ISE-Hate: A benchmark corpus for inter-faith,sectarian, and ethnic hatred detection on social media in Urdu

Social media has become the most popular platform for free speech. This freedom of speech has given opportunities to the oppressed to raise their voice against injustices, but on the other hand, this has led to a disturbing trend of spreading hateful content of various kinds. Pakistan has been dealing with the issue of sectarian and ethnic violence for the last three decades and now due to freedom of speech, there is a growing trend of disturbing content about religion, sect, and ethnicity on social media. This necessitates the need for an automated system for the detection of controversial content on social media in Urdu which is the national language of Pakistan. The biggest hurdle that has thwarted the Urdu language processing is the scarcity of language resources, annotated datasets, and pretrained language models. In this study, we have addressed the problem of detecting Interfaith, Sectarian, and Ethnic hatred on social media in Urdu language using machine learning and deep learning techniques. In particular, we have: (1) developed and presented guidelines for annotating Urdu text with appropriate labels for two levels of classification, (2) developed a large dataset of 21,759 tweets using the developed guidelines and made it publicly available, and (3) conducted experiments to compare the performance of eight supervised machine learning and deep learning techniques, for the automated identification of hateful content. In the first step, experiments are performed for the hateful content detection as a binary classification task, and in the second step, the classification of Interfaith, Sectarian and Ethnic hatred detection is performed as a multiclass classification task. Overall, Bidirectional Encoder Representation from Transformers (BERT) proved to be the most effective technique for hateful content identification in Urdu tweets.     

A machine learning based framework to identify unseen classes in open-world text classification

Classical supervised machine learning (ML) follows the assumptions of closed-world learning. However, this assumption does not work in an open-world dynamic environment. Therefore, the automated systems must be able to discover and identify unseen instances. Open-world ML can deal with unseen instances and classes through a two-step process: (1) discover and classify unseen instances and (2) identify novel classes discovered in step (1). Most existing research on open-world machine learning (OWML) only focuses on step 1. However, performing step 2 is required to build intelligent systems. The proposed framework comprises three different but interconnected modules that discover and identify unseen classes. Our in-depth performance evaluation establishes that the proposed framework improves open accuracy by up to 8% compared to the state-of-the-art models.     

Revisiting Cross-document Structure Theory for multi-document discourse parsing

Multi-document discourse parsing aims to automatically identify the relations among textual spans from different texts on the same topic. Recently, with the growing amount of information and the emergence of new technologies that deal with many sources of information, more precise and efficient parsing techniques are required. The most relevant theory to multi-document relationship, Cross-document Structure Theory (CST), has been used for parsing purposes before, though the results had not been satisfactory. CST has received many critics because of its subjectivity, which may lead to low annotation agreement and, consequently, to poor parsing performance. In this work, we propose a refinement of the original CST, which consists in (i) formalizing the relationship definitions, (ii) pruning and combining some relations based on their meaning, and (iii) organizing the relations in a hierarchical structure. The hypothesis for this refinement is that it will lead to better agreement in the annotation and consequently to better parsing results. For this aim, it was built an annotated corpus according to this refinement and it was observed an improvement in the annotation agreement. Based on this corpus, a parser was developed using machine learning techniques and hand-crafted rules. Specifically, hierarchical techniques were used to capture the hierarchical organization of the relations according to the proposed refinement of CST. These two approaches were used to identify the relations among texts spans and to generate multi-document annotation structure. Results outperformed other CST parsers, showing the adequacy of the proposed refinement in the theory.     

