Sentence Type Widget

The Sentence Type Widget helps you understand how people communicate about your brand, not just what they say. By using Natural Language Processing (NLP), this widget classifies each post into a specific sentence intent or communication type, such as complaints, inquiries, promotions, or customer support responses.

What It Does

The widget analyzes conversations and automatically assigns a sentence type label to each post based on its underlying intent. It classifies content into 11 standardized categories, optimized for Bahasa Indonesia, Bahasa Malaysia, and English to ensure consistent and accurate analysis across markets.

It enables you to:

• Identify the intent behind conversations (e.g., complaint vs inquiry)
• Detect customer needs and expectations in real time
• Monitor risk signals such as suspicious activity or phishing
• Understand how your audience engages with your brand messaging
• Prioritize actions based on conversation urgency and type

Instead of manually reading thousands of posts/publications, you get a structured view of audience intent at scale.

How It Works

For each selected campaign, dataset, or filter:

1. The system collects posts from supported channels
2. NLP models analyze the sentence structure and keywords
3. Each post/publication is assigned a sentence type label based on predefined patterns
4. If multiple sentence types are detected, the system applies a priority logic to determine the dominant type
5. The results are aggregated and visualized in charts or tables

Priority Logic (When Multiple Types Exist)

To ensure consistent classification, the system prioritizes critical intents:

1. Complaint
2. Suspicious Activity
3. Phishing
4. Promotional Offer
5. Official Update
6. Appreciation
7. Encouragement
8. Customer Support Response

This ensures that high-risk or high-impact conversations are always surfaced first.

Key Metrics

• Buzz – The overall level of conversation activity associated with each topic.
  This metric helps you compare which topics generate the most attention within your campaign scope.
• Mentions – The total number of mentions contributing to each topic cluster.
  This view helps you understand topic scale and relative importance across channels.

Visualization Types

• Table View – Detailed breakdown for reporting and export
• Bar/Stacked Chart – Compare trends across time or campaigns

Sentence Type Categories (Updated)

The widget now uses a simplified and more accurate classification:

• Complaint – Negative feedback or dissatisfaction
• Inquiry – Questions or requests for information
• Customer Support Response – Replies from brands or support teams
• Promotional Offer – Promotions, giveaways, invitations
• Official Update – Announcements, greetings, informational updates
• Appreciation – Thanking or positive acknowledgment
• Encouragement – Motivational or supportive messages
• Recommendation – Suggestions or advice
• Tips – Helpful guidance or how-to content
• Phishing – Potential scam or misleading offers
• Suspicious Activity – Fraud indicators or unusual behavior

Channels Applicable

This widget supports:

• Social Media Channels
• Mainstream Media Outlets

Availability depends on your campaign setup and data coverage.

Use Cases

Customer Issue Detection

Quickly identify spikes in complaints to take immediate action.

Example: A sudden increase in “Complaint” posts may indicate product issues or service disruptions.

Customer Intent Understanding

Understand what your audience needs from your brand.

Example: High “Inquiry” volume suggests users are actively seeking product information.

Customer Experience Monitoring

Track how well your support team responds to users.

Example: Monitor the balance between “Customer Support Response” and “Complaint” to evaluate service effectiveness.

Campaign Effectiveness Analysis

Measure how your messaging is received.

Example: An increase in “Appreciation” or “Encouragement” indicates positive campaign reception.

Risk & Fraud Monitoring

Detect harmful or suspicious conversations early.

Example: Spikes in “Phishing” or “Suspicious Activity” may indicate brand impersonation or scams.