Communication with Computers
“Sorry, can you repeat that?”
“Pardon me?”
And of course, my favorite: “What??”
Just by reading these phrases, you can probably think of a time that they were said to you, or by you. You may think of how annoying it is when someone doesn’t understand you, or how frustrating it is when you’ve asked someone three times and you still don’t know what they said so you just say ‘yes’ or nod awkwardly. Well nod no more! Keep reading to learn about some amazing artificial intelligence technologies that could change the way we think about communication!
Natural Language Processing (NLP)
Natural language processing is a branch of artificial intelligence that helps computers understand, interpret, and manipulate human languages. Its main goal is to bridge the gap of communication between computers and humans. There isn’t one best way to accomplish this because of all the different types of data and applications for it, but in general, NLP breaks down the language into shorter, elementary pieces. This helps it understand the relations between the pieces and how they all combine to form a meaning.
Some basic tasks of NLPs include:
- Tokenization: taking a text and breaking it up into individual words (tokens)
- Lemmatization /Stemming: taking an individual word and finding the root
- Part-of-speech tagging: tagging a word as a noun, verb, adjective, etc.
- Semantic labeling: assigns words a label like agent, patient, goal, etc.
Once the NLPs can do these tasks they can be used to understand a text or generate their own. The first idea, understanding text, is called natural language understanding (NLU) and is used so that the computer comprehends what you’re saying. The second idea, generating text, is called natural language generation (NLG) and is used to get the computer to create sentences and phrases in natural language.
Some challenges of NLP include:
- Lexical ambiguity: when a word could mean different things in a sentence. Ex. “I see a bat.” The ‘bat’ could be a flying creature or a baseball bat.
- Syntactic ambiguity: when a sentence could mean different things. Ex. “Jacob looked at the dog with one eye.” The sentence could mean that the dog has one eye, or that Jacob looked with one eye.
Emotion Recognition
“by 2022, your personal device will know more about you than your own family”. -Annette Zimmermann, vice president of research at Gartner
You may think that this statement is crazy, but just two months after this was said, the University of Ohio made an algorithm that was better at recognizing emotions than a human was. The team used images of 184 people’s 18 different facial expressions. The people were a mix of different genders, ethnicities, and skin tones, and had multiple pictures were taken of each expression.
They figured out that the hue of someone’s face determined how they were feeling more than their actual expression so the team meshed the hues of expression onto neutral faces and asked participants to guess how the person was feeling. Then they showed the same images to the computer to guess. The result, computers were able to guess 20% more accurately than humans for happy, 15% more accurately for anger, the same accuracy for sad, and also detected fear which the participants didn’t even recognize!
Their algorithm has inspired other people to start thinking about emotion detection in new ways that could be even more accurate. An example of a different idea to recognize emotions is Affectiva. Their idea is a two-step process where it detects emotions from facial expressions and speech. To identify emotions from faces it uses a computer vision algorithm to landmark key features like the corners of your mouth or the edges of your eyes. Those regions are then pixelated and analyzed by deep learning algorithms to classify expressions. The expressions are then combined and mapped to emotions. To identify emotions from speech their software analyzes how the words are said in features like tone, loudness, tempo, paralinguistics, and voice quality. Then the face identification and speech identification are combined to really identify your emotions.
I realize that this is only a small step to computers understanding how you feel, but it’s a step in the right direction.
Why does it matter?
Ok, so the computer is getting better at understanding you. Who cares? What does this have to do with life? Let me show you:
Yes, your personal assistant whether it’s an Alexa, Google Home, or any other device uses natural language processing to understand you. Every time you speak to it, it does all of that processing within a matter of seconds to best serve you. These types of devices are the most obvious uses but what are some more integrated uses that we may not realize we use? Well, have you ever used a text-checking software like Grammarly? How about a language translator like Wordreference? What about a job finding site like Indeed? These programs all use NLP to read and understand your text, then to correct it, translate it, or match it. An example of NLP that we don’t necessarily think about is for the blind or deaf. There are some softwares that use NLP to read text aloud for the blind or to translate speech into words for the deaf.
Cool, we’ve found uses for NLP but what about emotion recognition? How will a computer know what you’re feeling help? One of the more prominent reasons is healthcare. If an AI algorithm can see how you feel it will dramatically improve mental health apps. By being able to tell how you feel it can give better, more intuitive advice and be more empathetic. Another use of emotion recognition is targeted advertising. If you have your camera on while scrolling through Instagram, playing games, reading an article, or even watching t.v. the companies can use emotion recognition software to find what ads you best respond to and target those towards you. Emotion recognition could also be very useful with people on the autism spectrum. Some don’t express their emotions in a way that we understand, but with this software, we can try to connect and understand more.
Key Takeaways
- Artificial intelligence is changing how we communicate
- Natural language processing (NLP) is a branch of AI that helps computers understand and manipulate human languages
- Challenges of NLP include lexical and syntactic ambiguity
- AI can now tell what you are feeling based off your facial hue
- Affectiva determines emotions through facial and speech recognition
- Both NLP and emotion recognition change / will change aspects of our lives
Thank you for reading my article on communication with computers and I hope you enjoyed it! Please leave a comment down below letting me know what you liked and what I can improve!
If you want to read some of my other articles check out The science behind a modern day craze; Artificial Intelligence.
