Penyo Georgiev

Dr. Penyo Georgiev

orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 March 2019
Verified
Conference Paper
Closed access
Mina Nikolaeva AngelovaMina Nikolaeva Angelova
Mina Nikolaeva Angelova
Daniela PastarmadzhievaDaniela Pastarmadzhieva
Daniela Pastarmadzhieva
GDGergana Dimitrova
Gergana Dimitrova
PGPenyo Georgiev
Penyo Georgiev
As the studies and empirical data indicate that innovations are a significant factor for improving competitiveness, we believe that scientific research in this area may provide support for those who seek to make their business more innovative. There are many works dedicated to general effect of innovations on growth, but less are focused on the role of innovations in certain sectors. Such sector is might be the wine industry. The wine industry in Bulgaria has a long history, experience and has the potential for development. That is why we examine the scope of innovations that can be implemented in the wine industry. Alongside we study the potential of Bulgarian wine industry. The results show that the wine producers in Bulgaria make slow steps towards more contemporary development of the sector, but the more significant results are yet to come.

1 December 2018 in Journal of International Studies
Verified
Journal Article
Open access
Mina Nikolaeva AngelovaMina Nikolaeva Angelova
Mina Nikolaeva Angelova
Daniela PastarmadzhievaDaniela Pastarmadzhieva
Daniela Pastarmadzhieva
PGPenyo Georgiev
Penyo Georgiev
GDGergana Dimitrova
Gergana Dimitrova

8 September 2026 in Electronics
Verified
Journal Article
Open access
PDF file
SCStoyan Nikolov Cheresharov
Stoyan Nikolov Cheresharov
GGGeorgi Gustinov
Georgi Gustinov
VTVeneta Veselinova Tabakova-Komsalova
Veneta Veselinova Tabakova-Komsalova
ISIvan Stoyanov
Ivan Stoyanov
PGPenyo Georgiev
Penyo Georgiev
Semantic search over domain-specific corpora requires an effective embedding model and infrastructure. Elasticsearch’s native semantic_text field and ELSER sparse-vector inference require a commercial Enterprise subscription, inaccessible to most academic institutions. This paper documents the licensing barrier and presents a reproducible manual pipeline achieving comparable semantic search with free Elasticsearch components and open-source small language models (SLMs), on a four-node Raspberry Pi 4 cluster (8 GB RAM, three-node Elasticsearch) over 1088 USGS documents. Five models were evaluated —ELSER v2, .multilingual-e5-small, all-MiniLM-L12-v2, all-mpnet-base-v2, and msmarco-MiniLM-L12-cos-v5—from 35 screened candidates, plus a BM25 lexical baseline. Available process memory, not compute, is the binding constraint: Elasticsearch’s footprint consumes 5–6 GB of the 8 GB. The three 384-dimensional models reindexed the corpus in 2.1–2.2 h; the 768-dimensional all-mpnet-base-v2 took 10.1 h. On-disk size is unreliable for provisioning: .multilingual-e5-small expands from 1.4 GB on disk to 3.5 GB at runtime (2.5×). Retrieval quality was assessed with Precision@10, MRR, MAP@10, and nDCG@10 over 15 human-judged queries rather than raw similarity scores; embedding-based retrieval outperforms BM25, reaching significance for two of five configurations. Msmarco-MiniLM-L12-cos-v5 offers the strongest quality-per-resource trade-off among the 384-dimensional models for 8 GB ARM deployments. Pipeline and configuration artefacts are documented and reproducible.

21 September 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Galina IlievaGalina Ilieva
Galina Ilieva
Tania YankovaTania Yankova
Tania Yankova
MRMargarita Lyubenova Ruseva
Margarita Lyubenova Ruseva
DADelian Ivanov Angelov
Delian Ivanov Angelov
SKStanislava Klisarova-Belcheva
Stanislava Klisarova-Belcheva
ADAngel Enchev Dimitrov
Angel Enchev Dimitrov
TikTok supports fashion discovery and creator-mediated brand communication. This study examines associations among TikTok fashion-brand content involvement, brand awareness, brand engagement, and purchase-related intentions among Bulgarian users and illustrates a human-governed analytical architecture for structured and open survey data. An online survey yielded 295 responses; 268 eligible TikTok users comprised the sample. The architecture organizes primary measurement and structural analysis, secondary prediction and moderation, exploratory response-structure and text analysis, and descriptive multi-criteria synthesis within processing stages. Content involvement was positively associated with awareness (β = 0.683), engagement (β = 0.512), and purchase-related intentions (β = 0.283). Awareness was associated with engagement (β = 0.300) but had no incremental direct association with purchase-related intentions; engagement showed the strongest direct association (β = 0.496). The model explained 52.5% of purchase-related-intention variance, and PLS-PM prediction produced lower errors than the training-mean benchmark for all four outcome indicators. No demographic interaction received Holm-adjusted support. Illustrative multi-criteria synthesis ranked engagement first, involvement second, and awareness third under sample-derived criteria. Dictionary-assisted open-response categories concerning presentation, creators, reviews, demonstrations, music, and humor were manually checked against the responses. All reported relationships are cross-sectional associations and do not establish causal effects, intervention effectiveness, or marketing priorities. Concentrated sampling, broad working constructs, and internal validation require cautious interpretation and independent external replication.

