Social networks play a key role in providing a new approach to collecting information regarding mobility and transportation services. To study this information, sentiment analysis can make decent observations to support intelligent transportation systems (ITSs) in examining traffic control and management systems. However, sentiment analysis faces technical challenges: extracting meaningful information from social network platforms, and the transformation of extracted data into valuable information. In addition, accurate topic modeling and document representation are other challenging tasks in sentiment analysis. We propose an ontology and latent Dirichlet allocation (OLDA)-based topic modeling and word embedding approach for sentiment classification. The proposed system retrieves transportation content from social networks, removes irrelevant content to extract meaningful information, and generates topics and features from extracted data using OLDA. It also represents documents using word embedding techniques, and then employs lexicon-based approaches to enhance the accuracy of the word embedding model. The proposed ontology and the intelligent model are developed using Web Ontology Language and Java, respectively. Machine learning classifiers are used to evaluate the proposed word embedding system. The method achieves accuracy of 93%, which shows that the proposed approach is effective for sentiment classification.

Transportation sentiment analysis using word embedding and ontology-based topic modeling

Deep neural network for hierarchical extreme multi-label text classification

This study analyzed 928 full-text research articles retrieved from DESIDOC Journal of Library and Information Technology for the period of 1981–2018 using Latent Dirichlet Allocation. The study further tagged the articles with the modeled topics. 50 core topics were identified throughout the period of 38 years whereas only 26 topics were unique in nature. Bibliometrics, ICT, information retrieval, and user studies were highly researched areas in India for the epoch. Further, Spain and Taiwan showed common research trends and areas as India whereas India has quite distinct research interests from America and China. Therefore, researchers in Library and Information Science in India should pay more attention to the topics which are under-researched. Further, it was found that there were some unique sub-fields to Indian Library and Information Science research, such as open access; online exhibition; virtual libraries; multimedia libraries; open source software; library automation; and library management system. With the passage of time topics evolve over time, new topics emerge, and old ones become obsolete. Topic modeling not only helps the researcher to determine the trending themes or related fields with respect to their field of interest but also helps them to identify new concepts and fields over time.

Mapping of topics in DESIDOC Journal of Library and Information Technology, India: a study

Correction to Annals of Applied Statistics 1 (2007) 17--35 [doi:10.1214/07-AOAS114]

Correction: A correlated topic model of Science

In this paper we present an analysis on the usage of Deep Neural Networks for extreme multi-label and multiclass text classification. We will consider two network models: the first one is formed by a word embeddings (WEs) stage followed by two dense layers, hereinafter Dense, and a second model with a convolution stage between the WEs and the dense layers, hereinafter CNN-Dense. We will take into account classification problems characterized by different number of labels, ranging from an order of 10 to an order of 30,000, showing the different performances of the neural networks varying the total label number and the average number of labels for sample, exploiting the hierarchical structure of the label space of the dataset used for experimental assessment. It is worth noting that multi-label classification is an harder problem if compared to multi-class, due to the variable number of labels associated to each sample. We will even investigate on the behaviour of the neural networks as function of the training hyperparameters, analysing the link between them and the dataset complexity. All the result will be evaluated using the PubMed scientific articles collection as

Deep Convolution Neural Network for Extreme Multi-label Text Classification

With the explosive growth of academic writing, it is difficult for researchers to find significant papers in their area of interest. In this paper, we propose a pipeline model, named collective topical PageRank, to evaluate the topic-dependent impact of scientific papers. First, we fit the model to a correlation topic model based on the textual content of papers to extract scientific topics and correlations. Then, we present a modified PageRank algorithm, which incorporates the venue, the correlations of the scientific topics, and the publication year of each paper into a random walk to evaluate the paper’s topic-dependent academic impact. Our experiments showed that the model can effectively identify significant papers as well as venues for each scientific topic, recommend papers for further reading or citing, explore the evolution of scientific topics, and calculate the venues’ dynamic topic-dependent academic impact.

Collective topical PageRank: a model to evaluate the topic-dependent academic impact of scientific papers

