TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN: Pointers for a news recommender project?

3 点作者 killermouse0大约 7 年前
Hey HN,<p>I&#x27;m getting started with a little project to practice my nascent machine learning skills.<p>The idea is to have users list a few websites they like and then, based on those, suggest a few interesting news as they get posted. The news will be delivered by email. I want it to be privacy focused.<p>So, I already know this is not really original. This is really more a learning project than anything.<p>The &quot;battle plan&quot; so far is: 1. Build a landing page where people can subscribe and list their favorite news sources, with a few tags. Trying to &quot;crowd source&quot; the effort of finding content and cataloging 2. Drive traffic to the landing page somehow. I was going to check AdWords campaigns but they seem too pricey for something which is basically for fun more than profit. I&#x27;ll try to rely on social discussions &#x2F; comments. Any other suggestion on this would be appreciated! 3. Ship an periodic email with 5 links or such, which are supposedly interesting &#x2F; have the email include links to &quot;thumb up &#x2F; down&quot; the content, to further customize the recommending engine 4. If there&#x27;s any visible interest and &#x2F; or if providing it starts costing me more than I&#x27;m willing to pay, I&#x27;ll attempt to monetize it. For example by adding sponsored content &#x2F; or advertising maybe?<p>What I want to gain from this test is: 1&#x2F; Practice machine learning 2&#x2F; Form some habits that will help me shipping other side projects in the future 3&#x2F; Practice other skills such as SEO, Cloud architectures (aiming at the most serverless architecture I can imagine. Right now the landing page is hosted on S3+CloudFront, the form is a Google Form (yes, that one looks cheap))<p>I would love to hear your thoughts about this all. Any suggestion?<p>The landing page is here: http:&#x2F;&#x2F;newspulse.ai Feel free to comment on it, and to fill the form (with a fake email if you don&#x27;t want to hear about it again!), that will kickstart my news source catalog.<p>Thanks for any help!

1 comment

eggie5大约 7 年前
a big problem w&#x2F; recommender systems is called the cold-start problem. One way this is manifest is when new items enter the system. In one class of recess, collaborative filtering, since the new item doesn&#x27;t have any interaction history, it will never be recommended. Then in your case where news articles are the items, then you have an extreme case where every single item is cold.<p>Look at a talk from a Dutch group called Blendle at RecSys conf where they talk about these problems.