I've been developing my own algorithm for news shortening. Basically, it takes a news article from the site, does some calculations and spits out up to 5 the most important sentences. Demo can be seen here <a href="https://excerptdaily.com" rel="nofollow">https://excerptdaily.com</a><p>I created it because I read news and I hate reading Bible-size articles full of unnecessary information just to find the main point<p>It has a purpose, it really does solve a problem:
* Save people's time
* Inform you as fast as possible
* Give you the main point of an article in 5 sentences
* Save you from clickbait or half clickbait titles<p>Starting my own news website without any connections or audience doesn't make sense, also I'm bad at marketing. I firmly believe this is a very good solution. I just don't know yet how to utilize it?<p>Should I offer the power of algorithm to some podcast that have audience and their own news website, should I offer it to someone who wants to build a news website...
Can you articulate why this is different to other text summarising solutions?<p>It may be that yours is easier to integrate than using AWS APIs[1], or performs better than what's available on say, npm[2]. It may be that your algorithm is designed specifically for news articles.<p>If you can articulate where this fits into the market of other solutions - that will help inform how best to utilize it.<p>[1] <a href="https://aws.amazon.com/blogs/machine-learning/part-1-set-up-a-text-summarization-project-with-hugging-face-transformers/" rel="nofollow">https://aws.amazon.com/blogs/machine-learning/part-1-set-up-...</a><p>[2] <a href="https://www.npmjs.com/package/text-summary" rel="nofollow">https://www.npmjs.com/package/text-summary</a>
I would try to optimize it for one specific use case, not have a general API that does this. I think it's much easier to sell something more niche.<p>Find a job that requires reading long text. Let's say something in healthcare where they need to read a lot of journal articles. Now you're not a "summarization API" you're a way to reduce your time spent reading medical journals, time that could be better spent savings lives (your new tagline: "Less reading, more saving lives" -- half kidding). You can also optimize your tool to summarize journal articles which are written in a very specific way. When you sell, you can set yourself apart by being made especially for them.<p>I wouldn't sell to podcasters or anyone in media because they have no money (just look at how much writers are paid, look at media company valuations...).
Extractive summarisation of news isn't very hard, I guess it doesn't hurt to put it in an API wrapper and have a pay per use model, but don't expect this to sell.<p>Also FYI, depending on the News outlet the important info is usually at the top - it's maybe the first thing they teach you in Journalism (don't bury the lede). You don't need to read the "Bible-size" article if you read the first paragraph and it's well written.<p>However, if you did abstractive summarisation instead, that might be more interesting esp for financial news - you might have buyers.
I have reservations about this because in a lot of the news today the actual details, the important ones, are so buried in the body of the body that it probably won't be surfaced given all the fluff.<p>EDIT: this is to say, garbage in, garbage out. The product you have built is probably great.
It's worth noting that this is such an old problem that traditional news-writing styles incorporate a solution: they require that the author cram the most important info into the first few sentences, then expand on it for the rest of the article.<p>I'm not sure how well it's followed these days, and I'm sure that few places other than very traditional news publishers enforce it at all. Old archival newspapers (think 19th and early 20th century) tend to be entirely in this style, though.
A good bet is to create an API for it and sell it on Rapidapi or some site, with a possible free tier.<p>I think a lot of devs would benefit from a good text summarisation algorithm (haven't tested yours just general advice) and since youre good at programming and not marketing make some npm modules, composer packages, gems and the whole shebang.<p>Soon you'll be making a few $k a month depending on how good and fast your api is. The free tier will help you get some search engine traffic. An on page demo is also very useful in this regard.<p>Also this has chance of landing some big co with deep pockets who just finds your product a good fit and couldn't be bothered to hire a dev to do this just yet. So make sure you have a $$$ unlimited plan. Good luck.
Use the technology to identify which articles talk about the same topic. Then create a site where it is possible to get the full picture of an event by easily accessing the different sources.
Do people want this problem solved?<p>> Yes. Saving time is good.
