My friend and I noticed that most students don’t like reading long news articles because it takes time and it’s very boring. So, we decided to create a mobile application, Snip, that summarizes important news articles in 60 words or less for those that are too busy and find it boring to read long stuff. You can download it on Google Play Store and let us know what you think - https://play.google.com/store/apps/details?id=com.fifthtechnologies.snip
So I would like to know what process does the app takes to summarize the article?
Is there some form of AI going in the background ?
Is it like summly ?
At the moment, we manually curate it after passing the link through an algorithm that shortens the news. Nothing complex and by the way, I think Summly was a ‘scam’ as it never lived up to the whole AI hype
Manually curtate after passing article through algorithm (engine). Isn’t that what Summly’s ‘AI’ was supposed to simulate?
Plus are you thinking of automating the manual process as it isn’t scalable.
Just like Kene rightly said you need to automate the process in the long term. You might need great NLP here.
It looks hard but trust me it is easier and possible.
Yes, we understand the whole automation process and its importance but at the same time, it has its disadvantages (it doesn’t have that ‘intellect’ to know what the reader wants to read). As you said, it has to be automated in the long term but I feel having about 5-10 curators will do a better job. Summly’s AI wasn’t built with empty pockets, they had serious funding; something we don’t have. So for now, we’ll make do with what we have, though we are working towards automation.
What we did with Summly — which is different from anything else — is we’re the first to yield high-quality results that consumers would accept. Summly is an app that goes to hundreds of thousands of people that are consuming the content, so we wanted to make it really legible and readable, so, we came up with this very scalable system. We built a test, a summarizability filter, because not everything is summarizable in the current architecture; certain articles are just not written in a way from which you can take candidates and form a summary, so we detect that. We’ve done a lot of training and have a lot of data that we’ve used to come up with a rule-based filter that tells us if something is summarizable or not.
The only difference I see between Snip and Summly is $1.5million investment. Yahoo bought them, Verizon bought Yahoo and no one knows what has happened to the summarization tech. As I said it is something we are working on; you can check Snip to see what we have done so far. Thanks for the article though, it was insightful.