General-use unsupervised keyword extraction model for keyword analysis

Keyword extraction is a fundamental task of text mining, so extracting high-quality keywords is of great significance. Typical keyword extraction algorithms usually rely on the statistical features, but lack of the semantic information. At the same time, the supervised keyword extraction algorithms rely too much on sample labeling. Therefore, in this paper, an unsupervised keyword extraction algorithm based on multi-dimensional features called MDFKE is proposed, which combines statistical features, external knowledge-based features and semantic features. MDFKE mainly studies the semantic information of candidate keywords. LDA model is used to obtain text topic, and Word2vec word embedding is used to generate word vectors. Based on these, the similarity between candidate keyword and text topic is quantified as semantic feature. Nine specific features are extracted from five aspects: term frequency, length, position, external knowledge base, and semantics. Finally, this paper clusters on feature vectors to obtain the final keyword set. The experiment turns out that, compared with traditional keyword extraction algorithms based on statistical features, MDFKE can significantly improve extraction performance, and can also make up for the shortage of supervised learning overly relying on labels.

Large language models such as GPT-3 have revolutionised how we conduct natural language processing tasks by applying abundant pre-trained linguistic information. Other extraction models exhibit a brief scope of attention, making it difficult to map different words and phrases together. We propose one that takes the ability of the LLMs to generate fitting and suitable keywords for research works. Individuals have traditionally solved the keyword extraction tasks using algorithms such as Text Rank, TF-IDF (Term Frequency Inverse Document Frequency), RAKE (Rapid Automatic Keyword Extraction), and so many others. From the Comparative Analysis of Ten Large Language Models, we identify the GPT-2 model to be most suited for the process of Keyword Extraction.

A study of supervised term weighting scheme for sentiment analysis

News articles categorization is a supervised learning approach in which news articles are assigned category labels based on likelihood demonstrated by a training set of labeled articles. A system for automatic categorization of news articles into a standard set of categories has been implemented. The proposed work will use Term Frequency–Inverse Document Frequency (TF-IDF) term weighting scheme for optimization of classification techniques to get more optimized results and use two supervised learning approaches, i.e., Support Vector Machine (SVM) and K-Nearest neighbor (kNN) and compare the performances of both classifiers. Each news document is preprocessed and transformed into a term-document matrix (Tsoumakas et al. in Data mining and knowledge discovery handbook. Springer, Berlin, pp 667–685 (2010) [1]). After preprocessing and transforming each news article into a vector of weights, TF-IDF term weighting scheme was used for weighting the word. TF-IDF weighted the words calculating the number of words that appear in a document. An unknown news item is also transformed into a vector of keyword weights, and then categorized into suitable categories such as Sports, Business, and Science and Technology. The system purposed in research work was trained on the collection of approximately 300 categorized news articles extracted from the various Indian newspaper websites and tested on a different set of 60 randomly extracted news items from the same sources (Trstenjak et al. in Proc Eng 69:1356–1364 (2014) [2], Buana et al. Int J Comput Appl 50:37–42 [3]). It has been observed that the performances of both algorithms improve when TF-IDF approach is used.

Keyword extraction of complex hierarchical categories becomes a challenge in text mining since commonly used classification for flat categories results in low accuracy. This paper presents a method to improve keyword extraction from hierarchical categories considering terms occurred in category from a hierarchy as additional factors in term-weighting. The method is an enhancement of a basic TF-IDF calculation; thus, it can comfortably be used for keyword extraction and classification. By taking term frequency and inverse document frequency of categories hierarchically related to a focused category, we can determine how important terms are in their family categories. In this work, hierarchy relations used in calculation are sub-categories, supercategories and sibling-categories. From experiment results, we found that the proposed method gained higher accuracy for about 40% from a baseline in a classification task.

