Keyphrase Extraction Techniques Research Articles

Comparative Analysis on Automatic Keyphrase Extraction (AKPE) Techniques

Abstract— Extracting Keyphrase from the large corpus manually is a tedious task. To overcome this challenging task Automatic Keyphrase Extraction (AKPE) techniques have been used for Keyphrase extraction. Keyphrase extraction is the task of automatically selecting a small set of phrases that best describe a given free text document. Keyphrase constitutes a succinct conceptual summary of a document, which is very useful in digital information management systems for semantic indexing, faceted search, document clustering, and classification. This paper contains a brief description of automatic Keyphrase extraction techniques which consist of various algorithms and also a comparative study on automotive Keyphrase extraction techniques such as KEA, TF-IFD, RAKE, TextRank. Keywords— Keyphrase Extraction, Techniques, Comparison, Keyphrase Extraction Algorithm, Term Frequency- Inverse Frequency Document, Rapid Automatic Keyphrase Extraction, TextRank.

Contextual topic discovery using unsupervised keyphrase extraction and hierarchical semantic graph model

Recent technological advancements have led to a significant increase in digital documents. A document’s key information is generally represented by the keyphrases that provide the abstract description contained therein. With traditional keyphrase techniques, however, it is difficult to identify relevant information based on context. Several studies in the literature have explored graph-based unsupervised keyphrase extraction techniques for automatic keyphrase extraction. However, there is only limited existing work that embeds contextual information for keyphrase extraction. To understand keyphrases, it is essential to grasp both the concept and the context of the document. Hence, a hybrid unsupervised keyphrase extraction technique is presented in this paper called ContextualRank, which embeds contextual information such as sentences and paragraphs that are relevant to keyphrases in the keyphrase extraction process. We propose a hierarchical topic modeling approach for topic discovery based on aggregating the extracted keyphrases from ContextualRank. Based on the evaluation on two short-text datasets and one long-text dataset, ContextualRank obtains remarkable improvements in performance over other baselines in the short-text datasets.

Automatic Construction of Educational Knowledge Graphs: A Word Embedding-Based Approach

Knowledge graphs (KGs) are widely used in the education domain to offer learners a semantic representation of domain concepts from educational content and their relations, termed as educational knowledge graphs (EduKGs). Previous studies on EduKGs have incorporated concept extraction and weighting modules. However, these studies face limitations in terms of accuracy and performance. To address these challenges, this work aims to improve the concept extraction and weighting mechanisms by leveraging state-of-the-art word and sentence embedding techniques. Concretely, we enhance the SIFRank keyphrase extraction method by using SqueezeBERT and we propose a concept-weighting strategy based on SBERT. Furthermore, we conduct extensive experiments on different datasets, demonstrating significant improvements over several state-of-the-art keyphrase extraction and concept-weighting techniques.

A New Unsupervised Technique to Analyze the Centroid and Frequency of Keyphrases from Academic Articles

Automated keyphrase extraction is crucial for extracting and summarizing relevant information from a variety of publications in multiple domains. However, the extraction of good-quality keyphrases and the summarising of information to a good standard have become extremely challenging in recent research because of the advancement of technology and the exponential development of digital sources and textual information. Because of this, the usage of keyphrase features for keyphrase extraction techniques has recently gained tremendous popularity. This paper proposed a new unsupervised region-based keyphrase centroid and frequency analysis technique, named the KCFA technique, for keyphrase extraction as a feature. Data/datasets collection, data pre-processing, statistical methodologies, curve plotting analysis, and curve fitting technique are the five main processes in the proposed technique. To begin, the technique collects multiple datasets from diverse sources, which are then input into the data pre-processing step by utilizing some text pre-processing processes. Afterward, the region-based statistical methodologies receive the pre-processed data, followed by the curve plotting examination and, lastly, the curve fitting technique. The proposed technique is then tested and evaluated using ten (10) best-accessible benchmark datasets from various disciplines. The proposed approach is then compared to our available methods to demonstrate its efficacy, advantages, and importance. Lastly, the results of the experiment show that the proposed method works well to analyze the centroid and frequency of keyphrases from academic articles. It provides a centroid of 706.66 and a frequency of 38.95% in the first region, 2454.21 and 7.98% in the second region, for a total frequency of 68.11%.