远程监督实体关系抽取研究

【目的/意义】实体关系抽取是构建领域本体、知识图谱、开发问答系统的基础工作。远程监督方法将大规 模非结构化文本与已有的知识库实体对齐,自动标注训练样本,解决了有监督机器学习方法人工标注训练语料耗 时费力的问题,但也带来了数据噪声。【方法/过程】本文详细梳理了近些年远程监督结合深度学习技术,降低训练 样本噪声,提升实体关系抽取性能的方法。【结果/结论】卷积神经网络能更好的捕获句子局部、关键特征、长短时记 忆网络能更好的处理句子实体对远距离依赖关系,模型自动抽取句子词法、句法特征,注意力机制给予句子关键上 下文、单词更大的权重,在神经网络模型中融入先验知识能丰富句子实体对的语义信息,显著提升关系抽取性能。 【创新/局限】下一步的研究应考虑实体对重叠关系、实体对长尾语义关系的处理方法,更加全面的解决实体对关系 噪声问题。     

基于机器学习的扫描图书元数据自动抽取研究

在对纸本图书数字化加工过程中,元数据录入是必需的环节,然而手工录入工作量大、效率低,针对这一问题,提出了一种基于机器学习的扫描图书元数据自动获取方法。首先定义元数据的描述、管理和结构元素,然后以扫描页面的DjVu XML文档为数据源,分析页面的格式、结构等特征,以行作为初始特征向量,采用基于有监督的机器学习方法进行元数据抽取,实验表明该算法能够取得较高的准确率和召回率,能够显著的提高图书数字化的效率。     

Automatic extraction of bilingual word pairs using inductive chain learning in various languages

In this paper, we propose a new learning method for extracting bilingual word pairs from parallel corpora in various languages. In cross-language information retrieval, the system must deal with various languages. Therefore, automatic extraction of bilingual word pairs from parallel corpora with various languages is important. However, previous works based on statistical methods are insufficient because of the sparse data problem. Our learning method automatically acquires rules, which are effective to solve the sparse data problem, only from parallel corpora without any prior preparation of a bilingual resource (e.g., a bilingual dictionary, a machine translation system). We call this learning method Inductive Chain Learning (ICL). Moreover, the system using ICL can extract bilingual word pairs even from bilingual sentence pairs for which the grammatical structures of the source language differ from the grammatical structures of the target language because the acquired rules have the information to cope with the different word orders of source language and target language in local parts of bilingual sentence pairs. Evaluation experiments demonstrated that the recalls of systems based on several statistical approaches were improved through the use of ICL.     

Towards a big data framework for analyzing social media content

Modern companies generate value by digitalizing their services and products. Knowing what customers are saying about the firm through reviews in social media content constitutes a key factor to succeed in the big data era. However, social media data analysis is a complex discipline due to the subjectivity in text review and the additional features in raw data. Some frameworks proposed in the existing literature involve many steps that thereby increase their complexity. A two-stage framework to tackle this problem is proposed: the first stage is focused on data preparation and finding an optimal machine learning model for this data; the second stage relies on established layers of big data architectures focused on getting an outcome of data by taking most of the machine learning model of stage one. Thus, a first stage is proposed to analyze big and small datasets in a non-big data environment, whereas the second stage analyzes big datasets by applying the first stage machine learning model of. Then, a study case is presented for the first stage of the framework to analyze reviews of hotel-related businesses. Several machine learning algorithms were trained for two, three and five classes, with the best results being found for binary classification.     

Distributed search based on self-indexed compressed text

Query response times within a fraction of a second in Web search engines are feasible due to the use of indexing and caching techniques, which are devised for large text collections partitioned and replicated into a set of distributed-memory processors. This paper proposes an alternative query processing method for this setting, which is based on a combination of self-indexed compressed text and posting lists caching. We show that a text self-index (i.e., an index that compresses the text and is able to extract arbitrary parts of it) can be competitive with an inverted index if we consider the whole query process, which includes index decompression, ranking and snippet extraction time. The advantage is that within the space of the compressed document collection, one can carry out the posting lists generation, document ranking and snippet extraction. This significantly reduces the total number of processors involved in the solution of queries. Alternatively, for the same amount of hardware, the performance of the proposed strategy is better than that of the classical approach based on treating inverted indexes and corresponding documents as two separate entities in terms of processors and memory space.     

Content analysis of e-petitions with topic modeling: How to train and evaluate LDA models?

