Get Started with Einstein Language
- Describe natural language processing (NLP) and how it relates to AI.
- Explain what the Einstein Language APIs are.
- Explain the different scenarios in which to use the Einstein Language APIs.
You’re a developer working for Cloud Kicks—a retail company that makes custom sneakers. Cloud Kicks uses Salesforce in all their business processes, from managing their sales cycle to order creation and fulfillment to customer support.
Cloud Kicks is growing fast. So they’re looking for ways to scale their business processes and still continue to provide the stellar customer service that they’re known for.
Right now, on the Cloud Kicks website, there’s a service request form that customers or potential customers can fill out to ask a question or get help. The information in the form is sent to a single email address at Cloud Kicks; then it’s the responsibility of one of the customer service reps to route the emails to the correct department.
When they first implemented the service request form, one person would spend a few hours each week handling and routing the emails. But business is booming, and now three customer service reps devote a few hours each day to these tasks. Cloud Kicks wants to automate the routing of these emails so that they can handle more and more requests. Then their customer service reps can focus on tasks that can put their skills to better use.
The Cloud Kicks web developer handles the code for the service request form. But the team wants you to build something that identifies the type of service request, based on the text that the customer enters in the form. Depending on the type of help the user wants, the website routes the service request to the appropriate departmental email address. This means that the website sends the service request to the right department, and customer service reps can go back to helping customers.
This is not your typical software development challenge. Like any good developer, you decide not to panic and immediately start an internet search for solutions. You discover that what you need is artificial intelligence, and in particular, natural language processing.
Natural language processing—NLP— is a term that refers to the interaction between computers and human language. The “language” part of NLP specifically means the language that humans use to communicate with each other, including both written and spoken language. NLP is an aspect of artificial intelligence (AI) in which computers analyze and understand language.
NLP has been around for a relatively long time. For example, filtering of spam emails is a practical implementation of NLP that’s been around for a while. A spam filter analyzes components of an email, such as the the subject and the email content, and looks for occurrences of certain words, phrases, and punctuation. Based on this analysis, the filter classifies the email as spam or not spam.
There’s been a resurgence in interest in NLP because advances in deep learning have made NLP even more powerful and useful in solving real-world problems. We’ve introduced a new term here, deep learning, so let’s take a short detour to talk about what it is.
Deep learning is a subfield of machine learning. In deep learning, the focus is on creating systems that mimic the function of the human brain, functions like thought or decision making. In deep learning, sophisticated algorithms are run against a lot of data to create neural networks. These neural networks learn from the data provided and can then return predictions for new data. For example, given enough data, neural networks can analyze text and make decisions about what the text means.
The confluence of two factors has brought deep learning to the forefront of AI: the availability of large quantities of data required to create useful neural networks and computers with enough processing power and speed to work with that data. Data is important because the more data you can use to create an artificial neural network, the better or more intelligent that network is. And processing power is important because it takes a lot of computational muscle to work with that data and create a neural network.
Deep learning means that NLP can now analyze text and understand words, phrases, sentences, and paragraphs to understand meaning. NLP is a challenging area for computers because language is often ambiguous and lacks precision. Idioms, words that have multiple meanings, slang, sarcasm—these are all language constructs that our human brains immediately understand. Now NLP is becoming sophisticated enough to understand ambiguous language so that text analysis can really be useful and solve problems.
Take this sentence, for example: They are visiting relatives. That sentence could mean that they have gone to visit relatives or that they are relatives who are visiting. A human can read that sentence in context and understand what it means. Deep learning applied to NLP means that now computers can understand language subtlety that it was previously unable to. Given the context of a sentence, computers can now correctly interpret ambiguous grammar.
You’ve identified the problem to solve, and you have an idea of how to solve it. What now? Luckily for you, the Salesforce Platform includes Einstein Language. You can use the Einstein Language APIs to build natural language processing into your apps and unlock powerful insights within text. Einstein Language contains two NLP services: Einstein Intent and Einstein Sentiment.
- Determine which products prospects are interested in, and send customer inquiries to the appropriate sales person.
- Route service cases to the correct agents or departments, or provide self-service options.
- Understand customer posts to provide personalized self-service in your communities.
- Identify the sentiment or emotion in a prospect’s emails to trend a lead or opportunity up or down.
- Provide proactive service by helping dissatisfied customers first or extending promotional offers to satisfied customers.
- Monitor how people perceive your brand across social media channels, identify brand evangelists, and note customer satisfaction.
The good news is that the Einstein Language APIs give you the ability to build NLP into your apps, and you don’t need a data science PhD to do it. You don’t need to know about algorithms and statistics—Einstein Language handles that for you so you can focus on solving your particular problem.
At last, the puzzle pieces are coming together. You can use the Einstein Intent API to analyze the text that the user enters into the service request form. Based on the API’s analysis, the service request can then be routed to the correct department.
- Einstein Platform Developer Guide: Introduction to Salesforce Einstein Language (Beta)
- Forbes: What Is The Difference Between Deep Learning, Machine Learning and AI?
- Algorithmia: Introduction to Natural Language Process (NLP) 2016
- Adam Geitgey’s blog series on Medium: Machine Learning is Fun!
Rights of ALBERT EINSTEIN are used with permission of The Hebrew University of Jerusalem. Represented exclusively by Greenlight.