이 연구에서는 용어 클러스터링을 이용하여 단일문서의 키워드를 추출하는 알고리즘을 제안하고자 한다. 단락단위로 분할한 단일문서를 대상으로 1차 유사도와 2차 분포 유사도를 산출하여 용어 클러스터링을 수행한 결과, 50단어 단락에서 2차 분포 유사도를 적용했을 때 가장 우수한 성능을 나타냈다. 이후, 용어 클러스터링결과를 이용하여 단일문서의 키워드를 추출하기 위해 단순빈도와 상대빈도의 조합을 통해 다양한 키워드 추출 공식을 도출, 적용한 결과, 단락빈도(pf)와 단어빈도<TEX>$\times$</TEX>역단락빈도(<TEX>$tf{\times}ipf$</TEX>) 조건에서 가장 우수한 결과를 나타냈다. 이 결과를 통해, 본 연구에서 제안한 알고리즘은 좋은 키워드가 가져야 할 두 가지 조건인 주제성과 고른 빈도분포라는 측면에서 단일문서를 대상으로 효과적으로 키워드를 추출할 수 있음을 확인하였다. In this study, a new keyword extraction algorithm is applied to a single document with term clustering. A single document is divided by multiple passages, and two ways of calculating similarities between two terms are investigated; the first-order similarity and the second-order distributional similarity. In this experiment, the best cluster performance is achieved with a 50-term passage from the second-order distributional similarity. From the results of first experiment, the second-order distribution similarity was also applied to various keyword extraction methods using statistic information of terms. In the second experiment, pf(paragraph frequency) and <TEX>$tf{\times}ipf$</TEX>(term frequency by inverse paragraph frequency) were found to improve the overall performance of keyword extraction. Therefore, it showed that the algorithm fulfills the necessary conditions which good keywords should have.

Text mining is a process that uses data mining approaches to extract valuable information held in the hidden form in textual data. In this paper, we are proposing a framework for fuzzy clustering of news articles. These news articles originate on different news portals on the web. The data obtained need to be stored in a central database and then pre-processing reduces the noise. The keyword extraction is used to extract keywords from the text and then word-frequency vector is generated. On these vectors, distance measure or similarity measure function is used to find the similarity between articles. One article may belong to more than one cluster so fuzzy context vector must be generated. Mutual Information can be used to find fuzzy membership values. The threshold values are required for the identification of clusters. The proposed framework shows that fuzzy clustering does not restrict each news article to belong exactly to one cluster. Therefore this framework when applied to information retrieval systems or other application systems, system gives better performance and relevance to the users.

The online customer reviews information available on the internet about any product consider as an essential information resource concerning customer’s interest and their knowledge of the product. It is inscribed in the form of natural language and is unstructured data. To reduce the significant information in the form of a summary is vital to the firms that work on business intelligence. It will help in product recommendation and increase in customer understanding about the product. Therefore, there is much research for creating new methodologies to summarise the text in online customer reviews automatically. In this paper, RNN-Long short-term memory Tensor Flow model along with Recall-Vocabulary Again (RVA) and Copy mechanism has used for the task of generating summary in the form of term wise from the customer reviews and news article. The RNN, along with the RVA mechanism, has been trained through a feed-forward neural network with encoder–decoder to solve the general summarization. The method has validated for the efficiency of the Giga word and DUC dataset to minimize the problem of unknown words in a decoder and generate an accurate summary that contains more vital information.

