Back to Top

 

Blog

Q2 Feature Round-Up

A Complete List of Recent AlchemyAPI Updates

For most of us, summer means backyard BBQs, beach vacations, and relaxing by the pool. While our team members at AlchemyAPI have gotten their fill of hamburgers and potato salad, we have also been busy adding new features to our natural language processing (NLP) and computer vision solutions.

With AlchemyAPI’s NLP and computer vision APIs, customers gain access to updates the moment they become available. Read on to learn about the most recent enhancements and visit our Recent Updates timeline for additional information.

1. Combined Call
With AlchemyAPI’s combined call, you are able to analyze a single piece of content (URL, HTML, Text) with multiple text and image analysis features all at once.

2. Publication Date API
This newly available API helps you group articles by extracting publication date information from web pages and normalizing the data to give you standardized formats. This API solves the problem of determining a publication date when faced with the following challenges: varied date formats (05/10/2014 or May 10th, 2014 or the 10th of May 2014), placement on the webpage (header, h1, main body), and differentiating between multiple dates on a page.

Publication date extraction combined with other text analysis features enables the generation of tag clouds, sentiment towards specific topics, and more on a temporal basis.

3. New Web Page Cleaning
The new web page cleaning system is more accurate, has increased precision on extracting the main text from article pages, and recognizes a wider variety of types of pages and handles them appropriately. Overall, the new system makes text extraction results more meaningful and is able to extract data from a larger percentage of pages than was previously possible.

4. Constituent Parser
The constituent parser builds a parse tree from the words in a sentence. Parse trees are useful structures to show the relationship between words. For example, in the phrase "new cars and trucks," we know that the word "new" applies to both cars and trucks. AlchemyAPI's technology understands the structure of complex sentences and we are now exposing this powerful process to customers.

5. New Taxonomy
In Q2, we rolled out a significant taxonomy update, offering improved accuracy on web page content analysis. Confidence scores have been improved to more accurately convey when the results can be trusted. If you need custom categories, we can help with that, too. Contact us for more information.

6. Hashtag Sentiment
Sentiment hashtag decomposition makes it possible to determine the sentiment of hashtags by splitting them into individual words and commonly used phrases.

7. AlchemyVisionImage Link Extraction and Image Tagging
With AlchemyVision, you can now apply deep learning innovations to understand a picture’s content and context.

Given any URL, the Image Link Extraction API will scan the designated page to find the most prominent image and directly retrieve the URL for that image. It can then be appropriately classified and tagged. You can use the Image Link Extraction API to aggregate images and understand the context in which they are being served.

With the Image Tagging API, you can quickly categorize and organize image libraries at a massive scale. By understanding complex visual scenes in their broader context, you can automatically extract knowledge from images and act upon what you learn.

Share/Save

Case Study - How CrisisNET is Shaping the Future of Reporting

CrisisNET

Timely, relevant, and accurate information is a powerful ally for news organizations reporting on natural disasters or humanitarian conflicts. However, gathering that data can be quite the challenge. CrisisNET, an Ushahidi initiative, is solving this problem for journalists, data scientists, and developers with a platform that provides easy access to critical information.

CrisisNET partners with AlchemyAPI to pull in data from thousands of data sources, from individual Facebook posts to UNHCR refugee updates to Ushahidi deployments. The multi-source data is first normalized into a common structure and then enhanced with relevant metadata (e.g. language, reverse geo-coding, and keywords). In minutes, this data can be made available to users through multiple API endpoints.

Imagine you want to understand the content of photos being posted about the current Iraq conflict. Specifically, you want images containing reported members from the Islamic extremist group, ISIS. Can you name the hundreds of data sources sharing information on Iraq and find the related images? If not, that means you’ll have to spend days scouring the internet for sources, scanning each one, looking for relevant photos (excluding advertisements or propaganda), and tagging each image by hand. That is a lot of work.

As the CrisisNET team shares in this case study, they leverage AlchemyAPI’s image recognition technology to dig deeper into the data stream. With AlchemyAPI, CrisisNET understands the complex content of images, extracts ISIS-related photos, and identifies images of people. By using this technology to filter the data collected, CrisisNET was able to quickly document the current situation and share the timeline below in the “Faces of ISIS” blog post.

Faces of ISIS
CrisisNET recently published a social media timeline that shares relevant stories on the Iraq crisis. They used AlchemyVision to extract and tag photos from multiple sources.

