Sentiment analysis involves data mining large quantities of natural language text to determine the subjective viewpoints of masses of users about a given topic. In essence, sentiment analysis monitors huge amount of opinionated content to extract commercially valuable insight.
Sentiment analysis – part of the Big Data market – is often performed on streams of social media content, online reviews, survey responses, blogs, and new items. As such, the challenge of sentiment analysis is statistically quantifying material that typically contains many shades of nuanced human moods.
The Sentiment Analysis market (also known as the Text Analytics Market) is forecast to grow at a torrid CAGR of 24.2 percent from 2018 to 2025. The total market size is predicted to be 18.3 billion in 2025. SA software is a type of Big Data software. Indeed, many Big Data companies offers SA services.
Examples of the questions that users ask of sentiment analysis software include:
- Products: How are consumers responding to the new microwave?
- Movies: How is word of mouth forming the overall view of this film?
- Politics: Is this candidate resonating with the broader electorate?
This is done using sentiment analysis tools (see below), which employ artificial intelligence and deep learning techniques. These tools also employ technologies like computational linguistics and algorithmic text analysis.
At its most basic, sentiment analysis determines the positive-negative polarity of an in instance of user generated text. Does the text essentially approve, disapprove of the given topic? Or does the text offer a neutral assessment of a given topic?
Moving up in complexity, more advanced sentiment analysis offers a range of values across a spectrum of fondness to disdain.
Although the field of sentiment analysis is relatively new, its methods can be roughly grouped into two categories:
- Statistical Techniques: Statistical methods deploy tool sets from AI, deep learning and machine learning. These technologies dig into grammatical relationships, sometimes attempting to differentiate the subject of the opinion from the users offering those opinions.
- Knowledge Techniques: This methodology identifies “sign post” words that tend to be clear identifies of emotion, ranging from “love” to “hate” to “adore” to “lame.” At its more advanced, knowledge techniques can evaluate words on a spectrum of like/dislike.
Many sentiment analysis tools use a combined, hybrid approach of these two techniques. The goal of this hybrid approach is to mix tools and so create a more nuanced sentiment analysis portrait of the given subject.
How to Choose a Sentiment Analysis Tool
Sentiment analysis tools are variously described as performing opinion extraction, subjectivity analysis, opinion mining or sentiment mining. In either case, choosing the best sentiment analysis tool for your company typically includes considering the following:
- Volume of material: Estimate the amount you want to analyze – if your company is truly hoping for a deep, across-the-market sentiment analysis, you will need a larger, more robust tool.
- Test the software: Perhaps this one goes without saying, but…does the software give an accurate view of the market’s opinion? That is, if you look at the text yourself, does the software agree with human understanding?
- Total features: Sentiment analysis is a complex process. Is this tool comfortable with unstructured data. You will want to know the “F1 score,” which is a measure of a statistical test’s accuracy.
- Pricing vs. assumed value: Perhaps this one is hard to fully compute. But look at the price tag on the sentiment analysis software and then ask yourself: Are we going to make many more times that amount (in profit) based on the insight from this app? Is the app that good?
Top Sentiment Analysis Tools
The following sentiment analysis tools take a variety of different approaches, but these are all top solutions.
- IBM Watson Tone Analyzer
- Clarabridge Speech Analytics
- OpenText Sentiment Analysis
- SAP HANA Sentiment Analysis
- SAS Sentiment Analysis Tool
- Basis Technology Rosette API
- Linguamatics I2E
- Expert System’s Cognitive Automation
- Sentiment Analysis Vendors: Comparison Chart
IBM Watson Tone Analyzer
Backed by the major AI infrastructure of IBM Watson, the Tone Analyzer examines online emotions and “tones” in what users post, from Tweets to reviews to random social media.
Tone Analyzer has a particular focus on customer service and support. This may be its greatest strength. Tone Analyzer will report on support conversations to monitor whether phone agents are polite and helpful, and whether they truly answered the questions. Plus: what was the mood of the caller? Were they satisfied?
Tone Analyzer can be built into chatbots. A business can program it to create appropriate “dialogue strategies” so the conversation moves in productive patterns.
Clarabridge Speech Analytics
If the telephone is a major sales channel for you, then Clarabridge Speech Analytics may be the app for your business. The software captures and processes the speech from calls, processing the data to ascertain customer emotions and/or satisfaction.
The software automatically parses call records so key business insights are retained. To aid the software, Clarabridge analyzed massive amounts of recordings to create a portrait of typical patterns. Then its text and sentiment analytics engine plows through the material, creating metrics about the customers’ state of mind.
