Social Listening is a valuable tool for brands to gauge user perception on various topics. However, it requires a significant amount of configuration design and upkeep to properly group and categorizes mentions. Even with these efforts, there are limitations to automated techniques such as keyword-based classification, as they struggle to grasp the context in which certain terms are used, resulting in irrelevant data. This process can be time-consuming and burdensome, relying heavily on the skills of the implementer to differentiate the relevant mentions.

To address these challenges, Sprinklr has introduced an AI-based text classification tool called Text Classifier in AI Studio. This tool is designed for researchers, analysts, PR managers, and anyone dealing with brand reputation, making it easier to categorize and analyze data.

What is Text Classifier?

Text Classifier simplifies the process of categorizing messages that match a query or come from selected sources. It eliminates the need for implementing rules or maintaining keyword lists. By utilizing Text Classifier, irrelevant common terms can be filtered out, allowing for more precise analysis of data related to a brand, product, or industry.

Furthermore, Text Classifier is a valuable solution for those who wish to group mentions based on a specific customized objective without having to create a complex and maintenance-intensive configuration. It is a must-have tool for streamlining the process of message categorization and improving the accuracy of data analysis.

Types of Text Classifier models in AI Studio

When it comes to text classification, Sprinklr offers two options: Text Classifier (Single Select) and Text Classifier (Multi Select).

The main difference between the two is that the Text Classifier (Single Select) option only supports one classification per message, while the Text Classifier (Multi Select) option allows for more than one classification on each message.

• The Single Select option is ideal for use cases where only one classification is needed per message, such as determining whether a message is spam or not, or whether it is relevant or irrelevant. To use this option, the user would need to create a picklist custom field on a message level.

• Multi Select classifiers are better suited for cases where a single message can have multiple distinct classifications.

For example, messages related to sales leads for a product may include various categories including customer inquiries, information requests, sales pitches, and complaints, and one message could belong to two of these categories simultaneously, then the multi-select classifier would be the better option. To use this option, the user would need to create a multi-select picklist custom field on a message level.

In summary, while single select classifiers are suitable for use cases that require only one classification per message, multi-select classifiers provide a more granular level of classification when a single message can have multiple distinct classifications.

Use cases of Text Classifier

Following are some use-cases where Sprinklr's text classifier can be utilised in AI Studio –

1. Text Classifier uses contextual matching to accurately classify messages based on the query being searched. For example, if a brand wants to monitor mentions of their new product, Text Classifier can accurately identify and classify messages that discuss the product and its features.

2. Text Classifier eliminates the need to create and maintain keyword lists or rules, saving time and effort. For example, if a brand wants to monitor mentions of a new campaign, Text Classifier can automatically classify messages based on the context of the campaign.

3. Text Classifier eliminates the need for manual inspection of tagged irrelevant mentions due to ambiguous but irrelevant terms. For example, if a brand name has multiple meanings or connotations, Text Classifier can accurately classify messages based on the specific context of the message, eliminating the need for manual inspection. For example, the word ‘apple’ has multiple meanings, including the fruit, and the brand. Using this model, we can filter our brand and non-brand mentions.

4. Text Classifier allows customers to create their own training data for the AI models, ensuring that the AI is trained based on their specific needs. For example, a brand can use Text Classifier to train an AI model to accurately classify messages based on industry-specific terms or jargon.

Feature scope

• Sources – Sprinklr supports Text Classification in AI Studio for the following sources: Review Source, Topics, Topic Groups, Topic Tags, Themes, Theme Tags, Domain Lists, Domain List Tags, Keyword Lists, Channels, Account, Account Group, Message Type, Media Type, Post Type, Data Ingestion File Name, and Data Ingestion Import Tag.

• Languages – Sprinklr supports Text classification for 101 languages. Refer to the list of supported languages.

• Sprinklr supports the training of new Text Classification Models within the AI Studio interface itself.

• Sprinklr allows the validation of Text Classification predictions of deployed models, which can lead to additional model retraining for improved accuracy.

• The accuracy of Text Classification models can be calculated within the AI Studio interface with the help of Sprinklr.

List of languages supported in Text Classifier