7 September 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
Galina IlievaGalina Ilieva
Galina Ilieva
Tania YankovaTania Yankova
Tania Yankova
MRMargarita Lyubenova Ruseva
Margarita Lyubenova Ruseva
DADelian Ivanov Angelov
Delian Ivanov Angelov
SKStanislava Klisarova-Belcheva
Stanislava Klisarova-Belcheva
ADAngel Enchev Dimitrov
Angel Enchev Dimitrov
TikTok supports fashion discovery and creator-mediated brand communication. This study examines associations among TikTok fashion-brand content involvement, brand awareness, brand engagement, and purchase-related intentions among Bulgarian users and illustrates a human-governed analytical architecture for structured and open survey data. An online survey yielded 295 responses; 268 eligible TikTok users comprised the sample. The architecture organizes primary measurement and structural analysis, secondary prediction and moderation, exploratory segmentation and text analysis, and descriptive multi-criteria synthesis within processing stages. Content involvement was positively associated with awareness (β = 0.683), engagement (β = 0.512), and purchase-related intentions (β = 0.283). Awareness was associated with engagement (β = 0.300) but had no incremental direct association with purchase-related intentions; engagement showed the strongest direct association (β = 0.496). The model explained 52.5% of purchase-related-intention variance, and PLS-PM prediction produced lower errors than the training-mean benchmark for all four outcome indicators. No demographic interaction received Holm-adjusted support. Illustrative multi-criteria synthesis ranked engagement first, involvement second, and awareness third under sample-derived criteria. Dictionary-assisted open-response categories concerning presentation, creators, reviews, demonstrations, music, and humor were manually checked against the responses. Cross-sectional self-reports, concentrated sampling, broad working constructs, and internal validation require cautious, noncausal interpretation and independent external replication.

4 October 2026 in Trends in Higher Education
Verified
Journal Article
Open access
PDF file
Galina IlievaGalina Ilieva
Galina Ilieva
Tania YankovaTania Yankova
Tania Yankova
MRMargarita Ruseva
Margarita Ruseva
SKStanislava Klisarova-Belcheva
Stanislava Klisarova-Belcheva
PGPenyo Georgiev
Penyo Georgiev
GTGeorge Totkov
George Totkov
Generative artificial intelligence (GAI) is increasingly used in higher education for explanation, content generation, feedback, and personalized learning. However, conventional prompt–response chatbots may provide insufficient support for complex academic tasks requiring planning, evidence organization, comparison, justification, verification, and reflection. This study proposes the Agentic GAI-Supported Learning Framework for Higher Education (AGAI-HE), which positions GAI agents as bounded, human-supervised learning partners rather than substitutes for students or instructors. Following an instructor-led demonstration, a convenience sample of 130 students rated three approaches applied to the same e-commerce task: traditional e-learning without GAI, standard prompt–response GAI-chatbot support, and a prompt-based simulation of the AGAI-HE workflow. Students observed rather than individually executed the three approaches; no autonomous agent, persistent agent environment, API-based workflow, or integrated multi-agent system was deployed. Both AI-supported approaches received higher ratings than traditional e-learning on an overall perceived-support score, with large pairwise effects and a parallel descriptive pattern across four strongly overlapping facets. No statistically significant difference was found between the GAI chatbot and the AGAI-HE simulation. The findings constitute an exploratory, perception-based evaluation and do not validate the framework’s educational effectiveness, demonstrate an additional advantage of agentic support, or establish improved learning performance. The evaluation was limited to student perceptions; instructor acceptance, implementation feasibility, workload, and educational effectiveness were not assessed.