KEST: A graph-based keyphrase extraction technique for tweets summarization using Markov Decision Process

Multi-document summarization finds its application in many downstream information retrieval and natural language processing tasks. In the light of recent developments in social media data mining, Tweet summarization has emerged as a fundamental task of automatically detecting important keyphrases from a set of Tweets about current happenings. In the existing literature, the graph-based keyphrase extraction techniques are well-established unsupervised algorithms to capture summaries from dynamically evolving data. We argue that the traditional multi-tweet summarization technique may or may not capture user’s interest-specific keyphrases during tweet summarization. The nature of user-generated factual short-text is different from well-formed descriptive and perceptual long-text due to their repetitive nature. In this context, we introduce a simple yet effective interest-specific keyphrase extraction technique for tweet summarization as KEST: Key Extraction for Summarization of Tweets using Markov Decision Process (MDP). In this research work, we generate a path as evolving chain of highly interconnected words from sub-components in graph of words. We evaluate the effectiveness of our computationally, inexpensive, graph-based, abstractive keyphrase extraction approach over two datasets which we make publicly available.

Keyword Extraction for Medium‐Sized Documents Using Corpus‐Based Contextual Semantic Smoothing

Keyword extraction refers to the process of selecting most significant, relevant, and descriptive terms as keywords, which are present inside a single document. Keyword extraction has major applications in the information retrieval domain, such as analysis, summarization, indexing, and search, of documents. In this paper, we present a novel supervised technique for extraction of keywords from medium‐sized documents, namely Corpus‐based Contextual Semantic Smoothing (CCSS). CCSS extends the concept of Contextual Semantic Smoothing (CSS), which considers term usage patterns in similar texts to improve term relevance information. We introduce four more features beyond CSS as our novel contributions in this work. We systematically compare the performance of CCSS with other techniques, when implemented over INSPEC dataset, where CCSS outperforms all state‐of‐the‐art keyphrase extraction techniques presented in the literature.

Using Social Media to Analyze Public Concerns and Policy Responses to COVID-19 in Hong Kong

The outbreak of COVID-19 has caused huge economic and societal disruptions. To fight against the coronavirus, it is critical for policymakers to take swift and effective actions. In this article, we take Hong Kong as a case study, aiming to leverage social media data to support policymakers’ policy-making activities in different phases. First, in the agenda setting phase, we facilitate policymakers to identify key issues to be addressed during COVID-19. In particular, we design a novel epidemic awareness index to continuously monitor public discussion hotness of COVID-19 based on large-scale data collected from social media platforms. Then we identify the key issues by analyzing the posts and comments of the extensively discussed topics. Second, in the policy evaluation phase, we enable policymakers to conduct real-time evaluation of anti-epidemic policies. Specifically, we develop an accurate Cantonese sentiment classification model to measure the public satisfaction with anti-epidemic policies and propose a keyphrase extraction technique to further extract public opinions. To the best of our knowledge, this is the first work which conducts a large-scale social media analysis of COVID-19 in Hong Kong. The analytical results reveal some interesting findings: (1) there is a very low correlation between the number of confirmed cases and the public discussion hotness of COVID-19. The major public concern in the early stage is the shortage of anti-epidemic items. (2) The top-3 anti-epidemic measures with the greatest public satisfaction are daily press conference on COVID-19 updates, border closure, and social distancing rules.

TripleRank: An unsupervised keyphrase extraction algorithm

Automatic keyphrase extraction algorithms aim to identify words and phrases that contain the core information in documents. As online scholarly resources have become widespread in recent years, better keyphrase extraction techniques are required to improve search efficiency. We present two features, keyphrase semantic diversity and keyphrase coverage, to overcome limitations of existing methods for unsupervised keyphrase extraction. Keyphrase semantic diversity is the degree of semantic variety in the extraction result, which is introduced to avoid extracting synonym phrases that contain the same high-score candidate. Keyphrase coverage refers to candidates’ representativeness of other words in documents. We propose an unsupervised keyphrase extraction method called TripleRank, which evaluates three features: word position (a sensitive feature for academic documents) and two innovative features mentioned above. The architecture of TripleRank includes three sub-models that score the three features and a summing model. Though involving multiple models, there is no typical iteration process in TripleRank; hence, the computational cost is relatively low. TripleRank has led the experiment results on four academic datasets compared to four state-of-the-art baseline models, which confirmed the influence of keyphrase semantic diversity and keyphrase coverage and proved the efficiency of our method.