While unstructured data analysis and natural language understanding are often used by enterprises who want to learn about their customers, instances also exist in which these services are used to benefit society as a whole. “By removing barriers to accessing and analyzing unstructured information, we are hoping to impassion people who otherwise might not get involved and help drive local and global changes,” shares Jonathan Morgan, Co-founder and Technical Director at CrisisNET.

Learn more:

Share/Save

The AlchemyAPI Experience

Our Team is Building the Future of AI

By Marissa Kaufmann

When we talk about company culture at AlchemyAPI, we are not referring to the perks we offer like company-paid, twice daily Starbucks runs, team lunch on Fridays, gym memberships, bike rentals, or the breathtakingly awesome downtown office view. We are talking about the people who work at AlchemyAPI and the way they define who we are as a company.

Computer science geniuses, modern marketers and sales extraordinaires comprise our energetic team focused on delivering the latest and greatest artificial intelligence capabilities for our customers. Our team members describe us as “clever, inclusive, intellectual, entertaining, innovative, kind, and fun.” Working here offers a challenge, but we conquer it while wearing flip flops and drinking our grande triple lattes.

Colby Toland, engineer at AlchemyAPI says, “It's easy to get bored as an engineer. Eventually you master whatever you are working on. Here, I can count on the complexity of the problem to keep me exploring new knowledge every day.”

At AlchemyAPI, we seek to provide a comfortable environment for employees to grow personally, mentally, and professionally. “The majority of the work we do is on the cutting edge of what is possible,” shares Aaron Chavez, Chief Scientist.

By the way, we are always looking for smart people who want to work at a place where innovation is valued, everyone has a voice and team “bonding” happens naturally. Check out our open positions and let us know how you could make a difference as a part of our team.

Learn more about AlchemyAPI:

Share/Save

Twitter Hashtag Sentiment

Twitter Hashtag Sentiment Is Easy #ExceptWhenItsNot

By Audrey Klammer

Determining Twitter sentiment is already a challenge given the casual nature of tweets and the limits of 140 characters. But then you add hashtags: unspaced words or phrases with a ‘#’ sign in front of them. We look at a hashtag and are naturally able to break it down into the most probable set of words or phrases. A computer instead sees a hashtag as a compound word. It must first perform “sentiment hashtag decomposition” to break the hashtag into recognizable words and phrases before it can determine its sentiment.

How is hashtag sentiment performed? Instead of using the inefficient process of compiling a list of commonly used hashtags, the process of “compound word expansion” separates the hashtag into individual words using statistical language modeling. The hashtag, or the “compound word,” is “expanded” by adding spaces between what the system has determined is the most probable sequence of words. Since there are often multiple ways to chunk words in a hashtag, statistical language modeling helps to provide the most probable one.

For example, the hashtag #daretodream is a hopeful and positive statement used in a variety of ways, but adopted lately by World Cup fans. Our natural language processing (NLP) system recognizes #daretodream as a positive statement, and when it’s added to a neutral statement such as “Room on the bench for these lads?” the system realizes that it’s now a positive one (try it yourself using our online demo). Without Twitter hashtag decomposition the computer would see #daretodream as “daretodream,” instead of “dare to dream.”

Another example is the neutral statement, “Me and my bros after training,” made positive by these hashtags: #dfb #Brazil #WorldCup2014 #team #friends #training #poldi #aha #lp10 #germany.

One more example, the title of this post, “Twitter Hashtag Sentiment Is Easy #ExceptWhenItsNot” is found to be overall neutral however #ExceptWhenItsNot is correctly found to be negative.

To read more about the process of applying sentiment analysis to Twitter using AlchemyAPI, read developer Greg Case’s blog post on “Applying Sentiment Analysis to Twitter.”

Share/Save

Computer Vision Experiment Seeks to Define Beauty

What makes something #beautiful? Computer vision helps find the answer.

Is beauty really in the eye of the beholder? Richard Shaw, AlchemyAPI user, isn’t so sure. Recently, Shaw put this classic statement to the test by analyzing 1000 pictures that were tagged #beautiful on Instagram.

A person could spend hours trying to view and classify 1000 images by hand. And oftentimes, the results are not accurate or consistent enough to draw good conclusions. Shaw completed tagging his library of images in seconds using a free trial of computer vision API, AlchemyVision, to determine patterns in the data.


The bubble chart above shows the number of times each tag occurred in the 1000 sample images.

Shaw continued his experiment by looking for patterns in image color and composition for certain tags, creating a network graph to show relationship, and plotting common properties of #beautiful people to draw his conclusions. As for the question, is “beauty in the eye of the beholder”? Shaw’s experiment presumes that if it is, certain people in certain places are surely looking through the same eye.

