Great Ahrefs' article (by Mateusz Makosiewicz): How to Earn LLM Citations to Build Traffic & Authority What the LLMs looks for when selecting citations? The article lists several factors that tend to correlate with being cited: > Freshness: newer or recently updated content tends to be preferred. > Domain authority: Sites with strong backlink profiles / high domain rating tend to get cited more. > Semantic relevance: Content that directly addresses the user’s query, with clear, extractable answers. > Structured & accessible formatting: clear headings, paragraphs, data in text rather than only in images, etc.
Rodrigo Rocco’s Post
More Relevant Posts
-
This article is very timely (for me at least) as I have been reflecting on how LLMs obscure the hyperlinked structure of the web. The author posits workflows where LLMs and hyperlinks work hand-in-hand in what seems to me a very viable way in continuing to work _with_ hypermedia systems. https://lnkd.in/efSNhCZX
-
Excited to share my first Data Science project! I’ve developed a Student Success Prediction System (SSPS) — a machine learning–based dashboard that predicts whether a student is “Safe” or “At Risk” based on their attendance, midterm, assignments, and final exam scores. Built With: Python (Pandas, NumPy, Scikit-learn) Streamlit (for interactive dashboards) Matplotlib & Seaborn (for data visualization) XlsxWriter (for Excel export) 🎯 Key Features: Upload student data (CSV or Excel) Train and test predictive models Visualize insights through charts and heatmaps Export predictions as Excel reports This project is a step forward in applying Data Science for Educational Analytics, helping institutions make data-driven decisions to support student success. 🔗 Project Link: https://lnkd.in/d2DTBmGz #DataScience #MachineLearning #Python #Streamlit #EducationalAnalytics #BNBWU #AI #EducationInnovation
-
Most SEOs ignore schema markup. But for Google, structured data is like subtitles — it helps it “understand” your content clearly. 🧠 Use FAQ, HowTo, and Product schema where relevant. 📍 Add author and organization schema for EEAT. 📈 Validate it in Search Console regularly. You can’t always see schema working — but you’ll feel it in visibility, CTR, and clarity.
-
The best time to prepare for AI-driven search was last year. The second-best time is today. Start by: – Auditing your content for clarity and structure. – Adding schema markup for context. – Publishing authority-based content (case studies, tutorials, data). – Building credibility with mentions and backlinks from trusted sites.
-
-
Hi, all. I just wrapped up my first end-to-end analytics project, a sentiment analysis of headlines from popular news outlets. Basically, I wanted to explore whether the tone of headlines (sentiment) correlates with a source’s political bias or reliability rating, and whether left- and right-leaning media present headlines in different ways. Using Python, I wrote a series of scripts to fetch article objects from NewsAPI (executed multiple times using a cron job), match the news sources to the bias and reliability ratings from Ad Fontes Media, and then transform them into a useable dataset in csv form. Then, with a set of more than 2,000 headlines, I used Pandas/SciPy/Seaborn to run some basic EDA on the the bias and reliability scores, followed by a sentiment analysis on the headlines using VADER (Valence Aware Dictionary and sEntiment Reasoner). The results were actually fairly boring: - Left-leaning headlines had slightly more positive sentiment on average. - Right-leaning headlines showed more variability and tended to be somewhat more negative. - There was essentially no correlation between sentiment and bias or reliability. - Differences between the bias of left- and right-leaning headlines were statistically significant but not meaningfully large (Cohen’s d ≈ 0.19). So, no great revelations, but that's okay! It was still an interesting project, and there are enough limitations (discussed in the notebook) that more research could probably glean some greater insights. Here is the Github repository: https://lnkd.in/eB34feqX And here is the actual Jupyter Notebook: https://lnkd.in/e9fWk9dB Would love to hear any feedback from those more experienced than me!