Diverse feature set based Keyphrase extraction and indexing techniques

The internet changed the way that people communicate, and this has led to a vast amount of Text that is available in electronic format. It includes things like e-mail, technical and scientific reports, tweets, physician notes and military field reports. Providing key-phrases for these extensive text collections thus allows users to grab the essence of the lengthy contents quickly and helps to locate information with high efficiency. While designing a Keyword Extraction and Indexing system, it is essential to pick unique properties, called features. In this article, we proposed different unsupervised keyword extraction approaches, which is independent of the structure, size and domain of the documents. The proposed method relies on the novel and cognitive inspired set of standard, phrase, word embedding and external knowledge source features. The individual and selected feature results are reported through experimentation on four different datasets viz. SemEval, KDD, Inspec, and DUC. The selected (feature selection) and word embedding based features are the best features set to be used for keywords extraction and indexing among all mentioned datasets. That is the proposed distributed word vector with additional knowledge improves the results significantly over the use of individual features, combined features after feature selection and state-of-the-art. After successfully achieving the objective of developing various keyphrase extraction methods we also experimented it for document classification task.

TOP-Rank: A TopicalPostionRank for Extraction and Classification of Keyphrases in Text

Keyphrase extraction is the task of extracting the most important phrases from a document. Automatic keyphrase extraction attempts to itemize a document content as metainformation and facilitate efficient information retrieval. In this paper we propose TOP-Rank, an approach for keyphrase extraction and keyphrase classification. For keyphrase extraction, we build an approach based on the position of keyphrases in the document and expand it with topical ranking of keyphrases. In particular, keyphrase extraction technique analyzes the documents and extracts keyphrases from the document by giving a higher rank to topical phrases. After keyphrase extraction, we classify keyphrases as process, material and task. Our evaluation on diverse datasets shows that TOP-Rank achieves F1-score of 0.73 for keyphrase classification improving upon state-of-the-art methods by a huge margin.

TeKET: a Tree-Based Unsupervised Keyphrase Extraction Technique

Automatic keyphrase extraction techniques aim to extract quality keyphrases for higher level summarization of a document. Majority of the existing techniques are mainly domain-specific, which require application domain knowledge and employ higher order statistical methods, and computationally expensive and require large train data, which is rare for many applications. Overcoming these issues, this paper proposes a new unsupervised keyphrase extraction technique. The proposed unsupervised keyphrase extraction technique, named TeKET or Tree-based Keyphrase Extraction Technique, is a domain-independent technique that employs limited statistical knowledge and requires no train data. This technique also introduces a new variant of a binary tree, called KeyPhrase Extraction (KePhEx) tree, to extract final keyphrases from candidate keyphrases. In addition, a measure, called Cohesiveness Index or CI, is derived which denotes a given node’s degree of cohesiveness with respect to the root. The CI is used in flexibly extracting final keyphrases from the KePhEx tree and is co-utilized in the ranking process. The effectiveness of the proposed technique and its domain and language independence are experimentally evaluated using available benchmark corpora, namely SemEval-2010 (a scientific articles dataset), Theses100 (a thesis dataset), and a German Research Article dataset, respectively. The acquired results are compared with other relevant unsupervised techniques belonging to both statistical and graph-based techniques. The obtained results demonstrate the improved performance of the proposed technique over other compared techniques in terms of precision, recall, and F1 scores.

The structure of word co-occurrence network for microblogs

The study of structure and dynamics of complex networks is seeking attention of academic researchers and practitioners in recent years. Although Word Co-occurrence Networks (WCN) have been studied for different languages, yet there is the need to study the structure of WCN for microblogs due to the presence of ill-formed and unstructured data. In this research article, existing WCN based applications have been explored and microblog WCN have been analysed for multiple key parameters to uncover the hidden patterns. The key parameters studied for microblogs WCN are scale-free property, small world feature, hierarchical organization, assortativity and spectral analysis. The twitter FSD dataset has been used for experimental results and evaluation. Different mathematical, statistical and graphical interpretations proved that the microblog WCN are different from the WCN of traditional well-formed text. The robustness of the key parameters of microblogs WCN have been explored for keyphrase extraction from domain specific set of microblogs. The baseline methods used for comparisons are TextRank, TopicRank, and NErank. Extensive experiments over standard public dataset proved that the proposed keyphrase extraction technique outperforms the existing techniques in terms of precision, recall, F-measure, and ROUGE scores.