Clarabridge uses Natural Language Processing (NLP) and what it calls “Sentiment Scoring” to capture not only the voice, but the emotional context of the customer, from phrasing to voice tones.
OpenText Sentiment Analysis
The OpenText Sentiment Analysis module – geared for the enterprise – looks at text on a document and sentence level to get the proper context for its analysis of nuance and color. It then records whether a given comment is positive, negative, neutral or mixed. It can be deployed on-premise of (like most of the solutions in this list) via cloud computing.
The vendor also offer OpenText Semantic Navigation, a complete semantic search engine. This allows customers to get a full menu of visualization and analytic widgets, which enables a company to focus on highly specific, targeted use cases. Custom queries can be created, as can interactive visualization and alerts.
OpenText is a solution that handles social media across borders; it has full support for English, Spanish, French, German and Portuguese. As the US market grows increasingly cross-lingual (especially English-Spanish) this is an important characteristic.
SAP HANA Sentiment Analysis
SAP HANA is, of course, a large tool that handles many digital tasks. Just one of these is sentiment analysis. A strength of HANA in sentiment analysis is its ability to offer a visual view of the sentiment. The “tag cloud” offers a real time portrait of the many emotion-laden terms employed by users.
The software outputs three options: Analysis, Distinct Values and Raw Data.
The software can be set for a simple positive-negative analysis. Or it can analyze a full range of user sentiments, including Anger, Fear, Positive, Surprise, Anticipation, Joy, Sarcasm, and Negative.
The SAP HANA Sentiment Analysis tools displays a “word cloud” as part of its visual interface.
SAS Sentiment Analysis Studio
SAS Sentiment Analysis employs a sophisticated mix of linguistic and metric-based guidelines to determine if a body of text – unstructured data – is positive or negative or “unclassified.”
Users can define taxonomies at a variety of levels, with great flexibility. This is a key feature that can provide significant insight into a document/text/posting. The software can determine mood or affect from the entire document, a single idea container therein, or a given attribute or feature of a single idea. This flexibility offers great nuance into determining the flavor a text.
The software is available in English, or additional languages as application add-ons.
The SAS Sentiment Analysis Studio allows for a hybrid approach to sentiment analysis.
Basis Technology Rosette API
Basis in 2016 unveiled Rosette API, a solution that uses artificial intelligence to decipher natural language. Rosette was first developed to handle social media, complex compliance issues, and search challenges. The Rosette API takes these developments and – offered as a cloud-based tool – now enables document analysis and sentiment decoding.
Rosette includes morphological analysis, which identifies part of speech, and features lemmatization, which groups inflected forms of a word so they can be data mined as single concept.
Important for the global marketplace, Rosette includes name matching and name translation, which helps decode across languages and cultures. Its entity extraction capability helps contextualize organizations and peoples, relating them to idiomatic usage.
Among Linguamatics strong points: users can data mine text by asking questions using clear and simple language. Or, for more advanced projects, the system can handle queries requiring complex linguistic analysis.
The response to these questions – based on a deep sentiment analysis of the text – are structured clearly. Users can then place them in a variety of formats, from charts to conceptual maps to HTML tables.
A key differentiator is that I2E has a user interface that is accessible to both casual and professional uses. This feature saves significant staff time by allowing less skilled staff members produce results without help from advanced tech staff.
The software is available as on-premise or cloud-based.
Expert System’s Cognitive Automation
Expert System’s Cognitive Automation is geared to understand in a style that reflects typical human understanding. It is automated, providing the speed and machine-like consistency of software with, to an extent, the nuance and color context of a human reader.
Using a process it calls RPA (Robotic Process Automation) helps with this data-heavy analytics work.
Cognitive Automation’s use of natural language processing helps with its data mining of unstructured text. The vendor touts the application as being particularly well suited to banking and insurance text environments. As such, it can be set up for uses like claims automation, which speeds up the typically human-intensive work of claims handling.
Sentiment Analysis Vendors: Comparison Chart
|IBM Watson Tone Analyzer||particular focus on customer service and support||Backed by IBM Watson|
|Clarabridge Speech Analytics||Uses Natural Language Processing||Processes the speech from phone calls|
|OpenText Sentiment Analysis||Handles social media across borders||Complete semantic search engine|
|SAP HANA Sentiment Analysis||Offer a visual view of the sentiment||Three options for output|
|SAS Sentiment Analysis Studio||Sophisticated mix of linguistics||Define taxonomies at many levels|
|Basis Technology Rosette API||Deploys lemmatization||Uses artificial intelligence|
|Linguamatics I2E||Query in simple or complex language||Accessible to both casual and expert users|
|Expert System’s Cognitive Automation||Robotic Process Automation||Suited to banking and insurance|