Google NLP API Review With A Quick Comparative Study
The Google NLP API is a popular Cloud-based data analytics platform that provides text and sentiment analysis from various data sources. In this article, we look at the API in detail including its pros and cons from the perspective of a user, and also discuss the top APIs in the market that offer sentiment mining capabilities.
What Is Google Cloud NLP Service?
The Google Cloud NLP (Natural Language Processing) Service is a Cloud-based pay-as-you-go platform from Google for textual data analytics. This text analytics API works on machine learning (ML) and is used to extract meaningful insights from data such as customer feedback, search queries, chat histories, social media, text documents, etc.
The API provides accurate sentiment analysis results and is relatively easy to use by data analysts and developers after a basic understanding of the platform. It works on a credit system, which means you purchase credits based on the number of Unicode characters you may use.
An Overall Review Of The Google NLP API
Google’s NLP Cloud API service is a fairly simple tool that can be easily integrated with third-party services and applications through the REST API. Here is an overall Google NLP API review with respect to different features.
1. Pre-trained industry models
The API is available in multiple pre-trained models built for different industries that enable software developers to apply natural language understanding (NLU) to applications through entity sentiment analysis, content classification, entity identification, and such. However, to use these custom models, one needs to use Google’s Vertex AI software, which comes at an additional cost.
The service supports 11 languages including English for sentiment analysis but only supports 3 languages - English, Spanish, and Japanese - for entity sentiment analysis.
3. Entity detection
The API can identify and extract entities from text data, which makes it easy for contextual analysis.
4 Sentiment analysis
The Google NLP API analyzes emotion in the text at an overall and entity level. However, there is no transparency, and analysts working on the platform have no view of how sentiment or salience scores are calculated.
5. Topic-based & Aspect-based sentiment at extra cost
The API provides granular topic-level and aspect-level sentiment analysis at an extra cost because it relies on its Vertex AI. Otherwise, it does not have automatic topic classification.
6. Integration with Google Docs
The API, as expected, easily integrates with Google docs and allows you to efficiently categorize documents. This makes it quite helpful as it lets you know which content is being shared and published. This is useful in news services and communication fields.
7. Knowledge graphs
The tool provides relationship graphs or knowledge graphs in relation to entity identification. But this feature is tangible only when there is a huge amount of data, which can be a hindrance for campaign-based marketing functions that may not have terabytes of data.
8. Difficult to teach new users
The tool provides fairly simple insights but is built keeping software developers in mind. Thus, it is not easy to teach new users and even older employees who are not familiar with technology, which is often the case in most marketing and marketing operations teams.
9. No Spell-check
Even though you can analyze text to extract entities and get an overall sentiment score, there is no facility for spelling correction in case an item has been misspelled. This is an interesting point we came across during the Google NLP API review. This is where the features like fuzzy matching and semantic search are handy. So either you clean your data to every syllable and character or just keep paying to clean data or running the API hits again and again until the required level of accuracy is achieved.
10. No text extraction feature
The service does not allow text extraction which means that you need to highlight or input the sentences you want to analyze. The API does not find them automatically nor extract them from a document such as a news article. Thus features like world cloud; Topic based sentiment and Aspect based sentiment analysis become inconceivable.
11. Coding requirements
The platform does require coding for many parameters. However, it is in Python, which is widely used.
12. No billing alerts
The service does not have billing alerts to remind users that they are running out of credits, which may cause them to be caught off-guard mid-project.
How Much Does The Google NLP API Cost?
The Google NLP API is around US$2 per 1M Unicode characters. To understand this pricing, we need to know how data is measured. For the purposes of text analysis, 1 unit = 1 document comprising 1000 Unicode characters. Unicode characters include whitespace characters and any markup characters such as HTML or XML tags.
Therefore, if you want to analyze a document (text sentence) that has 700 characters, you are charged for 1 unit. If your document crosses 1000 characters, you are charged for 2 units. It is easy to burn through units when analyzing pages and pages of surveys, news articles, or social media comments. That’s why pricing is based on buckets of 5K+ - 1M units or 1M+ - 5M units and so on. This is the basic premise of the Google NLP API pricing.
This can be problematic for some. This is because the average page on a website, say for review sentiment analysis, usually spans hundreds of thousands of Unicode characters. If the data is cleaned and removed of CSS and HTML tags, this can reduce the number of characters. Interestingly, the API does not do this, and hence, you can end up using all of your credits pretty fast for very basic sentiment analysis.
This cost increases in case you need more efficient, fine-grained sentiment mining from critical aspects and topics in your business feedback data to analyze customer experience.
Checkout the basic pricing Calculator without Aspect Based Sentiment Analysis or Machine learning models or Semantics: Google NL API Pricing
For Pricing with Machine learning model and built your own ABSA feature: Google Vertex AI
What Are The Top 8 Sentiment Analysis APIs?
Below is a crisp overview (including the Google NLP API) of the top 8 sentiment analysis APIs available in the market.
1. Repustate API Repustate’s sentiment analysis API provides granular aspect-based sentiment analysis from any source of text data. This includes customer experience analysis, employee feedback data, surveys, and more. It identifies and extracts entities, topics, aspects, and themes to provide deep-rooted sentiment insights that can show sentiment trends, aspect trends, aspect-emotion co-occurrence, and other crucial insights.
Deep-learning-based semantic classification and named entity recognition ensure that results are accurate and precise. The API is available in industry-specific models, and in 23 languages. Further, it can provide analysis in real-time if needed as well as from current or historical data. The API is also available with a dashboard platform that offers more capabilities such as social media sentiment analysis, enabling you to find TikTok influencers, set alerts, and more.
