Merlin Text Generation (Beta)

The Merlin Text Generation connector is a powerful AI-driven tool in Tray's automation platform, designed to make text generation and summarization capabilities simple to use within workflows.

Overview

Copy

The Merlin Text Generation connector is a powerful AI-driven tool in Tray's automation platform, designed to make text generation and summarization capabilities simple to use within workflows. It offers a suite of advanced natural language processing operations that can be seamlessly integrated into various automation scenarios.

Operations

Copy

Generate Text

Copy

The Generate Text operation in the Merlin functions connector allows users to create AI-generated text based on given instructions.

Input Parameters

1. Instructions: A prompt that guides the text generation. It should be a clear and concise description of what you want the generated text to cover.

2. Temperature: A value that controls the randomness of the text generation (0.0 to 1.0). Lower values make the output more deterministic, while higher values introduce more creativity and variability.

3. Max tokens: Limits the length of the generated text.

In the following example, the Webhook trigger receives a notification when a new feature is released. Using the AI connector (AWS Bedrock), a detailed description of the new feature is generated.

The Generate Text operation then uses this description to create a title for the new feature.

Classify Text

Copy

The Classify Text function categorizes a given piece of text into one or more predefined categories based on its content.

Parameters

1. Text: The input text to be classified.

2. Categories: The list of categories to classify into, if not predefined in the model.

3. Examples: is used for in-context learning, where the model is given a few examples to guide its classification process.

In the following example, the workflow iterates over records from a Google Sheet that contains feature titles and descriptions.

The descriptions are provided as input to the Merlin Functions connector to classify each feature into specific categories. A list of categories is supplied to aid the categorization process. For more precise classification, you can also provide examples.

Sentiment Analysis

Copy

Sentiment analysis, also known as opinion mining, is a powerful tool that has a wide range of applications across different industries and domains.

Its primary purpose is to understand the emotional tone behind a body of text, helping organizations and individuals make data-driven decisions.

The sentiment of the text is typically categorized as POSITIVE, NEGATIVE, or NEUTRAL.

In the following example, the workflow is triggered whenever a new customer support ticket is created.

The Merlin Functions connector employs the Sentiment Analysis operation to analyze the data received from the trigger and categorize the sentiments as positive, negative, or neutral.

As depicted in the image below, the step output shows the sentiment categorized as 'NEUTRAL' along with its detailed score.

This detailed score provides actionable insights, enabling effective decision-making in how to respond and utilize the data. For instance, based on a negative sentiment score, appropriate actions like escalating the ticket can be taken if necessary.

Summarize Text

Copy

The Summarize Text operation in the Merlin Functions connector allows you to generate concise summaries of provided text. You can choose from different formats to customize how the summary is presented.

Parameters

• Text (Required): The full text you want to summarize.

• Format (Required): Determines the style of the summary. Choose the format that best suits your needs for summarizing the text:

  • Call Summary: Produces a structured summary with sections: Action Items, Decisions, Unanswered Questions, and Notes.

  • Short Title: Generates a brief title summarizing the content in 4-7 words.

  • Single Sentence: Condenses the content into a single, concise sentence.

  • Bulleted Recap: Creates a bulleted list outlining the key ideas from the content.

• Temperature: Controls the creativity of the summarization. Range is 0 to 1 (Default is 0.5). Lower values make the summary more focused and deterministic, while higher values introduce more creativity and variability.

• Max Tokens: Limits the number of tokens (words or parts of words) in the summary.

In the following example, the workflow receives a transcript from a meeting.

Transcript:

1
{
2
    "text": "We had a productive meeting today discussing the upcoming product launch. Key topics included finalizing the marketing strategy, assigning roles to each team member, and setting deadlines for the creation of promotional materials. We also identified potential risks, such as delays in content production, and brainstormed solutions to mitigate these risks. Additionally, we reviewed feedback from the previous product launch to improve our approach this time around. The meeting concluded with a list of action items, including creating a detailed timeline, scheduling regular check-ins, and ensuring clear communication across all teams."
3
}

The Summarize Text operation using the Call Summary format summarizes the above transcript as follows:

1
{
2
    "summary": {
3
        "Action Items": [
4
            "Create a detailed timeline for the product launch",
5
            "Schedule regular check-ins and meetings",
6
            "Ensure clear communication across all teams involved",
7
            "Develop promotional materials according to assigned roles and deadlines",
8
            "Implement strategies to mitigate potential risks like delays in content production"
9
        ],
10
        "Decisions": [
11
            "Finalized the marketing strategy for the upcoming product launch",
12
            "Assigned specific roles and responsibilities to each team member",
13
            "Set deadlines for the creation of promotional materials",
14
            "Identified potential risks and brainstormed solutions"
15
        ],
16
        "Unanswered Questions": [
17
            "What specific feedback from the previous product launch will be incorporated?",
18
            "What are the detailed roles and responsibilities assigned to each team member?",
19
            "What are the exact deadlines set for the creation of promotional materials?"
20
        ],
21
        "Notes": [
22
            "The meeting focused on planning for an upcoming product launch",
23
            "Key topics included marketing strategy, team roles, deadlines, and risk mitigation"
24
        ]
25
    }
26
}
27