Introduction: No accurate diagnostic tests exist for exocrine pancreatic insufficiency (EPI). As a result, EPI is misdiagnosed, including in chronic pancreatitis (CP), and there is an unmet need for clinical tools to help clinicians evaluate patients with CP at risk for EPI. The aim of this study was to generate a simple clinical tool to predict EPI in patients with CP. Methods: 49 variables from medical records of CP patients with or without EPI were entered into a Classification and Regression Tree (CART) model using the unimputed full analysis set and 4 other models (Logistic Regression with Least Absolute Shrinkage and Selection Operator regularization, Support Vector Machine, Random Forest, and Gradient Boosting Machine) using imputed full analysis set. EPI misclassification rate (mRate) served as primary metric to train the 5 prediction models. Each model’s generalizability was assessed using 5-fold cross-validation and the model that was best suited for clinician use based on mRate and other characteristics was selected. Results: Records of 274 patients with CP from 6 pancreatitis centers across the US were included. Among the 160 (58%)/114 (42%) patients with EPI/without EPI, 48%/55% were female, mean age was 55/52 years, 76%/69% were white, 29%/52% were never smokers, and 39%/53% were never alcohol users. All 5 prediction models demonstrated similar EPI mRates. The CART model was selected. The final CART decision tree includes 10 variables (in order of importance): smoking, pancreatic atrophy, steatorrhea, alcohol use, hemoglobin, serum vitamin D, pancreatic duct contour change, serum albumin, pancreatic parenchymal calcifications, and use of opioid analgesics. The mRate without and with 5-fold cross-validation of the CART was 0.153 (training error) and 0.314 (prediction error), and the area under the receiver operating characteristic curve was 0.889 and 0.682, respectively (Figure 1). Sensitivity/specificity without/with 5-fold cross validation was 0.888/0.789 and 0.794/0.535, respectively. Conclusion: A parsimonious set of 10 variables associated with EPI in CP was identified. The CART decision tree demonstrated good prediction performance for EPI and was selected as the most practical screening tool to aid clinicians in the assessment of EPI in patients with CP. Further external validation of the EPI screening tool is planned.Figure 1.: AUC-ROC of the CART model trained on the full analysis set and after 5-fold cross-validation. AUC, area under the curve; CART, classification and regression tree; CV, cross-validation; ROC, receiver operating characteristic.

With the availability of multigene panel testing (MGPT) for hereditary cancer risk assessment, clinicians need to assess the likelihood of pathogenic germline variants (PGVs) across numerous genes in parallel. This study's aim was to develop and validate a clinical prediction model (PREMMplus) for MGPT risk assessment. PREMMplus was developed in a single-institution cohort of 7,280 individuals who had undergone MGPT. Logistic regression models with Least Absolute Shrinkage and Selection Operator regularization were used to examine candidate predictors (age, sex, ethnicity, and personal/family history of 18 cancers/neoplasms) to estimate one's likelihood of carrying PGVs in 19 genes (broadly categorized by phenotypic overlap and/or relative penetrance: 11 category A [APC, BRCA1/2, CDH1, EPCAM, MLH1, MSH2, MSH6, biallelic MUTYH, PMS2, and TP53] and eight category B genes [ATM, BRIP1, CDKN2A, CHEK2, PALB2, PTEN, RAD51C, and RAD51D]). Model performance was validated in nonoverlapping data sets of 8,691 and 14,849 individuals with prior MGPT ascertained from clinic- and laboratory-based settings, respectively. PREMMplus (score ≥ 2.5%) had 93.9%, 91.7%, and 89.3% sensitivity and 98.3%, 97.5%, and 97.8% negative-predictive value (NPV) for identifying category A gene PGV carriers in the development and validation cohorts, respectively. PREMMplus assessment (score ≥ 2.5%) had 89.9%, 85.6%, and 84.2% sensitivity and 95.0%, 93.5%, and 93.5% NPV, respectively, for identifying category A/B gene PGV carriers. Decision curve analyses support MGPT for individuals predicted to have ≥ 2.5% probability of a PGV. PREMMplus accurately identifies individuals with PGVs in a diverse spectrum of cancer susceptibility genes with high sensitivity/NPV. Individuals with PREMMplus scores ≥ 2.5% should be considered for MGPT.

Recently, sustainable growth and development has become an important issue for governments and corporations. However, maintaining sustainable development is very difficult. These difficulties can be attributed to sociocultural and political backgrounds that change over time [1]. Because of these changes, the technologies for sustainability also change, so governments and companies attempt to predict and manage technology using patent analyses, but it is very difficult to predict the rapidly changing technology markets. The best way to achieve insight into technology management in this rapidly changing market is to build a technology management direction and strategy that is flexible and adaptable to the volatile market environment through continuous monitoring and analysis. Quantitative patent analysis using text mining is an effective method for sustainable technology management. There have been many studies that have used text mining and word-based patent analyses to extract keywords and remove noise words. Because the extracted keywords are considered to have a significant effect on the further analysis, researchers need to carefully check out whether they are valid or not. However, most prior studies assume that the extracted keywords are appropriate, without evaluating their validity. Therefore, the criteria used to extract keywords needs to change. Until now, these criteria have focused on how well a patent can be classified according to its technical characteristics in the collected patent data set, typically using term frequency–inverse document frequency weights that are calculated by comparing the words in patents. However, this is not suitable when analyzing a single patent. Therefore, we need keyword selection criteria and an extraction method capable of representing the technical characteristics of a single patent without comparing them with other patents. In this study, we proposed a methodology to extract valid keywords from single patent documents using relevant papers and their authors’ keywords. We evaluated the validity of the proposed method and its practical performance using a statistical verification experiment. First, by comparing the document similarity between papers and patents containing the same search terms in their titles, we verified the validity of the proposed method of extracting patent keywords using authors’ keywords and the paper. We also confirmed that the proposed method improves the precision by about 17.4% over the existing method. It is expected that the outcome of this study will contribute to increasing the reliability and the validity of the research on patent analyses based on text mining and improving the quality of such studies.

