Keyword Extraction Performance Analysis

In this paper we introduce Rapid Automatic Keyword extraction an unsupervised, domain independent and language independent method for extracting keywords from individual documents and compare this model with a graph based ranking algorithm(TextRank).In general TextRank consist of two unsupervised methods for both keyword and sentence extraction. Also we conduct a simple study regarding TextRank with the previously published methods

Data Mining is mainly used for storing and retrieve needed information. Reading and summarizing the contents of large entries of text into a small set of topics is difficult and time consuming for a human, so much that it becomes nearly impossible to accomplish with limited manpower as the size of the information grows. As a result, automated systems for retrieving information from large storage place, we are using the Concepts of Keyword and Keyword Extraction. It plays a vital role in the Web Based searches and normal searches of document or any important files etc in any organization. Keyword means word used in a text search, or a word in a text document that is used in an index to best describe the contents of the document. A word or phrase submitted to a search engine in an effort to locate relevant document or websites. Keyword Extraction means extracting the keyword from the implicit queries of the user. Extracting Keywords from the document helps the user to browse in fast manner, where time consumption is reduced. Keyword Extraction widely used many places like organization, college database, mail server, group discussion and internet etc.

International Journal of Advanced Computer Science and Applications, 2017

At each text there are a few keywords which provide important information about the content of that text. Since this limited set of words (keywords) is supposed to describe the total concept of a text (e.g. article, book), the correct choosing of keywords for a text plays an important role in the right representing of that text. Despite several efforts in this field, none of the so far published methods is accurate enough to elicit representative words for retrieving a vast variety of different texts. In this study, an unsupervised scheme is proposed which is independent on domain, language, structure and length of a text. The proposed method uses the words' frequency in conjunction with standard deviation of occurred location of words in text along with considering the conceptual relation of words. In the next stage, a secondary score is given to those selected keywords by the statistical criterion of TFISF in order to improve the basis method of TFIDF. Moreover, the proposed hybrid method does not remove the stopwords since they might be a part of bigram keywords while the similar approaches remove all stopwords at their first stage. Experimental results on the known SEMEVAL dataset imply the superiority of the proposed method in comparison with state-of-the-art schemes in terms of F-score and accuracy. Therefore, the introduced hybrid method can be considered as an alternative scheme for accurate keyword extraction.

2010 Workshops on Database and Expert Systems Applications, 2010

A common strategy to assign keywords to documents is to select the most appropriate words from the document text. One of the most important criteria for a word to be selected as keyword is its relevance for the text. The tf.idf score of a term is a widely used relevance measure. While easy to compute and giving quite satisfactory results, this measure does not take (semantic) relations between words into account. In this paper we study some alternative relevance measures that do use relations between words. They are computed by defining co-occurrence distributions for words and comparing these distributions with the document and the corpus distribution. We then evaluate keyword extraction algorithms defined by selecting different relevance measures. For two corpora of abstracts with manually assigned keywords, we compare manually extracted keywords with different automatically extracted ones. The results show that using word co-occurrence information can improve precision and recall over tf.idf.

International Journal of Information Retrieval Research, 2022

Retrieving keywords in a text is attracting researchers for a long time as it forms a base for many natural language applications like information retrieval, text summarization, document categorization etc. A text is a collection of words that represent the theme of the text naturally and to bring the naturalism under certain rules is itself a challenging task. In the present paper, the authors evaluate different spatial distribution based keyword extraction methods available in the literature on three standard scientific texts. The authors choose the first few high-frequency words for evaluation to reduce the complexity as all the methods are somehow based on frequency. The authors find that the methods are not providing good results particularly in the case of the first few retrieved words. Thus, the authors propose a new measure based on frequency, inverse document frequency, variance, and Tsallis entropy. Evaluation of different methods is done on the basis of precision, recall,...

arXiv (Cornell University), 2023

Automatic term extraction (ATE) is a Natural Language Processing (NLP) task that eases the effort of manually identifying terms from domain-specific corpora by providing a list of candidate terms. As units of knowledge in a specific field of expertise, extracted terms are not only beneficial for several terminographical tasks, but also support and improve several complex downstream tasks, e.g., information retrieval, machine translation, topic detection, and sentiment analysis. ATE systems, along with annotated datasets, have been studied and developed widely for decades, but recently we observed a surge in novel neural systems for the task at hand. Despite a large amount of new research on ATE, systematic survey studies covering novel neural approaches are lacking. We present a comprehensive survey of deep learning-based approaches to ATE, with a focus on Transformer-based neural models. The study also offers a comparison between these systems and previous ATE approaches, which were based on feature engineering and non-neural supervised learning algorithms.

Automatic keyword extraction is an important subfield of information extraction process. It is a difficult task, where numerous different techniques and resources have been proposed. In this paper, we propose a generic approach to extract keyword from documents using encyclopedic knowledge. Our two-step approach first relies on a classification step for identifying candidate keywords followed by a learning-to-rank method depending on a user-defined keyword profile to order the candidates. The novelty of our approach relies on i) the usage of the keyword profile ii) generic features derived from Wikipedia categories and not necessarily related to the document content. We evaluate our system on keyword datasets and corpora from standard evaluation campaign and show that our system improves the global process of keyword extraction.

