Skip to content

What is sentiment analysis? Using NLP in eDiscovery

| Written by Altlaw

As technology advances, we're seeing more artificial intelligence (AI) applications across various industries. One such application gaining traction in the legal sector is sentiment analysis.

Sentiment analysis is an AI-powered tool that allows legal professionals to better understand the documents they're reviewing by analysing the language used. 

This blog post will explore sentiment analysis, how it works and how legal professionals use it.



What is sentiment analysis?

Sentiment analysis uses natural language processing (NLP), machine learning and AI to analyse and determine the sentiment, opinion or emotion expressed in text or speech. The process involves the analysis of words and phrases used in communication, as well as the context in which they're used. By doing so, the tool can configure whether the overall sentiment is positive, negative or neutral. The sentiment analysis tool within the RelativityOne platform currently offers four sentiments — positivity, negativity, anger and desire.

Sentiment analysis can be used to analyse a wide range of communications, such as social media posts, news articles and legal documents. The tool aims to better understand the sentiment surrounding a topic or individual.


What can sentiment analysis be used for in law?

Sentiment analysis can be applied to various industries. In the legal industry, the tool can be utilised in several ways, including:

1. Analysing documents

Sentiment analysis works sentence by sentence to highlight emotions within a text by searching for indicator words that represent certain sentiments. The total amount of these words, combined with their individual rank, results in an overall score for the sentiment within a sentence.

Within the RelativityOne tool, each sentiment is assigned a colour, which helps reviewers easily identify which sentiments are present. The breakdown of sentiments allows reviewers to quickly identify important areas of a document, leading to more informed decisions on its relevance. Overall, this leads to a faster review.


2. Working with concept clustering

One of the most streamlined opportunities for collaboration between tools is combining sentiment analysis with concept clustering. After categorising your data into themed groups, you can analyse further by seeking the sentiments in each cluster.

Going one level deeper in your review can lead to more accurate document searching and batch building. For example, instead of simply searching for communications on hotels, you can search for communication regarding hotels that score negatively for sentiment. Therefore, your decisions will be more informed, and you can train your Active Learning algorithms using more relevant data.


3. Identifying emotion

The tool can be used to identify emotional distress within communications. This can be useful in cases where emotional distress is a key factor, such as harassment or discrimination.


New call-to-action


How it works

RelativityOne's sentiment analysis tool scans your documents and assigns a numerical score based on the likelihood of the sentence containing the sentiment you're looking for. The higher the confidence score, the more likely the sentence contains your desired sentiment.

The tool works by tying certain words to sentiments. For example, 'hate' would be attributed to anger and 'want' to desire. These factors ensure your results are as accurate as possible:

  • Maximum of 1,000 characters per sentence analysed.
  • No score or analysis is given to special characters or images.
  • Words such as 'and' and 'the' are not analysed.
  • Surrounding sentences do not influence the confidence score of each sentence.


Exploring the RelativityOne sentiment analysis tool

Running sentiment analysis using RelativityOne is simple. Start by choosing the document set you wish to sample (max. 300,000 documents per scan), select the sentiments you want to search for and begin your scan.

Much like other sentiment analysis tools, RelativityOne's version uses natural language processing (NLP) algorithms to identify and categorise emotions and attitudes expressed in text documents. RelativityOne's sentiment analysis can speed up reviews by:

  • Detecting unusual communicationReviewers can see a sudden tone shift within communication or trends over time.
  • Prioritising documentsDocuments to batch and send can be quickly identified based on their proximity to relevant communication.

These benefits can be useful for any investigation, but mostly those involving employee relations (harassment, wrongful termination, etc.), exfiltration or fraud.


Final thoughts on sentiment analysis and the need for human intervention

When utilising the sentiment analysis tool, it's vital to remember that the scores are predictions, not verdicts. Therefore, human intervention is imperative to ensure the most accurate results.

Reviewers must also be aware of several language considerations when handling sentiment analysis data. These include:

  • Cultural differences People from different cultures may express emotion in different ways, even if speaking the same language. Word choice and politeness standards are among the factors that can mislead analysis.
  • Age and gender Communication patterns vary among different age groups and genders.
  • Humour and sarcasm The tool will not detect sarcasm or language said in a joking manner. Words said jokingly in a negative manner or positively in a sarcastic manner will not be classed as such by the tool.
  • Contextual cues (images and emojis) Currently, sentiment analysis doesn't detect images or emojis. The presence of these can completely shift the intent of a sentence.

Overall, the sentiment analysis tool is a powerful one. It can help reviewers add an extra layer of detail to their reviews, combining topic relevance with sentiment. But that's not to say it's perfect — human intervention is still necessary, striking a balance between the reviewer and technology. It undoubtedly speeds up and produces accurate reviews, and much like most AI tools, it's expected to become more powerful in the future.


Want to learn more about RelativityOne?

RelativityOne offers several tools to provide litigation support for your eDiscovery needs, and Altlaw can ensure you get the most from the platform.

The platform offers numerous benefits and the tools to handle the largest and most complex cases. In the hands of an Altlaw expert, you can take your data further than ever before using the software.

If you're looking for a way to efficiently deal with large volumes of data in your review process, visit our RelativityOne page to learn more about the platform and how it can help you.

Visit our RelativityOne page