Content based recommendation engine

NeuroMarket has a request from a media news company to publish a real time recommendation engine based on what users read now, without using visitor personal data. All recommendation should be based just on news article data available from feed. More details you can find following the link:

https://neuromarket.ai/product/news-recommendation-engine/

Based on the fact that the user sees the 1st news, we need to recommend the list of next publications that will interest visitor based on the information from the previous one:

• Title
• News text
• Date and time
• Source
• Category

Engine should support 2 languages: Ukrainian (primary) and Russian

No need to show to user the same news, but from different author, so duplicates filtering should be included in solution

Technical requirements

• 25,000 news items appear every day
• 100,000 unique users in 1 minute read the news

News feed format

<rss xmlns="http://..." version="2.0">

   <channel>

      <title>...</title>

      <link>https://.../</link>

      <image>

         <url>https://....png</url>

         <link>https://.../</link>

         <title>...</title>

      </image>

      <description>Новини України</description>

      <language>uk</language>

      <lastBuildDate>Thu, 14 Jul 2022 12:33:34 +0300</lastBuildDate>

      <item>

         <title>

            <![CDATA[ Теракт у Вінниці: зросла кількість загиблих ]]>

         </title>

         <link>https://.../.../1100...</link>

         <guid>https://.../post/1100...</guid>

         <pubDate>Thu, 14 Jul 2022 12:31:47 +0300</pubDate>

         <description>

            <![CDATA[ Внаслідок влучання російських ракет у Вінниці 8 людей загинуло]]>

         </description>

         <category>Події</category>

         ...

      </item>

      ...

   </channel>

</rss>