Deep LDA : A new way to topic model

Bhat, Muzafar Rasool; Kundroo, Majid A.; Tarray, Tanveer A.; Agarwal, Basant

doi:10.1080/02522667.2019.1616911

Cited by 35 publications

(13 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the average accuracies for 18 models were relatively low even after well-established tuning parameters such as the dimension of input data set to 4,000, learning and dropout rate, or layer numbers. The proposed framework recommends two approaches for determining the correctness of topic labels in a coldstart situation based on deep learning: deep LDA [15] and GCN [8].…”

Section: Topic Labellingmentioning

confidence: 99%

“…The experiments followed best practices [15] in setting GCN model with two convolution layers plus the same hidden node numbers (330-130), and Adam optimizer. As mentioned previously, the experiments used 4 (four) selected topics and ± 27K article titles.…”

Section: Experiments On Parameters Of Gcn Modelmentioning

confidence: 99%

“…Those studies investigated topic identification from focused subjects like web documents of a selected issue crawled during three-year periods [1] or abstract texts from particular journals around two International Journal of Intelligent Engineering and Systems, Vol. 15 decades [2]. Aside from those studies that consider subject homogeneity to ensure the topic quality, other studies from more diverse news article subjects but a limited number of articles have identified topics then employed them in a classification task [3].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Graph Model and Deep Learning for Topic Labels in Classifying Short Texts of Scientific Article Titles

2022

IJIES

View full text Add to dashboard Cite

In an article searching system, topic categorization could guide researchers in finding the appropriate documents among the abundant availability of scientific articles. Because they are easily obtained from the internet, there is a preference for using short texts to full-text articles for data collection of the searching system. However, topic label scarcity becomes a problem, especially when preparing the system in a cold-start situation or without predefined topic categories. Typical topic analysis with a statistical-based unsupervised Latent Dirichlet Allocation (LDA) identifies clusters of words or topics based on the word distribution to overcome the label scarcity. For ease of use, words in LDA topics are manually observed, and some are set as topic names. Lower precision happened when tagging other articles using the LDA topics for the searching system preparation with categorization or classification approach. The precision values could be influenced by too many identified LDA topics. Thus, the overlapped context in the LDA results is possible since the same words appear in different topics, leading to many false positives. Here, the instigated problem is making classification results have comparable accuracy and precision values with the existing data condition of no topic labels and overlapped context when identifying topics. The problem solution is motivated to consider the word relations with others when identifying topics to differentiate the word context. Therefore, our contribution is to investigate LDA and relationships between words as a graph with a prevalent neural network model of deep learning called Graph Convolutional Network (GCN) for automatically determining the topics before examining them in classification tasks. Guided by the proposed framework, we synthesize training samples to make the dataset for LDA topics more similar in contexts. The empirical analysis through the experiments has thoroughly evaluated the LDA topics as a baseline to compare the results of statistical based (LDA) and Deep Learning based topic identification (Deep LDA) to ensure the topic quality. Then, we compared the usage of GCN with other frequently used text classifications. The classification results showed that the graph representation used to capture the subject context relationship between keywords in GCN has supported the balance between accuracy and precision.

show abstract

Section: Topic Labellingmentioning

confidence: 99%

Section: Experiments On Parameters Of Gcn Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Graph Model and Deep Learning for Topic Labels in Classifying Short Texts of Scientific Article Titles

2022

IJIES

View full text Add to dashboard Cite

show abstract

“…and the previous vectors w t−1 x ′ and w t−1 y ′ are const for the tth iteration, the solutions (w t x ′ and w t y ′ ) to equation (37) are also achieved by the GEV method.…”

Section: Figure 4 Directions Selection By Canonical Correlation Analysismentioning

confidence: 99%

“…Text-image recognition has drawn notable attention within machine learning and data mining communities [33][34][35]. In this paper, we extract two kinds of features for text-image recognition: the bag-of-visual SIFT (BOV-SIFT) vector from images [36] and the deep learning-based feature Deep Latent Dirichlet Allocation (DLDA) from texts [37]. (g) The BOV-SIFT feature.…”

Section: Feature Extraction For Text-image Recognitionmentioning

confidence: 99%