Based on user feedback, which highlighted a need for clear explanations alongside accurate results, together with multiple universities we developed the Result Insights feature. This feature shows you why the Keenious engine ‘thinks’ a certain result is relevant, focusing on factors such as text relevance, publication date and citation count. It's designed to give you a better understanding of our search and recommendation process by explaining why specific results are suggested and how they stand out from the rest.
How to access it
To access the Result Insights feature, simply follow these steps:
Conduct a search using Keenious to find articles relevant to your query.
In the search results, each article will be displayed on a result card.
On the result card, look for the "Result Insights" button. It's located at the top right, next to the cite and bookmark buttons.
Click the "Result Insights" button to view insights on why the article was recommended, focusing on text relevance.
How it works
Text Relevance
This indicates the balance between two primary analyses conducted by Keenious to assess the relevance of an article based on your input text. It shows the balance, not an absolute relevance score, because relevance scores can significantly vary across different searches. It's also important to note that Keenious evaluates only the title and the abstract of articles to judge their relevance.
Shared Terms: This highlights terms that appear in both your input text and the article's title or abstract. The importance of a term is influenced by its rarity and frequency. If shared terms dominate the balance, it shows that Keenious finds specific words and terms in your search more important within the article's title or abstract.
AI Predicted Meaning: This relevance is determined through AI analysis with a Large Language Model, comparing text embeddings from your search text and the article's text. If AI Predicted Meaning accounts for most of the relevance, it suggests the article's theme and topic are considered relevant by Keenious, even if the exact words are different.
Recency Boost
We increase the scores of newer articles because people usually find them more relevant. This increase is a percentage added to the article's final score. The more recent the article, the bigger the boost it gets with a maximum of 5%. For example, if an article was published a few years ago, it gets a higher boost. On the other hand, articles published more than ten years ago get a smaller boost, making them less likely to appear at the top of search results.
Citation Count Boost
Since there are many articles in our dataset that have 0 citations we have decided to give a boost to articles that have been cited at least a few times. This boost is similar to the recency boost but focuses on the number of times an article has been cited according to the OpenAlex dataset. A small number of citations can significantly increase an article's score, but this boost levels off at 100 citations with a maximum boost of 6%. Beyond this point, additional citations don't increase the boost.
Future Updates and Feedback
Please note that the scoring and features we've discussed are subject to change. Our current approach is part of our ongoing commitment to transparency, showing you what influences the search results. We plan to introduce more enhancements in the future to make it easier for you to understand how our system works.
If you have feedback on this feature, we encourage you to share it with us. Please use support@keenious.com to send your thoughts and suggestions. Your input is invaluable as we continue to improve and refine our platform.