Google Workspace Sentiment Analysis: How It Works

Sentiment analysis uses natural language processing (NLP) technology to interpret employee emotions and opinions through gathered data and employee feedback to gauge the overall mood and culture within an organization. Traditionally, HR departments have depended on surveys to understand the emotional and cultural temperature of their organizations. Modern-day businesses have embraced employee listening strategies in technologies like Google Workspace, which provide much more timely and multi-faceted data for business leadership to get a better snapshot of the pulse of their workforce.

Contents

• Understanding sentiment analysis and how companies can use it
• What are the different categories of sentiment analysis?
• How sentiment analysis works
• Benefits of analyzing sentiment data in Google Workspace
• What Google Workspace data should be part of your data analysis?
• Limitations to using Google Workspace add-ons for sentiment analysis
• Challenges of sentiment analysis
• How does Aware enable sentiment analysis on Google Workspace?

Understanding sentiment analysis and how companies can use it

NLP and AI/machine learning models can analyze collaboration messages at scale in real-time to evaluate the overall sentiment of an organization’s workforce. The technology extracts common topics, themes, and trends from thousands of messages and evaluates them for positivity, negativity, and toxicity. These messages come from company-owned collaboration tools and public channels to give business leaders insights into what subjects concern their employees and how they feel about them.

Companies can also pinpoint risks to the organization’s data, including insider threats, phishing scams, and malware delivery emails. Infosec leaders can protect intellectual property by setting alerts on keywords and topics to ensure valuable data and company property remain confidential. Leadership can surface negative employee sentiment in times of challenge or change and adjust the company messaging to assuage doubt, stamp out rumors, and prevent the spread of gossip.

Aware + Deloitte: Your guide to harnessing collaboration sentiment data while protecting employee privacy.

Get My Copy

What are the different categories of sentiment analysis?

There are three main categories of sentiment analysis: automated, rule-based, and hybrid.

Automated sentiment analysis

In automated sentiment analysis, machine learning algorithms classify the text. The automation algorithms are trained on large datasets to recognize patterns and predict sentiment. Key features include:

• Machine learning (ML) models, such as support vector machines, neural networks, and deep learning models, are trained on labeled datasets to learn positive, negative, and neutral characteristics.
• Natural language processing (NLP) uses tokenization, part-of-speech tagging, and syntactic parsing to preprocess text and extract features fed into ML models.
• Scalability. Automated artificial intelligence systems can process significant amounts of text quickly, so they’re suitable for ingesting messages at scale, such as the volume generated by collaboration tools like Google Workspace. These systems can then analyze the messages and return employee sentiment insights company leaders can use to understand the pulse of their workforce and monitor for risks.

Rule-based sentiment analysis

For rule-based sentiment analysis, a set of manually crafted rules determines the sentiment in the text analytics. These rules are based on linguistic heuristics and predetermined lexicons of positive and negative words. Key aspects include:

• Lexicons: Dictionaries of words or phrases annotated with their associated sentiment. For example, “happy” is positive, and “angry” is negative.
• Heuristics: Rules that account for linguistic constructs to modify sentiment, such as negations, intensifiers, and other descriptors. For example, “not happy” returns a negative sentiment.
• Pattern matching: Patterns in text can also indicate sentiment, such as certain adverbs and adjectives. Language is fluid, and the descriptors surrounding emotive words can change the meaning depending on context.

Rule-based systems are relatively straightforward with their interpretations, but they can be rigid and may not capture the fluidity of language well. New or varied contexts can sometimes confuse a rule-based system if no rule covers the language used.

Hybrid sentiment analysis

Hybrid sentiment analysis combines elements of automated and rule-based systems to leverage the strengths of each. This system typically involves:

• Integration of machine learning (ML) and rules: Using ML models to handle the bulk of sentiment classification while refining the results through a rule-based filter to improve the outcome.
• Contextual understanding: Employing rules to account for specific linguistic phenomena that ML models may miss, like sarcasm, abbreviations, slang, or industry-specific jargon.
• Flexibility and accuracy: Achieving a balance between scalability, automated methods, and the precision of rule-based systems is one way to address the fluidity of language.

Hybrid approaches intend to improve accuracy by shoring up the gaps in the other two systems alone.

