Conference paper · 2021
Novel User Preference Recommender System Based on Twitter Profile Analysis
Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020), Advances in Intelligent Systems and Computing, vol. 1248, pp. 85–93, Springer Singapore · Published
Summary
Vajjhala et al. (2021) built a recommender system, on the IBM Watson artificial intelligence platform, that mines and analyses a user’s Twitter profile and timeline to identify interests and recommend relevant products and services. The experimental output displayed the category of goods and services the user is most likely to consume, and the authors report a strong correlation between the category of products and services a user consumes and their tweets. The abstract does not report the number of users, evaluation protocol or correlation statistics.
Practical implication. E-commerce services can use a person’s own public social media activity as the preference signal, a cold-start-friendly source of personalisation.
- The Problem
- Conventional recommender systems lack interactive ways to adjust recommendation weights and ignore the interests users express on social media.
- The Methodology
- System design and experimental evaluation: mining and analysis of Twitter profiles and timelines on the IBM Watson platform to predict product and service categories, with a reported correlation between consumed categories and tweet content.
- The Core Finding
- A recommender built on IBM Watson that mines a user’s Twitter profile and timeline predicted the category of goods and services the user is most likely to consume, and the authors report a strong correlation between consumed categories and tweet content.
- The Citation
- Vajjhala, N. R., Rakshit, S., Oshogbunu, M., & Salisu, S. (2021). Novel User Preference Recommender System Based on Twitter Profile Analysis. In S. Borah, R. Pradhan, N. Dey, & P. Gupta (Eds.), Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020) (pp. 85–93). Springer Singapore. https://doi.org/10.1007/978-981-15-7394-1_7
The summary, key findings, methods, implications and relevance notes on this page are this website’s own description of the paper, written from the published abstract and text. The official abstract and citation details are given further down.
Key findings
- Finding 1. A key shortcoming of existing recommender systems identified by the authors is the lack of interactive methods to dynamically change the weights of recommendation algorithms.
- Finding 2. The proposed system uses a user’s Twitter profile and tweets to identify interests and recommends relevant products and services, by mining and analysing the Twitter timeline.
- Finding 3. The recommender is built on the IBM Watson AI platform; experimental results displayed the category of goods and services the user is most likely to consume.
- Finding 4. The authors report a strong correlation between the category of products and services a user consumes and the content of their tweets.
Source: Vajjhala et al. (2021), Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020), Advances in Intelligent Systems and Computing, vol. 1248, pp. 85–93, Springer Singapore. DOI: 10.1007/978-981-15-7394-1_7
The study in detail
What question does this paper answer?
Can a user’s Twitter profile and timeline be mined to infer their interests and recommend relevant products and services, addressing the lack of interactive ways to adjust recommendation weights in conventional recommender systems?
What did the study find, in detail?
The proposed system mines and analyses a user’s Twitter profile and tweets to identify interests and then recommends relevant products and services. Built on the IBM Watson AI platform, the experimental output displayed the category of goods and services the user is most likely to consume, and the authors report a strong correlation between the category of products and services a user consumes and their tweets.
Why does it matter?
Recommender systems help consumers make informed, personalised decisions, but most lack interactive control over algorithm weights. Using a person’s own public social media activity as the preference signal offers e-commerce services a cold-start-friendly source of personalisation.
Methods and data
| Question | Can Twitter profile and timeline analysis drive personalised product and service recommendations? |
|---|---|
| Design | System development with experimental demonstration |
| Data | Users’ Twitter profiles and tweets (timelines) |
| Platform | IBM Watson artificial intelligence platform |
| Main result | The system displayed the category of goods and services a user is most likely to consume, with a strong correlation between consumed categories and tweets |
| Implication | Social media text can serve as a preference signal for personalised e-commerce recommendation |
| Citation | Vajjhala et al. (2021) · DOI 10.1007/978-981-15-7394-1_7 |
Definitions and key terms
- Recommender systems
- Systems that can help provide preference-based personalised services to consumers and help them make informed decisions. — as used in Vajjhala et al. (2021)
- Twitter profile and timeline analysis
- Mining and analysis of a user’s Twitter profile and tweets to identify the user’s interests, which the proposed system uses to recommend relevant products and services. — as used in Vajjhala et al. (2021)
Abstract
Recommender systems can help provide preference-based personalized services to consumers and help them make informed decisions. However, a key shortcoming of the recommender systems is the lack of interactive methods to dynamically change the weights of recommendation algorithms. Our proposed system uses the Twitter profile and tweets to identify the interests of a user and then recommends the relevant products and services to that user. Our recommendation system is built to predict and personalize products and services based on the result of mining and analyzing the user’s Twitter timeline. The proposed recommender system is built upon an artificial intelligence platform called IBM Watson. The experimental result from the platform displayed the category of goods and services the user is most likely to consume. Our recommender system also showed a strong correlation between the category of products and services a user consumes and his/her tweets.
Abstract as published in Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020).
Keywords: Personalization; Collaborative; Recommender; E-commerce; Twitter; Mining; Hybrid; Retrieval
Limitations
- The abstract does not report the number of users, evaluation protocol or correlation statistics behind the reported relationship.
- Based on Twitter (now X) data and the IBM Watson platform as available at the time of the study.