A Flexible Keyphrase Extraction Technique for Academic Literature

A keyphrase extraction technique endeavors to extract quality keyphrases from a given document, which provide a high-level summary of that document. Except statistical keyphrase extraction approaches, all other approaches are either domain-dependent or require a sufficient amount of training data, which are rare at present. Therefore, in this paper, a new tree-based automatic keyphrase extraction technique is proposed, which is domain-independent and employs nominal statistical knowledge; but no train data are required. The proposed technique extracts a quality keyphrase through forming a tree from a candidate keyphrase; and later, it is expanded or shrunk or remained in the same state depending on other similar candidate keyphrases. At the end, keyphrases are extracted from the resultant trees based on a value, µ (which is the Maturity Index (MI) of a node in the tree), which enables flexibility in this process. A small µ value would yield many and/or lengthy keyphrases (greedy approach); whereas, a large µ value would yield lower and/or abbreviated keyphrases (conservative approach). Thereby, a user can extract his/her desired-level of keyphrases through tuning µ value. The effectiveness of the proposed technique is evaluated on an actual corpus, and compared with Rapid Automatic Keyphrase Extraction (RAKE) technique.

Performance Analysis of Graph based Keyphrase Extraction metrics for uncertain User-generated data

The keyphrase extraction, text summarization and word collocation algorithms have been different forms of summarizing the textual information. However, the unstructured user-generated data over social media is short-text and ill-formed. The use of existing traditional keyphrase extraction algorithm for unstructured data may give indifferent results. This problem has received minimal attention from academic researchers and practitioners. Thus in this article, the experimental evaluation of existing keyphrase extraction techniques have been performed. The performance evaluation have been done using ROUGE-1, ROUGE-2 and ROUGE-L score. The evaluation measures gives useful insights about keyphrase extraction techniques for social media data. The application of keyphrase extraction include topic detection, identifying events, trend detection and microblogs summarization.

Subject-based retrieval of scientific documents, case study

PurposeThe purpose of this paper is to introduce an approach for retrieving a set of scientific articles in the field of Information Technology (IT) from a scientific database such as Web of Science (WoS), to apply scientometrics indices and compare them with other fields.Design/methodology/approachThe authors propose to apply a statistical classification-based approach for extracting IT-related articles. In this approach, first, a probabilistic model is introduced to model the subject IT, using keyphrase extraction techniques. Then, they retrieve IT-related articles from all Iranian papers in WoS, based on a Bayesian classification scheme. Based on the probabilistic IT model, they assign an IT membership probability for each article in the database, and then they retrieve the articles with highest probabilities.FindingsThe authors have extracted a set of IT keyphrases, with 1,497 terms through the keyphrase extraction process, for the probabilistic model. They have evaluated the proposed retrieval approach with two approaches: the query-based approach in which the articles are retrieved from WoS using a set of queries composed of limited IT keywords, and the research area-based approach which is based on retrieving the articles using WoS categorizations and research areas. The evaluation and comparison results show that the proposed approach is able to generate more accurate results while retrieving more articles related to IT.Research limitations/implicationsAlthough this research is limited to the IT subject, it can be generalized for any subject as well. However, for multidisciplinary topics such as IT, special attention should be given to the keyphrase extraction phase. In this research, bigram model is used; however, one can extend it to tri-gram as well.Originality/valueThis paper introduces an integrated approach for retrieving IT-related documents from a collection of scientific documents. The approach has two main phases: building a model for representing topic IT, and retrieving documents based on the model. The model, based on a set of keyphrases, extracted from a collection of IT articles. However, the extraction technique does not rely on Term Frequency-Inverse Document Frequency, since almost all of the articles in the collection share a set of same keyphrases. In addition, a probabilistic membership score is defined to retrieve the IT articles from a collection of scientific articles.