2. Google NLP API
Google NLP API analyzes text documents seamlessly and efficiently. It can scan various sources like news, emails, chatbot histories, websites, and others. The API needs to be used by users familiar with technology as there is a bit of coding required. The platform is reasonably cost-effective but only provides overall sentiment and entity sentiment analysis and has a much slower processing time than Repustate.
The API provides multilingual data analysis for 11 languages but entity sentiment only for 3. One thing to note is that the basic level of the API does not give any deep insights into how the sentiment is arrived at nor provides any granularity in the insights from the data. It does not allow text extraction for analysis either, which is vital for the recognition of critical keywords and phrases for sentiment analysis.
To achieve the same granularity at Repustate, in order to make more intelligent discoveries from the data, you need to switch over to Google’s Vertex AI at an additional level and cost.
3. Microsoft Azure API
Microsoft Azure provides text analytics and sentiment mining from large amounts of data seamlessly and identifies entities as well as key phrases, this makes it more in-depth. Its advantage is the fact that it also analyzes quantifiable metrics like dates, currencies, percentiles, temperatures, etc, which enrich the insights from the qualitative analysis.
Yet, its drawback is that it cannot classify certain entities, such as people or places, and this limits the accuracy of the analysis due to there not being any contextual reference.
4. Dandelion API
Dandelion API provides efficient sentiment analysis through entity extraction that increases the accuracy of its sentiment analysis results. It can be used by startups as well as enterprises for a large range of text analysis from various data sources. Like Repustate, Dandelion also uses semantic technology to identify the syntactic and semantic relationship between words for better understanding and accuracy of results.
However, it doesn’t match noun “qualifiers” later in a sentence to nouns appearing earlier, which can be problematic. The API supports 7 languages compared to 11 as mentioned above in the Google NLP API review section.
5. Aylien API
Aylien API is a notable text analysis and sentiment mining software that is apt for news analysis and analyst research for critical business insights. The API can analyze over 80,000 sources for news feeds and extracts information from them, though it also includes paid news websites that have articles that cannot be opened.
It supports 6 languages and is a fairly easy software for overall trend analysis from news sources. However, granularity and the ability to handle complex syntaxes and sentence structures are not its strong suit. It also sometimes gives non-relevant news articles based on the search query. Nevertheless, it is used by small and mid-market companies as well as institutes like S&P Global.
6. Amazon Comprehend API
Amazon Comprehend API is a great sentiment analysis and text analytics tool that can extract sentiment from data easily. It can detect spelling errors and recognizes alternate names and aliases of entities. This enhances the accuracy of the insights it provides.
It currently provides multilingual data analysis for 6 languages including English. Additionally, similar to Repustate, the API can provide real-time sentiment analysis as well as analyze data from regular feeds.
7. spaCy API
spaCy API provides text analytics and emotion detection from text data similar to Google NLP API. It is an open-source library though, and has several ML capabilities such as named entity recognition, entity linking, text classification, and others. It provides good quality insights though it cannot identify misspelled words nor semantically classify words and phrases for contextual clarity. The API has multilingual capabilities and identifies 7 languages.
8. TextRazor API
TextRazor API is an intelligent text analysis API that has several features such as entity recognition, classification, key phrase extraction, automatic topic classification, and disambiguation capabilities for complex data. The API can provide sentiment analysis insights in 12 languages though it cannot identify misspelled words that can bring the accuracy down. After all, text cleaning is essential for machine learning, especially with Python.
How Can You Find The Best API For NLP Sentiment Analysis?
Your best bet for an NLP sentiment analysis tool suited to you is one that fulfills five of the major criteria mentioned below.
- Aspect-based sentiment analysis
Aspect-based sentiment analysis means that the NLP sentiment analysis API is able to identify the key issues that the review or comment mention, which set the tone of the entire comment as positive or negative. This gives you actionable marketing or business insights rather than just an overall sentiment score.
- Entity Extraction
An entity extraction capability ensures that the API can recognize entities that occur in the data and extract them for contextual analysis, which increases the accuracy in sentiment mining. This capability depends on how powerful the tool’s named entity recognition (NER) feature is.
This is one of the most important criteria your NLP API must have. The more robust the tool’s ability to recognize and extract topics, themes, entities, aspects, keywords, important phrases, and so on, the more accurate the insights will be.
- Native Multilingual NLP
Having a multilingual sentiment analysis capability is also very important because if the tool uses translations instead of native part-of-speech taggers, it is going to dilute the meaning of the text and thus lower the accuracy rate.
- Speed & Scale
Speed and scalability are more important than you’d think. They are critical in time-sensitive industries such as fintech and healthcare and in time-critical marketing functions such as market research for new product launches, monitoring event performances, and such.
The Google NLP API is as good as they come. What you need to do is first identify your business objectives, inspect your target audience, and see what kind of insights you are exactly looking for. Are you looking for TikTok social insights, or overall sentiment insights for your brand performance, or are you searching for market insights from online sources and publications? You have to keep these end goals in mind.
However, if you are looking for more in-depth sentiment analysis results that can give you action-oriented insights for customer success, employee engagement, or brand amplification strategies, then you need to widen the scope and compare more tools as the Google NLP API review showed earlier.
Explore Repustate’s sentiment analysis API for far better-targeted sentiment insights. Whether it’s social listening on Instagram or tracking sentiment drivers across timelines, data sources, or aspects, get the most comprehensive and precise CX insights with no hidden costs, or for that matter, coding involved.