Practical implications
- E-commerce services can treat social media text, such as a user’s Twitter profile and tweets, as a preference signal for personalised product and service recommendations.
- Researchers citing the reported strong correlation between consumed categories and tweets will not find the number of users, evaluation protocol or correlation statistics in the abstract.
- For researchers studying user profiling from Twitter/X data, the recommender reflects Twitter (now X) data and the IBM Watson platform as available at the time of the study.
Drawn from Vajjhala et al. (2021), Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020), Advances in Intelligent Systems and Computing, vol. 1248, pp. 85–93, Springer Singapore.
When this research may be relevant
This paper may be relevant to researchers studying social-media-based recommender systems, user profiling from Twitter/X data, personalisation in e-commerce, hybrid recommendation approaches, and applications of the IBM Watson platform.
Research topics addressed: recommender systems; personalisation; Twitter data mining; social media analytics; user profiling; IBM Watson; e-commerce; hybrid recommendation; text mining; consumer preference prediction
How to cite
- APA 7
- Vajjhala, N. R., Rakshit, S., Oshogbunu, M., & Salisu, S. (2021). Novel User Preference Recommender System Based on Twitter Profile Analysis. In S. Borah, R. Pradhan, N. Dey, & P. Gupta (Eds.), Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020) (pp. 85–93). Springer Singapore. https://doi.org/10.1007/978-981-15-7394-1_7
- MLA 9
- Vajjhala, Narasimha Rao, et al. “Novel User Preference Recommender System Based on Twitter Profile Analysis.” Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020), edited by Samarjeet Borah et al., Springer Singapore, 2021, pp. 85–93. https://doi.org/10.1007/978-981-15-7394-1_7.
- IEEE
- N. R. Vajjhala, S. Rakshit, M. Oshogbunu, and S. Salisu, “Novel User Preference Recommender System Based on Twitter Profile Analysis,” in Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020), S. Borah, R. Pradhan, N. Dey, and P. Gupta, Eds. Springer Singapore, 2021, pp. 85–93, doi: 10.1007/978-981-15-7394-1_7.
BibTeX
@inproceedings{vajjhala2021twitter,
title = {Novel User Preference Recommender System Based on Twitter Profile Analysis},
author = {Vajjhala, Narasimha Rao and Rakshit, Sandip and Oshogbunu, Michael and Salisu, Shafiu},
booktitle = {Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020)},
series = {Advances in Intelligent Systems and Computing},
editor = {Borah, Samarjeet and Pradhan, Ratika and Dey, Nilanjan and Gupta, Phalguni},
pages = {85--93},
year = {2021},
publisher = {Springer Singapore},
doi = {10.1007/978-981-15-7394-1_7},
url = {https://doi.org/10.1007/978-981-15-7394-1_7}
}RIS
TY - CPAPER TI - Novel User Preference Recommender System Based on Twitter Profile Analysis AU - Vajjhala, Narasimha Rao AU - Rakshit, Sandip AU - Oshogbunu, Michael AU - Salisu, Shafiu ED - Borah, Samarjeet ED - Pradhan, Ratika ED - Dey, Nilanjan ED - Gupta, Phalguni T2 - Soft Computing Techniques and Applications: Proceeding of the International Conference on Computing and Communication (IC3 2020) T3 - Advances in Intelligent Systems and Computing PY - 2021 DA - 2020/11/28 VL - 1248 SP - 85 EP - 93 PB - Springer Singapore SN - 2194-5357 DO - 10.1007/978-981-15-7394-1_7 UR - https://www.narasimharao.net/research/twitter-profile-recommender-system-ibm-watson/ LA - en ER -
Links and files
- DOI: https://doi.org/10.1007/978-981-15-7394-1_7
- Publisher’s page: https://link.springer.com/chapter/10.1007/978-981-15-7394-1_7
- This page as text: https://www.narasimharao.net/research/twitter-profile-recommender-system-ibm-watson/paper.md
- Permanent address of this page: https://www.narasimharao.net/research/twitter-profile-recommender-system-ibm-watson/
Related research
Related publications on machine learning and data science applications
The survey organizes visual data mining techniques, tools and collaborative filtering approaches, and concludes that combining visual data mining with collaborative filtering has led to substantial improvement and can alleviate the shortcomings of traditional collaborative filtering.
Across 92 primary studies (2016–2025), Diffusion Transformers and Gaussian Splatting account for about 36% of studies dated 2024 or later while GAN use has fallen below 3%; training data are demographically skewed and under-documented, and the dominant metrics (PSNR, SSIM, SyncNet) do not measure what deployment requires.
Fine-tuned DINOv2 and EfficientNet-B0 performed best on nine-class DeepWeeds data, but differences among several high-performing models were not consistently statistically significant across random seeds.
On ten years of daily data for Sofia, Berlin and Tokyo, linear regression matched random forest for temperature at a fraction of the run time and clearly outperformed it for humidity (R² 0.940–0.964), while decision trees performed worst.
A prefetching approach that builds an online navigation graph from server logs and clusters navigation behaviour with WEKA outperformed conventional prefetching techniques on domain client-group logs, particularly once server idle time allowed prefetching across all user predictions.