OnePetro Paper Recommendation System: Smarter Tools for Faster Insight
Suradech Kongkiatpaiboon
Offshore Technology Conference Asia (OTC Asia), Kuala Lumpur, Malaysia, 2026 · OTC-36312-MS
Engineers lose time searching the technical literature for case studies relevant to their work. This project built a low-cost, open-source pipeline that clusters more than 2,000 conference abstracts by topic, builds a profile of each user's interests, and pushes personalized weekly paper recommendations through a chat app. Early users found most recommendations relevant and clicked on them often, which shows a practical way to bring technical knowledge into engineers' everyday tools.
Key results
- Seven coherent topic clusters emerged, including subsurface, process optimization, digital transformation, sustainability and energy transition
- Manual validation with 5 test users gave 82% precision at N=10
- Weekly push notifications achieved an average click-through rate of 59.6%
- The hybrid ranking beat a content-only baseline, especially for users with mixed interests, and digitalization and sustainability topics have grown since 2020
Approach
- Collected and filtered more than 2,000 abstracts and their metadata, and merged them with an earlier dataset of industry podcast and vendor white-paper content
- Preprocessed text (lower-casing, stop-word removal, stemming/lemmatization, tokenization) and vectorized it with TF-IDF
- Grouped abstracts into topic clusters with unsupervised modelling (K-means and LDA topic modelling are both described)
- Ranked papers with a hybrid score that combines TF-IDF cosine similarity, Jaccard category overlap and click-feedback weighting from declared interests and interaction logs
- Delivered results through a LINE chatbot using a Python service on Google Cloud Functions, Google Apps Script and the LINE Messaging API; the code is open-sourced on GitHub
The problem
Engineers spend too much time manually searching large technical-paper libraries and can miss emerging topics relevant to design and brownfield operations.
Tools & methods
Summary written for this site from the published paper. For figures, data and the full method, read the paper on OnePetro.