I thought this was interesting the first time I saw it... but you've submitted this to Show HN 13 times now, including multiple domains that redirect to the same site, and you're also using multiple HN accounts to keep resubmitting it.
Concept is great! The screenshots don't inspire a sense of it working effectively though. I don't think raw sentiment analysis is enough. 'Good' news to me has one or more of the following qualities:
- It's actionable
- It's informative in a pragmatic sense
- It invokes a positive emotion, such as curiosity, wonder, appreciation, hope, fascination, delight, laughter
- It doesn't aimlessly invoke apathy/fear/anger/disgust/disappointment/hopelessness
- It has substance and is more than a mere report or thin and reflexive reaction
- It isn't tailored to reinforce a specific political agenda
- It's within a domain of interest
- It makes be better equipped to deal with the world
That rules out almost all news articles and virtually all of the examples in the screenshots. I don't think that kind of selecting can be done purely with a sentiment analysis.
The political agenda one is important. "Good" news to me in some areas of society are very much not good news to some other people (say, an abortion-related court case being decided).
I was going to post here that it may be "good news" if you voted for him, but this brings up an interesting point: I know people that even voted for him that no longer would consider that kind of thing "good news". Makes you wonder how long it would take for the app to figure out that you no longer consider something "good news". Sort of like how I had a friend stay with me for about a month, and he used by Netflix profile, and rated a bunch of stuff (which perhaps is the rudest thing a houseguest could do, I think). I'm still recovering proper recommendations.
I think that's something really important to consider -- related the tailoring/personalization question in this thread too.
I looked at the page and every article (picked by AI) was quite negative in the sense that I believe it. They were articles about geopolitical grandstanding, extrajudicial executions of drug dealers, etc. Even if you took the pro-con Trump angle out (if possible), and put in <your favorite politician> I'd still not consider them positive topics.
My immediate reaction was this was really incongruent. I imagine it'll be for many others too.
I'm far less interested in "positive" news than in news that is relevant to me. So much of online news has become clickbait political nonsense. I'd really like a news reader that is human curated to filter out the fake nonsense that makes it into Google News, while using ML to pick out the stories that matter to me from publishers I trust.
Hello, Hacker News users! It's my first Android app and first experience of Android development :-).
SentiNews is based on sentiment analysis which helps to classify news as positive or negative. What do you think about general concept of the app? Will be waiting for your feedback.
My first reaction was : "well, if it's sentiment analysis, it doesn't know anything about if the news is bad or good, only about the mood of writer". But I then realized this is actually even better. I don't want to filter out news that are not good news, this would be plain denial. For a same news, an article can be written in a positive and analytical way, or trying to incite hate or bad feelings. This is the later I want to filter, and sentiment analysis is probably the perfect tool for that.
I would love to know how you built your training dataset (how good and bad labels were decided), because that's ultimately the choice that shapes the whole decision process. Maybe this should be a standard kind of page for products offering ML based filtering.
Also, thanks a lot for providing an "all stories" tab, additionally to "good stories" and "bad stories", this is something automatic curated content misses too often. I really love the "stories to read" mode of google now, which provides me stories based on my interests, that's basically the first thing I check every morning. But I always wonder, when I read news from there : "is this a thing for the whole world or just for me?". We need referential, the possibility to see the whole picture and to switch easily between "content for me" and "content for the world" to take advantage of the bubble without being harmed by it.
Thanks a lot for the suggestions and feedbacks to everyone.
My nearest plans: small interface redesign, improvement of sentiment analysis algorithm, personalization of news feeds and the iOS launch. Keep on updates!
Is there any reason you didn't include either the Atlantic or the National Review? For US based news, they seem pretty hard to beat if you can bear to give reporters a few hours to digest stories and provide meaningful insight.
So how is "sentiment analysis" done? I assume that this will be personalised, rather than the binary good/bad news classification?
Congratulations though, I think the concept itself is good :) But, I think it may work better if it focuses solely on the "uplifting" type of good news. (As others have pointed out, a piece of news could be good for some people, bad to others.) Most of us get upsetting news everyday, so imagine if your app is the go-to place to rinse out all that $h!++y aftertaste with positive, inspiring stories :)
Thanks a lot for the link. I've used the technology called Paragraph Vectors https://cs.stanford.edu/~quocle/paragraph_vector.pdf for sentiment features extraction. Training collections were created in a semi-automatic mode and included news title+short description gathered from popular RSS feeds.
After collection of possible positive/negative features with weights I used logistic regression classifier with some modifications (e.g. position algorithm) to classify the article. It determines the article polarity based on features (words, phrases and etc.)
I wonder if the user will like it , some authors will use terms that are all good for items considered bad by the customers -
depending on the bias of the author. Sentiment analysis will fail if an opposite political bias is considered.
Don't you think that's a bit much?
https://news.ycombinator.com/submitted?id=Yeroniomus
https://news.ycombinator.com/submitted?id=sentinewsteam
https://news.ycombinator.com/submitted?id=hacakton
https://news.ycombinator.com/from?site=sentinewsmob.ml
https://news.ycombinator.com/from?site=sentinews.ml
https://news.ycombinator.com/from?site=intellexer.com