What's in your digital collaboration data? Get your free, customized report to find out.

Claim Now

How sentiment analysis works

Transforming raw text data into meaningful insights about the emotional tone and context of the messages requires steps taken in various stages.

Clean and prep the data, removing irrelevant information

Data cleaning and preprocessing are critical to getting the data into its purest plain text form.

1. Removing HTML tags, special characters, and irrelevant symbols.
2. Converting text to lowercase to normalize words.
3. Tokenization, which breaks text into individual words or phrases.
4. Removing stop words (“the” or “and”) that carry no sentiment.
5. Correcting spelling errors and accounting for slang, abbreviations, and jargon.
6. Stemming, or lemmatization, to reduce words to their root form, (e.g., congrats, congratulations).

Purifying the dataset reduces noise and sets the text within up for more accurate text analysis.

Natural language processing (NLP) for understanding speech

NLP enables ML models to understand and interpret language by breaking text into components. It’s a bit like a grammar lesson, but with extra analysis to understand the meaning and context.

1. Part-of-speech tagging to identify the grammatical function of each word.
2. Syntactic parsing to analyze the grammatical structure of sentences.
3. Named entity recognition to identify and classify named entities in text.
4. Semantic analysis to understand the meaning and context of words and phrases.

From here, the models grasp the nuance and styles of human language, a key piece of sentiment analysis accuracy.

Use NLP tools to identify the sentiment of each message

Once the message is processed and understood, NLP tools can determine the message’s sentiment. This involves:

1. Lexicon-based approaches that use dictionaries of words associated with specific sentiments.
2. Machine learning algorithms trained on labeled datasets that classify the text as positive, negative, or neutral.
3. Deep learning models to recognize complex patterns in language.
4. Aspect-based sentiment analysis to identify sentiment.

These tools analyze messages and assign sentiment scores or categories based on the emotional tone detected.

Aware’s proprietary NLP for sentiment analysis

Aware’s NLP technology provides a powerful solution for sentiment analysis specifically designed and built for real-world datasets like Google Drive. Key features of Aware’s sentiment analysis include:

• Proprietary algorithms developed and trained on real collaboration messages, ensuring high accuracy and relevance.
• Proven performance that outperforms leading competitors, including Meta, Microsoft, and Google for near-human accuracy.
• Easy integration and intuitive dashboard, making sentiment analysis accessible and efficient for businesses.

Aware’s NLP technology leverages advanced AI/ML technologies and a deep understanding of language nuance to provide sentiment analysis results with near-human accuracy. Aware’s models are normalized for your unique digital culture to capture subtle emotional cues and context-specific sentiment other, more generic solutions may miss.

The combination of thorough data preparation, sophisticated NLP techniques, and Aware’s cutting-edge proprietary technology allows for comprehensive and accurate sentiment analysis enabling businesses to glean valuable, data-driven insights for decision-making, crisis management, risk mitigation, and improving company culture.

Webinar: Legal and IT experts on how to secure your data by educating your people.

Watch Now

Benefits of analyzing sentiment data in Google Workspace

Google Workspace is a rich source of datasets companies can use for sentiment analysis. Applying the findings from this data can provide many benefits for stakeholders, including:

• Leveraging insights from structured and unstructured data. Employees using Google Workspace generate a significant number of messages and pieces of content that can be used to make informed decisions and get the pulse of the workforce.
• Improve change management. Leadership can use sentiment analysis to lead adaptation strategies, address concerns, and smooth transitions when organizational changes are happening.
• Protect valuable data assets. Monitoring communications for keywords, topics, and toxicity can help infosec leaders watch for signs of discontent, potential data breaches, negligence, and insider threats. They can safeguard sensitive information, maintain data integrity, and support data loss prevention.
• Mitigate threats and detect anomalies. Further secure data assets through sentiment analysis by detecting deviations from usually behavior to surface threats from bad actors, identify phishing scams and malware attacks, and intercept before they succeed in exfiltrating data or further compromising company systems.
• Crisis management. Monitoring sentiment analysis in real time helps leadership in times of crisis, giving comms leaders insight for improved messaging, highlighting areas where responses are most needed, and surfacing false information to mitigate the spread of rumors and reputation-damaging information.

What Google Workspace data should be part of your data analysis?