E-commerce has become a vital part of peoples lives. Moreover, Online customer reviews have become a crucial reference for customers decision-making process. A large number of studies have emphasized the role of review helpfulness and sentiment. However, little attention has been given to the actual relationship between review length and its informative value. This study primarily examines whether review length can be a reasonable indicator of the reviews informativeness. This study uses the Amazon Product Review dataset from Kaggle, which contains more than 568000 reviews from different categories. Using Python as the primary tool, this study conducts quantitative data analysis, data modeling, and visualization skills to explore the correlation between the two variables. The Term Frequency-Inverse Document Frequency (TF-IDF) represents the lexical richness of the review, indicating informativeness in this study. The results reveal a statistically significant positive correlation, with diminishing returns as length increases. The findings suggest that while longer reviews are generally more informative, length alone cannot fully determine review quality. This study provides exploratory insights for consumers and merchants regarding the interpretation and evaluation of online reviews.

Increasing influence of internet has led to huge amount of born-digital news articles being published on the internet. It is also becoming increasingly difficult to forage through the vast warehouse of these documents for preventing duplicity. Keywords are the most salient words in any textual document. We have introduced a graph-based approach for keyword extraction, using term co-occurrence in the textual news articles and integrating weighted closeness centrality (CC) with weighted clustering coefficient (WC). We have also proposed a metric namely co-occurrence index (CI) based on the extracted keywords for finding the amount of similarity between any two textual news articles. Our proposed method is independent of the ‘bag-of-word model’ and has shown significant performance improvement over the other existing methods.

In the last years, news agencies have become more influential in various social groups. At the same time, the media industry starts to monetize online distributed articles with contextual advertising. However, the efficiency of online marketing highly depends on the popularity of news articles. In our work, we present an alternative and effective way for article popularity forecasting with two–step approach: article keywords extraction and keywords-based article popularity prediction. We show the benefits of this technique and compare with widely used methods, such as Text Embeddings and BERT–based methods. Moreover, the work provides an architecture of the model for dynamic keyword tracking trained on the newest dataset of Russian news articles with more than 280k articles and 22k keywords for the popularity of forecasting purposes. KeywordsOnline news popularity forecastingKeyword extractionPopularity predictionBERTText embedding

Objective: With the rapid increase in the number of patent documents worldwide, demand for their automatic categorization has grown significantly. The automatic categorization of patent documents is the organization of such documents in digital form, thus replacing the manual time-consuming process. In this work, we proposed a system that can automatically categorize patent document by considering the structural information of the patents. Methods: We propose a three-stage mechanism for automatic categorization. In the first stage, we apply a pre-processing mechanism to reduce unwanted noise that can influence the categorization process. Such noise includes terms that have less structural meaning in the document. In the second stage, feature selection is conducted based on the term frequencies. Feature vectors are constructed from the structural information of the patent. In the third stage, classifications are conducted using a Random Forest (RF), Support Vector Machine (SVM), and Naïve Bayes (NB) classifier. Findings: It was found that the semantic structural information of a patent document is an important feature set in constructing the terms of a document for the categorization. The experimental results also show that feature reduction using Information Gain (IG) is beneficial for obtaining a higher accuracy rate in a reduced dimensional space. Applications: The results reveal the importance of the proposed method for automatic categorization of patent documents.Keywords: Classification, Feature Selection, Patent categorization, Structural information