E-petitions have become a popular vehicle for political activism, but studying them has been difficult because efficient methods for analyzing their content are currently lacking. Researchers have used topic modeling for content analysis, but current practices carry some serious limitations. While modeling may be more efficient than manually reading each petition, it generally relies on unsupervised machine learning and so requires a dependable training and validation process. And so this paper describes a framework to train and validate Latent Dirichlet Allocation (LDA), the simplest and most popular topic modeling algorithm, using e-petition data. With rigorous training and evaluation, 87% of LDA-generated topics made sense to human judges. Topics also aligned well with results from an independent content analysis by the Pew Research Center, and were strongly associated with corresponding social events. Computer-assisted content analysts can benefit from our guidelines to supervise every process of training and evaluation of LDA. Software developers can benefit from learning the demands of social scientists when using LDA for content analysis. These findings have significant implications for developing LDA tools and assuring validity and interpretability of LDA content analysis. In addition, LDA topics can have some advantages over subjects extracted by manual content analysis by reflecting multiple themes expressed in texts, by extracting new themes that are not highlighted by human coders, and by being less prone to human bias.     

Automatic extraction of titles from general documents using machine learning

In this paper, we propose a machine learning approach to title extraction from general documents. By general documents, we mean documents that can belong to any one of a number of specific genres, including presentations, book chapters, technical papers, brochures, reports, and letters. Previously, methods have been proposed mainly for title extraction from research papers. It has not been clear whether it could be possible to conduct automatic title extraction from general documents. As a case study, we consider extraction from Office including Word and PowerPoint. In our approach, we annotate titles in sample documents (for Word and PowerPoint, respectively) and take them as training data, train machine learning models, and perform title extraction using the trained models. Our method is unique in that we mainly utilize formatting information such as font size as features in the models. It turns out that the use of formatting information can lead to quite accurate extraction from general documents. Precision and recall for title extraction from Word are 0.810 and 0.837, respectively, and precision and recall for title extraction from PowerPoint are 0.875 and 0.895, respectively in an experiment on intranet data. Other important new findings in this work include that we can train models in one domain and apply them to other domains, and more surprisingly we can even train models in one language and apply them to other languages. Moreover, we can significantly improve search ranking results in document retrieval by using the extracted titles.     

Web page title extraction and its application

This paper is concerned with automatic extraction of titles from the bodies of HTML documents (web pages). Titles of HTML documents should be correctly defined in the title fields by the authors; however, in reality they are often bogus. It is advantageous if we can automatically extract titles from HTML documents. In this paper, we take a supervised machine learning approach to address the problem. We first propose a specification on HTML titles, that is, a ‘definition’ on HTML titles. Next, we employ two learning methods to perform the task. In one method, we utilize features extracted from the DOM (direct object model) Tree; in the other method, we utilize features based on vision. We also combine the two methods to further enhance the extraction accuracy. Our title extraction methods significantly outperform the baseline method of using the lines in largest font size as title (22.6–37.4% improvements in terms of F1 score). As application, we consider web page retrieval. We use the TREC Web Track data for evaluation. We propose a new method for HTML documents retrieval using extracted titles. Experimental results indicate that the use of both extracted titles and title fields is almost always better than the use of title fields alone; the use of extracted titles is particularly helpful in the task of named page finding (25.1–30.3% improvements).     

A deep learning-based expert finding method to retrieve agile software teams from CQAs

Currently, many software companies are looking to assemble a team of experts who can collaboratively carry out an assigned project in an agile manner. The most ideal members for an agile team are T-shaped experts, who not only have expertise in one skill-area but also have general knowledge in a number of related skill-areas. Existing related methods have only used some heuristic non-machine learning models to form an agile team from candidates, while machine learning has been successful in similar tasks. In addition, they have only used the number of candidates’ documents in various skill-areas as a resource to estimate the candidates’ T-shaped knowledge to work in an agile team, while the content of their documents is also very important. To this end, we propose a multi-step method that rectifies the drawbacks mentioned. In this method, we first pick out the best possible candidates using a state-of-the-art model, then we re-estimate their relevant knowledge for working in the team with the help of a deep learning model, which uses the content of the candidates’ posts on StackOverflow. Finally, we select the best possible members for the given agile team from among these candidates using an integer linear programming model. We perform our experiments on two large datasets C# and Java, which comprise 2,217,366 and 2,320,883 posts from StackOverflow, respectively. On datasets C# and Java, our method selects, respectively, 68.6% and 55.2% of the agile team members from among T-shaped experts, while the best baseline method only selects, respectively, 49.1% and 40.2% of the agile team members from among T-shaped experts. In addition, the results show that our method outperforms the best baseline method by 8.1% and 11.4% in terms of F-measure on datasets C# and Java, respectively.     

