NewsSift, First Impressions Matter in Search
Posted by Jeff as Uncategorized
I’ve been curious about the Financial Times’ news search service, called NewsSift, because it promises to leverage the substantial amount of metadata about financial news held by the FT for a pseudo semantic search experience. In other words, better targeting and categorization about news stories.
So I fired up the browser and headed on over to NewsSift only to be confronted by a less than obvious user experience. The obligatory keyword search box is up top along with a 6 category boxes for Organization, People, Place, Theme, and Business Topic. As you type the boxes under the keyword search term dynamically change to reflect found elements, which while useful is also not obvious what is happening in the absence of UI cues to guide the user.
I decided to watch the online demo of the service after having determined through several less than fruitful searches that I wasn’t educated on how the service works. The demo is clear and having got the gist of it I gave it another run, this time using the search term “Apple” as was used in the demo video. Here’s what it returned:
There are some interesting sidebar features, as you can see above, that point to usefulness despite the less than stellar search results displayed above. I am also curious how they differentiate online news from blogs and why news portals should be included at all if they are largely aggregating the other sources, i.e. how is AP via Yahoo Finance different from AP in any other category? At any rate, that news source widget is dynamic and does change according to the search you do. I would have also liked to see the inclusion of wikipedia-like sources as these are often very valuable in the context of search.
Even with the potential for skewing, the sentiment potentially does present a feature largely unavailable in other search services and it is useful, as evidenced by the the search I did on Geithner. What was interesting about this example is that the sentiment on the blogs category largely mapped to the online news source category.
The NewsSift service shows promise as it attempts to extract key entities and phrases for a better search experience but this thing need more time in the oven before the FT can claim a win. There are many features that are badly needed, like sentiment analysis, but if the search results are noisy then you really can’t assign too much value to the extras.
While I am actually quite positive on NewsSift the fact remains that if they do not deliver better search results the service will be destined to be a follower and not a leader. First impressions matter tremendously in search and no one can claim that within minutes of the first search they do on a newly found service that they haven’t formed an opinion, it’s binary and if you are running a search service there is nothing more important than the accuracy of the results (especially when using a search term that is used in the online demo).