-
Im still grinding my "RAG" project!💀 Just pushed some updates to improve the interface clarity - nothing fancy yet, still looks pretty rough around the edges... 😅 The real challenge is that i'm diving into semantic search optimization. I readed a blog about multi-vector search approaches for RAG systems, and now i have a lot of questions.... https://lnkd.in/eRczDRzQ The main challenge with most "RAG" systems seems to be the "relevant context retrieval". * Multi-vector embeddings * Hybrid search strategies * Better chunk segmentation * Query expansion techniques For anyone here who worked with RAG systems 🚀 Can you give me any tips for multi-vector approaches or other techniques that significantly improved your retrieval accuracy? I would love to read your knowledge of this kind of approach for rag systems 🧠💡
-
-
[Spatial and Network Analysis System: Mathematics and Computation]: It provides analysis tools for city/urban planning and development among other real-world applications: https://lnkd.in/e-GB7PK2 It aims to build scalable ML-based predictive analysis system. It contains mathematics and computer codes for Optimization and Linear Algebra which are becoming CS and EE topics especially for big data and ML technology. The codes were tested on and migrated from SeaWulf HPC cluster at Stony Brook University. It discusses probabilistic and statistical models for spatial/future predictions, and mathematical and numerical methods for matrix multiplication and Newton-based minimization problems. Many suggest to look at the elementary econometrics and probability theory. From my experience, the best and fastest way of learning is to begin with a pretty well-written code and mathematical solution along with textbooks and hands-on practices. dh3kim at gmail dot com
-
Still exploring Semantic Link Labs report package, I looked into what we can see about bookmarks. It's not everything, but it can still be helpful. Check it out! https://lnkd.in/gyrEJSDe
-
The role of metadata in publishing success Metadata is the silent driver of discoverability in publishing. It helps readers, libraries and search engines find content efficiently. Without quality metadata, even excellent research can remain invisible. Strong metadata should include: - Accurate author and affiliation details - Keywords that match subject relevance - Structured abstracts and DOIs - Consistent file naming and tagging At Siliconchips Services we ensure metadata accuracy across all formats, from XML to ONIX, improving search visibility and indexing speed. In today’s digital-first environment, well-structured metadata is as vital as the content itself, it ensures that your research reaches the audience it deserves. #research #quality #publishing #metadata #academia #digitalpublishing #success 📺 Explore more insights on our YouTube channel: 👉 https://lnkd.in/ebtc83km Don't forget to subscribe!
-
-
Using what Michael Otto learned at the CLARIAH-AT Summer School on Machine Learning for Digital Scholarly Editions (https://lnkd.in/erwYEUVw), he created an interactive Map from the letters in the Hugo Schuchardt Archive: Through grouping the specific paragraphs in every letter by topic using a custom BERTopic, he visualized every letter on a time and place axis and tag its main contents. The code is available via GitHub at: 👉 https://lnkd.in/dVmngS8w
I am delighted to report on my participation in the CLARIAH-AT summer school on Machine Learning for Digital Scholarly Editions: https://lnkd.in/ePpsGFKU. It was weeklong, in-depth examination of all the steps necessary to generate usable data and its customized visualizations from TEI-encoded corpora of text. We mainly engaged with the BERTopic library and went into detail discussing all the main components making a Topic Modelling Pipeline. Researchers of the Know Center and the Department of Digital Humanities provided helpful insights on the theoretical groundwork underlying all of the steps necessary to transform our data. We first had a look at the concepts of Embedding, Dimensionality Reduction and Clustering before moving on to Tokenization, Weighting and Finetuning. We then also talked about the importance of proper pre- and postprocessing of data in the humanities and common python libraries used in manipulating huge amounts of TEI-encoded files. I was lucky to be provided with two datasets by my Institute beforehand, so I had the chance to deploy those concepts in extracting multiple visualizations from the corpora I was working with. Using what I learned, I created an interactive Map from the letters in the Hugo Schuchardt Archive. Through grouping the specific paragraphs in every letter by topic using a custom BERTopic, I was able to visualize every letter on a time and place axis and tag its main contents. You can find the source code for this here: https://lnkd.in/eHY_bx6p.