>> Read more about Shaw's experiment.
>> Learn more about AlchemyVision or try the demo.

Share/Save

AlchemyAPI Named a Gartner 'Cool Vendor'

The Smart Machine Obsession - AlchemyAPI Recognized as 'Cool Vendor'

Smart machines, known for completing human-like tasks with astounding accuracy and speed, continue to gain popularity as enterprises and tech enthusiasts employ them to solve unique and complex data problems. These systems are helping the world's data-minded predict the spread of diseases, tag expansive image libraries, monitor social media posts for brand sentiment and so much more. But, smart machines aren't just 'cool'. Savvy developers use them to power their applications because they are effective.

Recently, Gartner, a technology and research firm, named AlchemyAPI to its “Cool Vendors in Smart Machines 2014” report.* AlchemyAPI joins a short list of leaders for its wide range of scalable AI services that help both enterprises and independent software vendors build unstructured data applications.

“We are honored to be recognized for our commitment and passion to make deep learning systems available to companies of all shapes and sizes. We’re excited to be at the forefront of the industry and look forward to the explosion of AI-powered applications coming in the near future,” shares Elliot Turner, CEO & Founder of AlchemyAPI.

Learn more with these resources:

* Gartner Inc., "Cool Vendors in Smart Machines 2014" by Tom Austin, Alexander Linden, Carol Rozwell, Kenneth F. Brant, Adib Carl Ghubril, Anurag Gupta; April 11, 2014.

Share/Save

Enhancing Cloud Apps with AI

Building AI Into Your Apps is Just a REST Call Away

The notion of artificial intelligence and smart machines may bring to mind images of apocalyptic robots, Skynet or love affairs with operating systems. But the AI we see today looks a lot different in real life and takes the form of smarter applications you probably use everyday like the U.S. Mail, Siri, GoogleNow, Pandora and Netflix.

AI is here – it’s just not what you thought it would be.

So, how do you take technology that traditionally takes thousands of Ph.D hours to create and tune, and make it useful in your own applications, for cheap? Use REST calls. Here’s how.

Here’s an example of how you can offload AI’s expensive computational work and still realize all of AI’s benefits. In this scenario, an application developer is building a voice and visual assistant that allows hikers around the world to understand the things they see on their adventures.

Let’s look at steps the application developer might take:

1) Define use case
- Answer verbal questions about plants and animals photographed on a hike
2) Define features
- Must be multi-lingual, accept voiced questions about smartphone photos
3) Determine how features can be implemented
- Hire Ph.D data scientists and buy AI processing servers OR use cloud APIs
4) Tie technologies to features and outline your workflow
5) Build the code using REST calls (download this presentation for some examples)

Whether you want to use computer vision to find each and every moment Leonardo DiCaprio appears in a film, machine learning to identify fraudulent account activity, or sentiment analysis to track how people feel about your products – numerous forms of artificial intelligence are already available for you to use in your apps. And it’s now easier than ever.

Share/Save

Sentiment-Based Music Player App

Hackers Create Mooddio - Playlists Based on Your Emotional State

Recently, Binghamton University students, Lucas Eager Leavitt and Yuval Shafir participated in the Bitcamp Hackathon at the University of Maryland. After 36 hours of intense work, Mooddio was unveiled. The music player application built with AlchemyAPI’s Sentiment Analysis API, along with a couple others, creates playlists for users based on the emotions conveyed by recent tweets. If you are tweeting about how relaxing your beach vacation is, you're likely to get an easy listening playlist, whereas if you're tweeting about poor customer service, you're going to get more tormented tunes throughout the day. Here’s what Leavitt and Shafir had to say about their app.

1. Tell us about your app. What does it do? How did you come up with the idea?
Mooddio is a music player web app that uses the sentiment of your tweets to build mood-based playlists. The user inputs their Twitter handle and the app begins playing music that is classified as happy, excited, sad, angry or relaxing, based on the sentiment of their most recent tweets. We came up with the idea for Mooddio on the bus ride to Bitcamp after trying to think of a simpler way to find music that suits your emotional state at any given moment.

2. Which feature of AlchemyAPI did you use? Which APIs did you use? Which programming language is the app written in?
We used AlchemyAPI’s Sentiment Analysis API. It was used along with the Twitter API in order to determine the disposition of a large block of tweets. This “mood” was then passed to The Echo Nest API to generate track keys for a playlist. The track keys were then used to construct a playlist and music player using the Rdio API. Our web app is written in Python, using the Flask web framework. The site itself is built using HTML, CSS and JavaScript.