Google apps included in Workspace are Gmail, Drive, Google Sheets, Docs, Meet, Chat, Calendar, and more. Any of them can be used for conducting sentiment analysis, although organizations may choose to exclude some datasets (for example, Sheets) where sentiment analysis is less likely to be informative.

Google Workspace itself does not offer sentiment analysis capabilities for any of its data sets, but there are a few options for adding the functionality to Google’s suite.

• Third-party add-ons.
• Google Cloud services (not part of Google Workspace, but can be integrated with Workspace) and data in Google’s cloud console.
• Duet AI, which Google recently introduced, includes NLP capabilities for some sentiment analysis, though it’s not explicitly offered for the purpose.

Each of these options requires varied levels of technical knowledge.

Limitations to using Google Workspace add-ons for sentiment analysis

Google Workspace’s add-ons provide some sentiment analysis capabilities, but users may struggle with some limitations of these add-ons, particularly those that haven’t been purpose-built.

1. Limited user access controls: Add-ons may not provide robust user access controls and permissions, which can lead to security and privacy concerns regarding sensitive data.
2. Finding suitable add-ons: Locating a product that precisely fits your sentiment analysis needs can take time, and in some cases, a suitable option meeting all your needs may not yet exist.
3. Difficulty searching large datasets: Google Workspace is a large environment, making searching for specific data a challenge. The platform doesn’t use metadata tags, which increases the difficulty and complication of locating and analyzing relevant information.
4. Version control limitations: Managing different versions of documents may become problematic, leading to inconsistencies in the sentiment analysis results over time.
5. Limited customization: While some add-ons are customizable, they may be less flexible than what’s required for complex or specialized sentiment analysis tasks.
6. Lack of advanced features: Some add-ons offer only basic functionality, lacking advanced capabilities like context-aware analysis or multi-language support.
7. Integration challenges: Integrating sentiment analysis results with other systems or workflows may be difficult when an add-on is only capable of operating within Google Workspace.

For more comprehensive and tailored solutions, businesses may need to consider dedicated sentiment analysis platforms to better address their requirements.

Aware Research: the risks of human behavior in workplace tools—and how to mitigate them.

Download

Challenges of sentiment analysis

While NLP and AI/ML models have come a long way, there are still challenges in how they operate and how companies use sentiment analysis tools to further their business interests.

• NLP cannot always identify humor and sarcasm in conversations, and as such, may misunderstand context when classifying messages as positive, negative, or neutral.
• Because private information may be included in the dataset, there are ethical considerations that the business world is still navigating.
• Language nuances such as slang, jargon, and abbreviations are still sometimes difficult to interpret. New communication tools, like emojis, also complicate context.

How does Aware enable sentiment analysis on Google Workspace?

Aware has a library of proprietary ML models that are purpose-built to interpret sentiment in collaboration tool messages. The majority of enterprise-level AI used today is trained on publicly available data. In contrast, Aware’s models are trained on real, hand-labeled collaboration messages. This means Aware’s models can determine positive, negative, or neutral sentiments with near-human accuracy.

With Aware, organizations using Google Drive can:

• Monitor and analyze communications in real-time, as messages are ingested as they happen.
• Understand the sentiment and context of messages and identify potential threats such as toxicity, malicious intent, or negligence that could disrupt teams or endanger the security of data.
• Surface employee sentiment surrounding change management and tailor messaging to address any employee concerns.
• Handle crisis communications in real time to minimize disinformation and rumors, and reduce employee uncertainty and reputational damage.
• Proactively manage threats to data security, including identifying phishing scams and malware attacks as they occur.

Aware in action: A large retail company adopted a task tool to facilitate internal workflows and communications initiatives between front and back-office employees. Very quickly, the operations team began receiving negative feedback and needed to understand why. They turned to Aware to understand the problem.

With Aware technology in place, they discovered their training protocols for the new task tool had gaps, and employees had resorted to old processes, which were no longer sufficient. Once the company added training tutorials for the new tool and increased support through their help desk, the negative feedback turned positive. Leadership learned how to improve future process rollouts, and to better communicate with their employees.

Unlock the power of employee sentiment insights in Google Drive. Request a demo today!

Request a demo to learn more about Google sentiment analysis from Aware