RankUp: Enhancing graph-based keyphrase extraction methods with error-feedback propagation

In recent years, unsupervised, graph-based ranking algorithms have been successfully applied to keyphrase extraction tasks. These methods have the advantage of taking into account global information, such as text structure and relations between words, phrases, and sentences, rather than relying solely on local, vertex-specific information. Graph-based approaches for keyphrase extraction, however, have a particular drawback, which comes from their frequency-based analysis methods. The weakness is that many common, less relevant terms may get a higher ranking, particularly in short articles. The converse situation also occurs, where less common (and possibly more relevant) terms obtain lower rankings. We propose an unsupervised method—RankUp—that enhances graph-based keyphrase extraction approaches by applying an error-feedback mechanism similar to the concept of backpropagation. Experiments have been performed on almost 3,300 short texts from a variety of domains. Our experiments show that error-feedback propagation can boost the quality of keyphrases in graph-based keyphrase extraction techniques.

Creation and evaluation of large keyphrase extraction collections with multiple opinions

While several automatic keyphrase extraction (AKE) techniques have been developed and analyzed, there is little consensus on the definition of the task and a lack of overview of the effectiveness of different techniques. Proper evaluation of keyphrase extraction requires large test collections with multiple opinions, currently not available for research. In this paper, we (i) present a set of test collections derived from various sources with multiple annotations (which we also refer to as opinions in the remained of the paper) for each document, (ii) systematically evaluate keyphrase extraction using several supervised and unsupervised AKE techniques, (iii) and experimentally analyze the effects of disagreement on AKE evaluation. Our newly created set of test collections spans different types of topical content from general news and magazines, and is annotated with multiple annotations per article by a large annotator panel. Our annotator study shows that for a given document there seems to be a large disagreement on the preferred keyphrases, suggesting the need for multiple opinions per document. A first systematic evaluation of ranking and classification of keyphrases using both unsupervised and supervised AKE techniques on the test collections shows a superior effectiveness of supervised models, even for a low annotation effort and with basic positional and frequency features, and highlights the importance of a suitable keyphrase candidate generation approach. We also study the influence of multiple opinions, training data and document length on evaluation of keyphrase extraction. Our new test collection for keyphrase extraction is one of the largest of its kind and will be made available to stimulate future work to improve reliable evaluation of new keyphrase extractors.

Bridging the gap: effect of text query reformulation in multimodal retrieval

Multimodal Retrieval provides new paradigms and methods aimed at effectively searching through the enormous volume of data. Multimodal retrieval is a well studied problem often used in image retrieval. Most of the existing works in image retrieval under the pretext of multimodality stress on bridging the semantic gap by using both textual and visual features. In this paper, we use relevance feedback from the user-generated documents associated with the images for expanding textual query and study its effect on both image and text retrieval. We employ a topic decomposition based keyphrase extraction technique to expand the textual queries. Our results articulate the fact that an insightful textual query expansion always improves retrieval performance for both textual or image retrieval. Also, we adopt optimum weight learning scheme to combine the modalities in a privileged way. We perform a comparative study with two well established keyphrase extraction techniques which are used for textual query expansion. A detailed set of experiments on a standard real world dataset is also carried out for the same.

Multimodal Retrieval using Mutual Information based Textual Query Reformulation

Multimodal Retrieval is a well-established approach for image retrieval. Usually, images are accompanied by text caption along with associated documents describing the image. Textual query expansion as a form of enhancing image retrieval is a relatively less explored area. In this paper, we first study the effect of expanding textual query on both image and its associated text retrieval. Our study reveals that judicious expansion of textual query through keyphrase extraction can lead to better results, either in terms of text-retrieval or both image and text-retrieval. To establish this, we use two well-known keyphrase extraction techniques based on tf-idf and KEA. While query expansion results in increased retrieval efficiency, it is imperative that the expansion be semantically justified. So, we propose a graph-based keyphrase extraction model that captures the relatedness between words in terms of both mutual information and relevance feedback. Most of the existing works have stressed on bridging the semantic gap by using textual and visual features, either in combination or individually. The way these text and image features are combined determines the efficacy of any retrieval. For this purpose, we adopt Fisher-LDA to adjudge the appropriate weights for each modality. This provides us with an intelligent decision-making process favoring the feature set to be infused into the final query. Our proposed algorithm is shown to supersede the previously mentioned keyphrase extraction algorithms for query expansion significantly. A rigorous set of experiments performed on ImageCLEF-2011 Wikipedia Retrieval task dataset validates our claim that capturing the semantic relation between words through Mutual Information followed by expansion of a textual query using relevance feedback can simultaneously enhance both text and image retrieval.

Keyword and Keyphrase Extraction Techniques: A Literature Review

In this paper we present a survey of various techniques available in text mining for keyword and keyphrase extraction.