TELKOMNIKA, 2023

In recent times, the trend of online shopping through e-commerce stores and websites has grown to a huge extent. Whenever a product is purchased on an e-commerce platform, people leave their reviews about the product. These reviews are very helpful for the store owners and the product’s manufacturers for the betterment of their work process as well as product quality. An automated system is proposed in this work that operates on two datasets D1 and D2 obtained from Amazon. After certain preprocessing steps, N-gram and word embedding-based features are extracted using term frequency-inverse document frequency (TF-IDF), bag of words (BoW) and global vectors (GloVe), and Word2vec, respectively. Four machine learning (ML) models support vector machines (SVM), logistic regression (RF), logistic regression (LR), multinomial Naïve Bayes (MNB), two deep learning (DL) models convolutional neural network (CNN), long-short term memory (LSTM), and standalone bidirectional encoder representations (BERT) are used to classify reviews as either positive or negative. The results obtained by the standard ML, DL models and BERT are evaluated using certain performance evaluation measures. BERT turns out to be the best-performing model in the case of D1 with an accuracy of 90% on features derived by word embedding models while the CNN provides the best accuracy of 97% upon word embedding features in the case of D2. The proposed model shows better overall performance on D2 as compared to D1.

2021

Keyword extraction is the task of identifying words (or multi-word expressions) that best describe a given document and serve in news portals to link articles of similar topics. In this work, we develop and evaluate our methods on four novel data sets covering less-represented, morphologically-rich languages in European news media industry (Croatian, Estonian, Latvian, and Russian). First, we perform evaluation of two supervised neural transformer-based methods, Transformer-based Neural Tagger for Keyword Identification (TNT-KID) and Bidirectional Encoder Representations from Transformers (BERT) with an additional Bidirectional Long Short-Term Memory Conditional Random Fields (BiLSTM CRF) classification head, and compare them to a baseline Term Frequency - Inverse Document Frequency (TF-IDF) based unsupervised approach. Next, we show that by combining the keywords retrieved by both neural transformer-based methods and extending the final set of keywords with an unsupervised TF-IDF b...

2019 International Conference of Advanced Informatics: Concepts, Theory and Applications (ICAICTA), 2019

Keyphrase extraction as a task to identify important words or phrases from a text, is a crucial process to identify main topics when analyzing texts from a social media platform. In our study, we focus on text written in Indonesia language taken from Twitter. Different from the original joint layer recurrent neural network (JRNN) with output of one sequence of keywords and using only word embedding, here we propose to modify the input layer of JRNN to extract more than one sequence of keywords by additional information of syntactical features, namely part of speech, named entity types, and dependency structures. Since JRNN in general requires a large amount of data as the training examples and creating those examples is expensive, we used a data augmentation method to increase the number of training examples. Our experiment had shown that our method outperformed the baseline methods. Our method achieved. 9597 in accuracy and. 7691 in F1.

Journal of Information Processing Systems, 2012

The paper presents three machine learning based keyphrase extraction methods that respectively use Decision Trees, Naïve Bayes, and Artificial Neural Networks for keyphrase extraction. We consider keyphrases as being phrases that consist of one or more words and as representing the important concepts in a text document. The three machine learning based keyphrase extraction methods that we use for experimentation have been compared with a publicly available keyphrase extraction system called KEA. The experimental results show that the Neural Network based keyphrase extraction method outperforms two other keyphrase extraction methods that use the Decision Tree and Naïve Bayes. The results also show that the Neural Network based method performs better than KEA.

Information Processing & Management, 2007

Keywords can be considered as condensed versions of documents and short forms of their summaries. In this paper, the problem of automatic extraction of keywords from documents is treated as a supervised learning task. A lexical chain holds a set of semantically related words of a text and it can be said that a lexical chain represents the semantic content of a portion of the text. Although lexical chains have been extensively used in text summarization, their usage for keyword extraction problem has not been fully investigated. In this paper, a keyword extraction technique that uses lexical chains is described, and encouraging results are obtained.

Bulletin of Electrical Engineering and Informatics

Automatic keyphrases extraction (AKE) is a principal task in natural language processing (NLP). Several techniques have been exploited to improve the process of extracting keyphrases from documents. Deep learning (DL) algorithms are the latest techniques used in prediction and extraction of keyphrases. DL is one of the most complex types of machine learning, relying on the use of artificial neural networks to make the machine follow the same decision-making path as the human brain. In this paper, we present a review of deep learning-based methods for AKE from documents, to highlight their contribution to improving keyphrase extraction performance. This review will also provide researchers with a collection of data and information on the mechanisms of deep learning algorithms in the AKE domain. This will allow them to solve problems encountered by AKE approaches and propose new methods for improving key-extraction performance.

Lecture Notes in Computer Science, 2018

In this work, we propose a lightweight approach for keyword extraction and ranking based on an unsupervised methodology to select the most important keywords of a single document. To understand the merits of our proposal, we compare it against RAKE, TextRank and SingleRank methods (three well-known unsupervised approaches) and the baseline TF.IDF, over four different collections to illustrate the generality of our approach. The experimental results suggest that extracting keywords from documents using our method results in a superior effectiveness when compared to similar approaches.

International Journal on Semantic Web and Information Systems, 2016

In this work the authors propose a novel Selectivity-Based Keyword Extraction (SBKE) method, which extracts keywords from the source text represented as a network. The node selectivity value is calculated from a weighted network as the average weight distributed on the links of a single node and is used in the procedure of keyword candidate ranking and extraction. The authors show that selectivity-based keyword extraction slightly outperforms an extraction based on the standard centrality measures: in/out-degree, betweenness and closeness. Therefore, they include selectivity and its modification – generalized selectivity as node centrality measures in the SBKE method. Selectivity-based extraction does not require linguistic knowledge as it is derived purely from statistical and structural information of the network. The experimental results point out that selectivity-based keyword extraction has a great potential for the collection-oriented keyword extraction task.