Appinions Developer Site

Welcome to the Appinions Developer Site

Appinions is a powerful and flexible platform that discovers and extracts opinions from around the Web, and then lets the results to be displayed in a variety of ways. For developers, we provide several different ways to create customized applications to meet your needs. This includes configuring our existing applications, or using our API to build new applications. An example of how our platform in action is http://dev.infomous.com/site/app/egypt/, which displays opinions about Libya.

Featured App: Infomous Bubble for Libya

 

A Simple Recipe: It's Elementary My Dear Watson

In case you missed it, IBM pitted their brainy computer,  Watson, against Jeopardy champs Ken Jennings and Brad Rutter and came out on top---despite thinking that Toronto was a US City.   Let's show a simple example of how you might use our APIs to gather data around this topic for purposes of building an application.    If you're disappointed that we're not using an example involving Kim Kardashian---tough! We're nerds.  

Our Grouping API is a good way to get the lay of the land around a particular topic.    It provides aggregate information across a number of dimensions like topic, publish date, author, publisher, polarity

Perhaps you're interested in the the number of opinions over time that mention Jeopardy.  The following query would give you that:

http://api.appinions.com/search/v2/groups?appkey=<yourkey>&sent=jeopardy&group=publish_date&group_rows=30

<value name="2011-02-13">81</value>
<value name="2011-02-14">231</value>
<value name="2011-02-15">220</value>
<value name="2011-02-16">265</value>
<value name="2011-02-17">345</value>
<value name="2011-02-18">75</value>
<value name="2011-02-19">28</value>
<value name="2011-02-20">64</value>
<value name="2011-02-21">74</value>
<value name="2011-02-22">119</value>
<value name="2011-02-23">94</value>
<value name="2011-02-24">95</value>

 

As you'd expect the number of opinions peaked during the three days of the match(February 14-16th), with the most occurring in the day following the match.   The real power however comes from the fact that we extract opinions and opinion holders.   Using a different grouping parameter we can get the top opinion holders for Jeopardy.

http://api.appinions.com/search/v2/groups?appkey=<yourkey>&sent=jeopardy&group=opholder_id&publish_date=[2011-02-13 TO 2011-02-17]

<value name="Ken Jennings" id="50b25a0d-1534-4a16-a116-628252f48835">48</value>
<value name="Alex Trebek" id="7b538c89-7339-4f2c-8a40-ed765c859b19">20</value>
<value name="IBM" id="7c154cab-6331-45c2-8870-ae0ae829af16">17</value>
<value name="Thomas J Watson" id="65bbaf46-6ef3-419e-b7bb-b3aab57b74d2">16</value>
<value name="David Ferrucci" id="f094b93d-cbca-4eec-90c4-969ca26fc351">12</value>
<value name="Harry Friedman" id="55ddbdb3-87dd-4e1e-b561-8cf4729c68e0">11</value>
<value name="Stephen Baker" id="c0f57a44-d5b6-408d-b8bf-c401b15620e5">11</value>
<value name="Watson" id="25937e43-bc1c-4203-b003-e9b52535ffc7">5</value>

With queries like this we can find the top authors on a particular topic, or the most popular topic in the last hour, or the polarity/sentiment distribution(number of postive or negative opinions) over a particular topic by a particular author.  We see that IBM is a frequent opinion holder on Jeopardy.   Let's see what else IBM is opining about by performing another grouping call to get the the most frequent topics discussed by IBM.  We pull the opinion holder id and use that as a constraint and get the top topics:

http://api.appinions.com/search/v2/groups?appkey=<appkey>&opholder_id=7c154cab-6331-45c2-8870-ae0ae829af16&group=topic_id

<value name="IBM" id="7c154cab-6331-45c2-8870-ae0ae829af16">548</value>
<value name="Mira" id="decce67f-6896-473d-87f8-db824c635385">26</value>
<value name="Thomas J Watson" id="65bbaf46-6ef3-41bb-b3aab57b74d2">20</value>
<value name="Watson" id="25937e43-bc1c-4203-b003-e9b52535ffc7">16</value>
<value name="China" id="a509a374-881d-4205-b67a-bb851e5d276f">11</value>
<value name="India" id="c6c3a960-c4f5-4b22-9831-a9ccde56e7a4">11</value>
<value name="N Y" id="c95203f1-45d6-4120-84e7-02dd09b83fa8">11</value>
<value name="Argonne National Laboratory" id="33b1c563-cad1-45dde8b53">9</value>
<value name="Ken Jennings" id="50b25a0d-1534-4a16-a116-628252f48835">9</value>
<value name="supercomputer" id="ac30079f-9701-3e69-8d22-00f56a094818">9</value>

We see that IBM is talking about 'Mira', its 10-petaflop supercomputer, China, etc.   We have another API that allows us to drill into the raw opinions themselves.   Using the Opinion Search API, let's look at the opinions by IBM about China.  With a query that constrains the opinion holder and topic we get back all opinions where IBM is the opinion holder and the topic is China.   Let's add a filter to only return the sentence text.

http://api.appinions.com/search/v2/opinions?appkey=<appkey>&topic_id=c6c3a960-c4f5-4b22-9831-a9ccde56e7a4&opholder_id=7c154cab-6331-45c2-8870-ae0ae829af16&rows=100&filter=sent

  • IBM says there has been growing demand for data centers and cloud computing in China.
  • IBM announced on Tuesday it is collaborating with China-based Range Technology Developmen to build a cloud computing data center in Langfang City in Hebei province.
  • China, which IBM says is dominated by Unix servers (I am pretty sure I don't believe that) nonetheless is fond of the IBM i platform in the banking and financial services sector.

There are varying degrees of authoritativeness in the opinions found above.  The first two are quoted opinions that more than likely reflect the true thoughts of the company---or at least its spokesperson.  The last one is from a trade journal.  The opinion that IBM thinks China is dominated by Unix server is perhaps less persuasive.    If you are looking for authoritative opinions you might restrict your opinions to certain publications or exclude blogs and tweets. Also, Natural Language Processing is hard!   We are working on pesky problems like person name disambiguation that are hard to crack.   We think our results are pretty cool, but  sometimes we will tell you that Toronto is in the US.    

Hope that gives you a flavor for the kinds of data you can get back from our API and sparks some ideas for the kinds of applications you want to build. 

Get Started Right Now!

Here's how you can access our API:

  1. Register for a user account.
  2. Apply for an API key.
  3. Browse the documentation