> No, as a news reader I prefer accuracy to speed. As a news reader I care about information coming directly from a source rather than summarised. As a reader I want easy access to a source of information and not have to go through a hoop of using another website/app. As a reader I want to read on mobile/laptop interchangeable and with the same interface. As a reader I want to take in deep knowledge and not summaries. As a reader either I a) read high density articles deeply and do not want a summary or b) read low density articles quickly and do not need a summary.<p>If so, are there current solutions?<p>> Yes, see other comments.<p>If so, are you able to do this cheaper or to a higher quality than the current solutions?<p>> Find the cost of other services and compare that to the cost you can deliver this solution at scale to actual users.
> Find a metric to compare your services accuracy/speed to other peoples solution.
> X-axis; quality. Y-axis; cost. Plot all the solutions.
Is your cheaper for some level of accuracy than a competitor? That’s promising!<p>If so, how can you get this packaged to users?<p>> mobile, web
> premium sources and pay walls
> reader apps<p>Just a few stray thoughts. A friend and I worked on something very very similar. Ultimately we stopped as we found that no one really wants to use this and pay for it. It’s “cool engineering” but people like reading just fine. Also tweets exist! You can find summaries easily that are human created and better synthesis.
This sounds like a <i>very</i> useful feature for a personal RSS reader, like the text-equivalent of a thumbnail, so you can decide if the article is worth reading.
Depending on the quality, you could offer it as a tool for lawyers to access all the paperwork they have to process.<p>> If dispute resolution is the social function of the law, what we have is far from the most efficient way to reach fair or reasonable resolutions. Instead, modern litigation can be understood as a massive, socially unnecessary arms race, wherein lawyers subject each other to torturous amounts of labor just because they can. In older times, the limits of technology and a kind of professionalism created a natural limit to such arms races, but today neither side can stand down, lest it put itself at a competitive disadvantage.<p><a href="https://www.newyorker.com/news/daily-comment/you-really-dont-need-to-work-so-much" rel="nofollow">https://www.newyorker.com/news/daily-comment/you-really-dont...</a><p><a href="https://news.ycombinator.com/item?id=31787599" rel="nofollow">https://news.ycombinator.com/item?id=31787599</a>
The website exactly the way you have it seems pretty great to me. You could add a non intrusive ad for revenue (I personally dislike ads, but what can you do) and then market it to the best of your ability. I think you've already made something really cool here.
One bad example may not be useful, but FWIW the first blurb I read on your demo got the salient facts of the article badly wrong:<p>"What Range Anxiety? The Mercedes-Benz EQS 580, Reviewed
Mercedes-Benz first gave us a glimpse at its electrification strategy in 33, with its first battery-electric vehiclethe EQC 23 crossovergoing on sale in Europe in 21. Sporting a range of around 2770 miles, 354 km the 402 hp 296 kW SUV never made it over to this side of the Atlantic."<p>I was quite interested in an EV with a 2770-mile range!
Maybe it could be useful to summarise YouTube videos. If you can extract the captions and display a summary, it would be great (to summarise tutorials, travel guides, and other videos).<p>macOS also has a built-in summary feature (in System Preferences > Keyboard > Shortcuts > Services > Summarise), you can use that to summarise news articles in Safari and other apps, but it doesn't work on videos, only text.
It might be interesting to expose it as a browser extension that provides a summary without leaving a site. Many non-news sites link out to news articles as references. Reducing the amount of time reading the references would be nice, but I'm unsure how it'd work in practice. It might be worth experimenting with.
Man, that's awesome. How does it work for recipes? I know there are a few solutions out there for this, but I absolutely cannot stand the 10 pages of narrative and prose before getting to the actual list of ingredients/steps.<p>I'd be curious to know if you've tested it against recipes.
This space already has a few solutions:<p><a href="https://tldrthis.com/" rel="nofollow">https://tldrthis.com/</a><p><a href="http://autotldr.io/" rel="nofollow">http://autotldr.io/</a><p>It might make sense to look at their monetization strategies.
Can we see a demo? Best I've seen in this space is <a href="https://smmry.com" rel="nofollow">https://smmry.com</a>, but probably a customized GPT-3 model would be even better.
Do people actually want news summarized? It's entertainment, parading as something intellectual. I don't want my Netflix summarized, for example.