Data Mining: Text Mining, Visualization and Social Media

Zero Tolerance Search : 24 year old neuroscientist

Matthew Hurst — 2012-05-12T17:44:21-04:00

[The idea behind 'zero tolerance search' posts is to illustrate real life search interactions that show how far we have to go in leveraging the explicit and implicit data in the web and elsewhere.]

Yesterday, I heard part of an interview on NPR. The interview was around a new book on determinism and neuroscience. The only thing I remember about the author was his young age. I wanted to recover the name of the author and the title of his new book so that I could comment on his argument against determinism (which was, essentially, 'I'm afraid of determinism therefore it can't be right').

The query {24 year old neuroscientist} is very clear to a human reader. The goal is to find information about a person who functions in the role of 'neuroscientist' and whose age is 24 years.

The text matching approaches of both Bing and Google essentially fails at this task. The errors they make include:

Matching on 'year[s] old' without recognizing the requirement that the result has to be about specifically a 24 year old.
Matching on the bare number '24'.
Articles about neuroscientists which also mention a '24 year old donor'
Matching on '18-24 year old samples' of the population.
A 24 year old who was studying neuroscientist before having a hit song and making a career change.
'24 year old' in the body of the document versus 'neuroscientist' in the 'about' section for a blogger.

Google edges out Bing by returning a single result in position 5 that does pertain to a 24 year old neuorscientist.

Given all the advances and trumpets employed in search these days, I still, shall we say, interested in results that ignore simple elements of document structure (the bio of the author being mixed with the content of a blog post) and inattention to elemental linguistics in the query (the '24' in '24 year old' really shouldn't match the '24' in 'March 24').

What would be a killer answer to this query would be returning a page about a person who was indeed a 24 year old neuroscientist but where the age and occupation of the individual were not present in the document.

Excellent Visualization of Network Effect

Matthew Hurst — 2012-05-12T17:28:24-04:00

]]>

Graphing Twitter Attention

Matthew Hurst — 2012-04-30T10:39:42-04:00

track // microsoft (and games and movies) now includes a simple graph indicating the attention being given to each cluster of posts. This graph shows the total of tweets per hour for all posts in the cluster. Below is an example from the cluster around Steve Wozniak's positive comments for his Windows Phone.

Microsoft's Windows Phone 8 Problem - A Solution

Matthew Hurst — 2012-04-22T11:08:02-04:00

Briefly - there is plenty of chatter (see track // microsoft) about the possibility that Microsoft won't be upgrading existing handsets to Windows Phone 8. However, Hal makes a very interesting point in his post on the topic. The problem is not the upgrade, it is the users. Simply give them all a new Windows Phone 8 hand set for free - problem solved.

Finding New Story Links Through Blog Clustering

Matthew Hurst — 2012-04-20T01:15:53-04:00

The basic mechanism used in track // microsoft to cluster articles is similar to that used by Techmeme. A fixed set of blogs are crawled and clustered based on specific features such as link structure and content (and in the case of Techmeme, additional human input). However, what about blogs that aren't known to the system?

I recently added a feature to track // microsoft which analyses clusters for popular urls and adds those to the bottom of the cluster. The title of the web page is used as a simple description of the popular page.

In the recent story about Nuno Silva's mistaken comment regarding the future of Windows Phone devices, there were many links to Nuno's own blog post. In addition to the large cluster of known blogs that were determined to be talking about the story, track // microsoft also surfaced Nuno's post through analysing the popular links discovered within the cluster.

This can be seen in this screen shot of the cluster currently appearing on the site.

Local Lens : A Hyperlocal Retrospective

Matthew Hurst — 2012-04-19T09:57:52-04:00

Thanks to a recent talk by Gonzalo, I discovered his video capturing the UX developed for our hyperlocal content aggregation and browsing application called Local Lens.

Search Engine Land's Mediocre Post on Local Search

Matthew Hurst — 2012-04-18T11:43:53-04:00

A colleague brought to my attention a post on the influential search blog Search Engine Land which makes claims about the quality of local data found on search engines and local verticals: Yellow Pages Sites Beat Google In Local Data Accuracy Test. The author describes surprise at the outcome reported - that Yellow Pages sites are better at local search than Google. Rather, we should express surprise at how poorly this article is written and at the intentional misleading nature of the title.

The article describes an analysis done by Implied Intelligence. The analysis looks at 1, 000 local businesses in the US. Here is the first problem - these businesses exclude chains and franchises. In addition, if a website wasn't known for the business, it too was excluded. With some general assumptions about the definition of local business, it is safe to assert that firstly there are many instances of chains and franchises out there and secondly that many (if not most) businesses don't have a website (the distribution varies by category of course). Quite where the original sample of 1, 000 came from is not reported.

This biases the analysis - Google, like Bing is intersted in all local entities.

The initial part of the analysis is reasonable - looking at coverage (% in the sample found on the site) and quality (duplicates, phone number errors and adderss errors). Note, however, that this is a measure of the local data, not of local search. A search product includes a relevance component and it is quite possible that a well tuned relevance algorithm might suppress duplicates.

The last table in the analysis sees us swinging back to bad reporting. It describes the percentage of records that have a certain attribute: URL, Hours of Operation and 'additional info'. Did you see what they did there? This is what we call the coverage of an attribute, and it tells us nothing as to the quality of the value. I can quite easily populate a local database with 100% coverage for all attributes. They might all be wrong, but the coverage could be 100%. Consequently, this table is reasonably close to meaningless. If they had included the precision of these values then coverage can be used to compute recall, but that wasn't done.

In summary, an important search publication has either written an intentionally misleading article, or has demonstrated that it doesn't really get data.

Tracking Microsoft - Weekly Roundup

Matthew Hurst — 2012-04-14T11:57:49-04:00

I thought I'd cull some stories from track // microsoft that summarize the week's news:

Steve Wozniak - famous for standing in line for Apple products, tweeted his intention to get into the smartphone game with a Lumia 900
The reaction of a girlfriend to the continuing Windows Phone / Lumia challenge failure of her boyfriends Android phone
Splashtop - an iPad app the emulates the Windows 8 Metro experience to give developers and UX designers a head start
From the perenial department: a rumour that Facebook would acquire Bing
AT&T states that Lumia sales have exceed expectations

Keep up to date on Microsoft news by visiting track // microsoft - a site dedicated to discovering and surfacing stories from the blogosphere about my employer!

Lumia Review Cluster

Matthew Hurst — 2012-04-08T11:25:44-04:00

Briefly, track // microsoft (a buzz tracking site for Microsoft that I'm experimenting with) is currently sporting a large story cluster focused on the reviews for the new (today!) Nokia Lumia 900.

[note - I updated this image due to fixing a bug I recently discovered with the clustering]

Tracking Video Game Buzz

Matthew Hurst — 2012-04-02T10:22:52-04:00

Briefly, I pushed out an experimental version of track // games to track tropics in the blogosphere relating to video games. As with track // microsoft it uses gathers posts from blogs, clusters them and uses an attention metric based on Bitly and Twitter to rank the clusters, new posts and videos.

Currently at the top of the stack is Bungie Waves Goodbye To Halo.

Enjoy!