This﻿article﻿describes﻿how﻿text﻿documents﻿are﻿a﻿major﻿data﻿structure﻿in﻿the﻿era﻿of﻿big﻿data.﻿With﻿the﻿ explosive﻿growth﻿of﻿data,﻿the﻿number﻿of﻿documents﻿with﻿multi-labels﻿has﻿increased﻿dramatically.﻿ The﻿popular﻿multi-label﻿classification﻿technology,﻿which﻿is﻿usually﻿employed﻿to﻿handle﻿multinomial﻿ text﻿documents,﻿is﻿sensitive﻿to﻿the﻿noise﻿terms﻿of﻿text﻿documents.﻿Therefore,﻿there﻿still﻿exists﻿a﻿huge﻿ room﻿ for﻿ multi-label﻿ classification﻿ of﻿ text﻿ documents.﻿ This﻿ article﻿ introduces﻿ a﻿ supervised﻿ topic﻿ model,﻿named﻿labeled﻿LDA﻿with﻿function﻿terms﻿(LF-LDA),﻿to﻿filter﻿out﻿the﻿noisy﻿function﻿terms﻿ from﻿text﻿documents,﻿which﻿can﻿help﻿to﻿improve﻿the﻿performance﻿of﻿multi-label﻿classification﻿of﻿text﻿ documents.﻿The﻿article﻿also﻿shows﻿the﻿derivation﻿of﻿the﻿Gibbs﻿Sampling﻿formulas﻿in﻿detail,﻿which﻿ can﻿be﻿generalized﻿to﻿other﻿similar﻿topic﻿models.﻿Based﻿on﻿the﻿textual﻿data﻿set﻿RCV1-v2,﻿the﻿article﻿ compared﻿the﻿proposed﻿model﻿with﻿other﻿two﻿state-of-the-art﻿multi-label﻿classifiers,﻿Tuned﻿SVM﻿and﻿ labeled﻿LDA,﻿on﻿both﻿Macro-F1﻿and﻿Micro-F1﻿metrics.﻿The﻿result﻿shows﻿that﻿LF-LDA﻿outperforms﻿ them﻿and﻿has﻿the﻿lowest﻿variance,﻿which﻿indicates﻿the﻿robustness﻿of﻿the﻿LF-LDA﻿classifier. KeyWoRDS Function Terms, Gibbs Sampling, Graph Model, Multi Label, Parameter Estimation, Probability Generation Process, Text Classification, Topic Model

LF-LDA: A Supervised Topic Model for Multi-Label Documents Classification

The textual data grows explosively with the advent of the era of big data, a significant portion of textual data is text documents labeled with multi-label such as the papers with keywords. Multi-label classification is a power technology to handle the multi-labeled textual data, but a huge room stays for improving the effect of multi-label classifying for textual data. This paper introduces labeled LDA with function terms (LF-LDA), a topic model that extracts noisy function terms from textual data to improve the performance of multi-label classification. The experimental result on RCV1-v2 textual dataset shows that LF-LDA can outperform the other two state-of-art multi-label classifiers: Tuned SVM and L-LDA on both Macro-F1 and Micro-F1 metrics. The low variance also indicates LF-LDA is a robust classifier.

LF-LDA: A Topic Model for Multi-label Classification

Chinese text classification problem was studied based on domain ontology graph (DOG) of semi-supervised conceptual clustering to solve the problem that English word disambiguation method cannot be applied to Chinese text classification. Structure model of domain ontology graph, text classification algorithm in HowNet dictionary and KLSeeker ontology and so on were used to realize accurate classification of Chinese text and display effectiveness of algorithm. Chinese text classification model in domain ontology graph based on conceptual clustering was developed from the angle of decreasing human participation in ontology construction as much as possible in the paper. Aimed at application domain of Chinese web text, the algorithm can generate DOG of knowledge conceptualization automatically. At the same time, document ontology graph (DocOG) was defined to represent contents of individual text document. DocOG extracting target realized text classification based on ontology by matching of single document ontology and domain ontology. Finally, example calculation analysis and actual data test set experiment were given in experimental stage. The result shows that proposed Chinese text classification method has higher classification accuracy and reflects effectiveness of design.

Semi supervised classification of scientific and technical literature based on semi supervised hierarchical description of improved latent dirichlet allocation (LDA)

Scientific papers play an important role in the scientific research. Scientific researchers understand the research trends of their interested scientific topics. But the huge amount of papers makes them difficult to get a global view of scientific topics. In this paper, we proposed an automatic LDA-based method to extract scientific topics and track their evolutions. The method firstly divides the papers into some subsets according to the paper’s published year, then extracts scientific topics for each subset. To tract the evolutions of topics and their relations, a directed graph is constructed in terms of KL distances. The experimental results on Amind ACM-Citation-network dataset shows that our method is reasonable.

Jialin Ma

Papers

Collective topical PageRank: a model to evaluate the topic-dependent academic impact of scientific papers

LF-LDA: A Supervised Topic Model for Multi-Label Documents Classification

LF-LDA: A Topic Model for Multi-label Classification

Semi supervised classification of scientific and technical literature based on semi supervised hierarchical description of improved latent dirichlet allocation (LDA)

Extraction and Tracking of Scientific Topics by LDA