steps, you can easily extract keywords from your Intercom support call transcripts and gain valuable insights into your customers’ needs and concerns.

Using AirOps to perform Keyword Identification

With AirOps, you can easily extract relevant keywords and phrases from your text-based data using the Keyword Identifier data app. Here's how:

Select "Keyword Identifier" from the Data Apps page. The input required for Keyword Identifier is the "text_field" which is the input text data.
Decide where you want the analysis to be performed and stored. The Keyword Identifier data app can be easily used in the AirOps Data App page and via API, but in this example, the analysis will be performed in Snowflake through an external function called AIROPS_KEYWORD_IDENTIFIER.

Here is an example SQL query:
```
SELECT
AIROPS_KEYWORD_IDENTIFIER(text_field) as result
FROM
your_table
```
Execute the keyword extraction analysis by running the SQL query. The output will contain an array of keywords and phrases extracted from the input text data.

Example Input:

"Hello, I am having trouble with my account. I cannot seem to log in and I have tried resetting my password multiple times."

Example Output:
```
"keywords": ["trouble", "account", "log in", "resetting", "password", "multiple times"],"summary": "A customer is having trouble logging into their account and has tried resetting their password multiple times."
```

Using AirOps to perform Sentiment Analysis

With AirOps, you can easily perform sentiment analysis on any text data such as reviews, support tickets, or sales calls using Sentiment Analyzer. Here’s how:

Select "Sentiment Analyzer" from the Data Apps page. The only input for Sentiment Analyzer is some text to analyze.
Decide where you want the analysis to be performed and stored. The Sentiment Analyzer data app can be easily used in the AirOps Data App page and via API, but in this example, the analysis will be performed in Snowflake through an external function called AIROPS_SENTIMENT_ANALYZER.

Here is an example SQL query:
```
SELECT
AIROPS_SENTIMENT_ANALYZER(text_field) as result
FROM
your_table
```
Execute the sentiment analysis by running the SQL query. The output will contain a sentiment score and sentiment summary, as well as a list of positive and negative keywords extracted from the input text data.
Input:
"I'm sorry to say that I had a terrible experience with your product. The customer service was unresponsive and the product didn't work as advertised."
Output:
```
"positive_keywords": [],"negative_keywords": ["terrible experience", "customer service", "unresponsive", "product", "didn't work", "advertised"],"score": -0.8,"sentiment": "Very Negative"
```

Using AirOps to perform Text Classification

With AirOps, you can easily perform classification using generative AI. Here’s how:

Select "Text Classifier'' from the Data Apps page. Below are the possible inputs for Text Classifier.text_field: The input text data.categories (optional): Categories can be specified as a comma-separated list. Leave empty for automatic determination.multi_category: Set to “true” if the text can belong to multiple categories, or “false” if it can only belong to one category.
Decide where you want the analysis to be performed and stored. The Text Classifier data app can be easily used in the AirOps Data App page and via API, but in this example, the analysis will be performed in Snowflake through an external function called AIROPS_CLASSIFIER.

Here is an example SQL query:
```
SELECT
AIROPS_CLASSIFIER(text_field, categories, multi_category) as result
FROM
your_table
```
Execute the classification analysis by running the SQL query. The output will contain a list of keywords extracted from the input text data that are relevant to the identified categories and a list of categories that the input text data belongs to based on the provided categories or automatic determination.

If you have feedback, please let us know!