SiNoptiC: swarm intelligence optimisation of convolutional neural network architectures for text classification

Ferjani, Imen; Hidri, Minyar Sassi; Frihida, Ali

doi:10.1504/ijcat.2022.123237

Cited by 8 publications

(4 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our experiments, the feature extraction used model is similar to the CNN model described in [3]. It consists of an embedding layer followed by a convolutional layer based on Rectified Linear Units (ReLU).…”

Section: Baseline Modelmentioning

confidence: 99%

See 1 more Smart Citation

Dynamic Knowledge Expansion: Real-time Text Classification with Deep Convolutional Neural Networks

Ferjani,

Hidri,

Frihida

2024

Preprint

Self Cite

View full text Add to dashboard Cite

Convolutional Neural Networks (CNNs) have shown exceptional prowess in text classification in the dynamic landscape of Natural Language Processing (NLP). It is difficult, however, to adapt to evolving concepts when overwhelmed with streaming data and retraining is cumbersome. We present an incremental learning solution tailored for real-world, computationally intensive applications in this paper. Using dynamic updates, we introduce a new taxonomy for text data stream classification that addresses the formidable challenge of concept drift. Using a modest dataset, we train a lean CNN. Using transfer learning, this model intelligently gleans insights from the data stream, leveraging knowledge gained from previous datasets. In order to classify data streams in real time, an online learning mechanism is deployed. A pretrained CNN is updated periodically to capture emerging features, accelerating learning even further. Our experimental results demonstrate the superiority of our approach, setting new benchmarks in performance. Our method allows NLP systems to remain agile, efficient, and state-ofthe- art in an era of relentless data flow.

show abstract

Section: Baseline Modelmentioning

confidence: 99%

“…Next, it has a max-pooling layer followed by a fully connected layer containing a Softmax. In comparison with the model described in [3], this model has fewer filters for time optimization reasons. The model has demonstrated impressive performance and is used in this study with three different classifiers including NB, Softmax, and LR.…”

Section: Baseline Modelmentioning

confidence: 99%

Dynamic Knowledge Expansion: Real-time Text Classification with Deep Convolutional Neural Networks

Ferjani,

Hidri,

Frihida

2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…To address this issue, an innovative approach is proposed to bridge the gap between user queries and document vocabulary. One promising direction involves leveraging optimal deep learning architectures (Ferjani et al, 2022;Khoei et al, 2023) that offer a promising avenue for mitigating the impact of vocabulary differences in IR while enabling more accurate matching between user queries and document content. By learning representations that transcend individual terms, the proposed models effectively handle vocabulary mismatches and facilitate personalized information access.…”

Section: Introductionmentioning

confidence: 99%

Learning-Based Models for Building User Profiles for Personalized Information Access

Sassi Hidri

2024

IJIKM

View full text Add to dashboard Cite

Aim/Purpose: This study aims to evaluate the success of deep learning in building user profiles for personalized information access. Background: To better express document content and information during the matching phase of the information retrieval (IR) process, deep learning architectures could potentially offer a feasible and optimal alternative to user profile building for personalized information access. Methodology: This study uses deep learning-based models to deduce the domain of the document deemed implicitly relevant by a user that corresponds to their center of interest, and then used predicted domain by the best given architecture with user’s characteristics to predict other centers of interest. Contribution: This study contributes to the literature by considering the difference in vocabulary used to express document content and information needs. Users are integrated into all research phases in order to provide them with relevant information adapted to their context and their preferences meeting their precise needs. To better express document content and information during this phase, deep learning models are employed to learn complex representations of documents and queries. These models can capture hierarchical, sequential, or attention-based patterns in textual data. Findings: The results show that deep learning models were highly effective for building user profiles for personalized information access since they leveraged the power of neural networks in analyzing and understanding complex patterns in user behavior, preferences, and user interactions. Recommendations for Practitioners: Building effective user profiles for personalized information access is an ongoing process that requires a combination of technology, user engagement, and a commitment to privacy and security. Recommendation for Researchers: Researchers involved in building user profiles for personalized information access play a crucial role in advancing the field and developing more innovative deep-based networks solutions by exploring novel data sources, such as biometric data, sentiment analysis, or physiological signals, to enhance user profiles. They can investigate the integration of multimodal data for a more comprehensive understanding of user preferences. Impact on Society: The proposed models can provide companies with an alternative and sophisticated recommendation system to foster progress in building user profiles by analyzing complex user behavior, preferences, and interactions, leading to more effective and dynamic content suggestions. Future Research: The development of user profile evolution models and their integration into a personalized information search system may be confronted with other problems such as the interpretability and transparency of the learning-based models. Developing interpretable machine learning techniques and visualization tools to explain how user profiles are constructed and used for personalized information access seems necessary to us as a future extension of our work.

show abstract

“…• Recall is the ratio between the number of TP and the total number of positives, i.e., the fraction of relevant objects found in the recommendation list, i.e., the probability that a relevant object is recommended [37,38]. It is given by Equation (4).…”

mentioning

confidence: 99%

Learning-Based Matched Representation System for Job Recommendation

et al. 2022

View full text Add to dashboard Cite

Job recommender systems (JRS) are a subclass of information filtering systems that aims to help job seekers identify what might match their skills and experiences and prevent them from being lost in the vast amount of information available on job boards that aggregates postings from many sources such as LinkedIn or Indeed. A variety of strategies used as part of JRS have been implemented, most of them failed to recommend job vacancies that fit properly to the job seekers profiles when dealing with more than one job offer. They consider skills as passive entities associated with the job description, which need to be matched for finding the best job recommendation. This paper provides a recommender system to assist job seekers in finding suitable jobs based on their resumes. The proposed system recommends the top-n jobs to the job seekers by analyzing and measuring similarity between the job seeker’s skills and explicit features of job listing using content-based filtering. First-hand information was gathered by scraping jobs description from Indeed from major cities in Saudi Arabia (Dammam, Jeddah, and Riyadh). Then, the top skills required in job offers were analyzed and job recommendation was made by matching skills from resumes to posted jobs. To quantify recommendation success and error rates, we sought to compare the results of our system to reality using decision support measures.

show abstract

SiNoptiC: swarm intelligence optimisation of convolutional neural network architectures for text classification

Cited by 8 publications

References 0 publications

Dynamic Knowledge Expansion: Real-time Text Classification with Deep Convolutional Neural Networks

Dynamic Knowledge Expansion: Real-time Text Classification with Deep Convolutional Neural Networks

Learning-Based Models for Building User Profiles for Personalized Information Access

Learning-Based Matched Representation System for Job Recommendation

Contact Info

Product

Resources

About