3. What are your plans for Mooddio?
We are planning to add a few features and deploy the app for public use in the coming weeks. We received a lot of encouraging feedback at Bitcamp and quite a few suggestions to enhance our app. We are excited to continue working on it.

4. Who was on the team?
Lucas Eager Leavitt: I am a sophomore Computer Science major at Binghamton University. I began programming during my sophomore year of high school. This is the first year I’ve been involved in the hacker scene. Bitcamp was my second hackathon and my first experience in building a larger scale web app while working with multiple APIs.

Yuval Shafir: I am a sophomore majoring in Computer Science. I began programming just one month before entering Binghamton University. I am also a co-founders of HackBU and we recently hosted BU’s first hackathon. I first learned about AlchemyAPI at MHacks and quickly knew that I wanted to use it in a future project. When we came up with the idea for Mooddio, I realized that AlchemyAPI would perfectly fit our needs.

5. Anything else you'd like to add?
We are excited to share that Mooddio was selected for "Best Use of the Rdio API" at Bitcamp and was featured as a top startup idea in the InTheCapital article, “11 Awesome Startup Ideas Launched in Just 36 Hours at UMD”. Thanks to AlchemyAPI for the amazing sentiment analysis tool. It was the heart of our app and made it all possible.

Share/Save

Digital Disruption at The NYT

How the New AI Can Help The New York Times (and all media companies)

by Richard Leavitt, CMO at AlchemyAPI

Raising a family in a small Colorado town means you never get called a liberal wingnut, but I have always loved The New York Times and Washington Post. They're so iconic, almost mythical in stature and influence over generations of news junkies like me. Last year, Bezos bought the Post to add the digital savvy of Amazon to that storied institution. And now, The NYT is abuzz with The New York Times' Innovation Report 2014 that was recently leaked by Buzzfeed (and, here's the easy-to-download full report). If you're developing for social, mobile, media or ad:tech, read it. It's lengthy, but at 100 pages long, it is full of insights into the challenges of reaching digital media readers and thereby securing their engagement and consequently, more advertising dollars.

Many others like Mashable and Forbes have commented on its call-to-arms for attracting readers like recommending older stories, suggesting related stories, and better packaging and personalizing content to individual readers' interests. The report especially emphasizes the need to reach readers where they are, which is often on their mobile devices.

The big story for me is that these problems are already being solved everyday by companies like Pocket and AdTheorent. Plus, look at the solutions delivered by SimplyMeasured, Outbrain, Taboola, and dozens of others who know how to use natural language processing and computer vision to find, tag, categorize, rank, recommend, and personalize content in ways that appeal to readers and advertisers alike.

Take a look at The NYT action plan below and see how the new AI matches the challenge.


The New York Times' Innovation Report 2014 identifies significant problems the news organization must address. Many of these problems are already being solved by younger companies using the newest natural language processing and computer vision technologies.
Share/Save

Announcing AlchemyVision

New Computer Vision System Available to Dev Teams, Large and Small

We are excited to announce that AlchemyVision, a computer vision platform that combines our core NLP technology with 3D image analysis is available for general use today. In its first release, AlchemyVision includes APIs for image extraction and tagging. Future releases will include APIs for image search.

It's safe to say that there is a multitude of unstructured data to be mined from the millions of images shared on the web each day. If you could capture that information and associate it with complex concepts, would that be valuable? Our customers tell us it would. Social media measurement company, Simply Measured develops applications for gauging and reporting brand interest on Instagram and Pinterest. Aviel Ginzburg, CPO and Co-Founder at Simply Measured, tells us that his team accurately tagged and classified a substantial number of images with minimal human effort using AlchemyVision.

As you would anticipate, a great deal of interest in computer vision is fueled by the possibility of making a giant leap forward in tying relevant images to eCommerce. When you can turn images into engaging calls-to-action, tailor product recommendations and measure how well the images in a campaign resonate, you have a distinct competitive advantage.

There are plenty of other applications as well. For instance, media companies have massive file libraries indexed with unstructured metadata. Their workflows grind to a halt when searching for files. Eventually, some of them buy what they are looking for even though they may already have it.

The visual nature of the web is a big part of what makes it so engaging. And, AI must take vision into account since more and more of the unstructured data being added to the web is images. At AlchemyAPI, we want to transform all of the world's data into something more valuable. AlchemyVision is a start.

Learn more about AlchemyVision and try the AlchemyVision Demo.

Share/Save

Pages

Subscribe to Blog