Textual keyword extraction and summarization: State-of-the-art

With the advent of Web 2.0, there exist many online platforms that results in massive textual data production such as social networks, online blogs, magazines etc. This textual data carries information that can be used for betterment of humanity. Hence, there is a dire need to extract potential information out of it. This study aims to present an overview of approaches that can be applied to extract and later present these valuable information nuggets residing within text in brief, clear and concise way. In this regard, two major tasks of automatic keyword extraction and text summarization are being reviewed. To compile the literature, scientific articles were collected using major digital computing research repositories. In the light of acquired literature, survey study covers early approaches up to all the way till recent advancements using machine learning solutions. Survey findings conclude that annotated benchmark datasets for various textual data-generators such as twitter and social forms are not available. This scarcity of dataset has resulted into relatively less progress in many domains. Also, applications of deep learning techniques for the task of automatic keyword extraction are relatively unaddressed. Hence, impact of various deep architectures stands as an open research direction. For text summarization task, deep learning techniques are applied after advent of word vectors, and are currently governing state-of-the-art for abstractive summarization. Currently, one of the major challenges in these tasks is semantic aware evaluation of generated results.     

The impact of deep learning on document classification using semantically rich representations

This paper presents a semantically rich document representation model for automatically classifying financial documents into predefined categories utilizing deep learning. The model architecture consists of two main modules including document representation and document classification. In the first module, a document is enriched with semantics using background knowledge provided by an ontology and through the acquisition of its relevant terminology. Acquisition of terminology integrated to the ontology extends the capabilities of semantically rich document representations with an in depth-coverage of concepts, thereby capturing the whole conceptualization involved in documents. Semantically rich representations obtained from the first module will serve as input to the document classification module which aims at finding the most appropriate category for that document through deep learning. Three different deep learning networks each belonging to a different category of machine learning techniques for ontological document classification using a real-life ontology are used.Multiple simulations are carried out with various deep neural networks configurations, and our findings reveal that a three hidden layer feedforward network with 1024 neurons obtain the highest document classification performance on the INFUSE dataset. The performance in terms of F1 score is further increased by almost five percentage points to 78.10% for the same network configuration when the relevant terminology integrated to the ontology is applied to enrich document representation. Furthermore, we conducted a comparative performance evaluation using various state-of-the-art document representation approaches and classification techniques including shallow and conventional machine learning classifiers.     

E-word of mouth sentiment analysis for user behavior studies

Nowadays, online word-of-mouth has an increasing impact on people's views and decisions, which has attracted many people's attention.The classification and sentiment analyse in online consumer reviews have attracted significant research concerns. In this thesis, we propose and implement a new method to study the extraction and classification of online dating services(ODS)’s comments. Different from traditional emotional analysis which mainly focuses on product attribution, we attempted to infer and extract the emotion concept of each emotional reviews by introducing social cognitive theory. In this study, we selected 4,300 comments with extremely negative/positive emotions published on dating websites as a sample, and used three machine learning algorithms to analyze emotions. When testing and comparing the efficiency of user's behavior research, we use various sentiment analysis, machine learning techniques and dictionary-based sentiment analysis. We found that the combination of machine learning and lexicon-based method can achieve higher accuracy than any type of sentiment analysis. This research will provide a new perspective for the task of user behavior.     

Information retrieval and machine learning for probabilistic schema matching

Schema matching is the problem of finding correspondences (mapping rules, e.g. logical formulae) between heterogeneous schemas e.g. in the data exchange domain, or for distributed IR in federated digital libraries. This paper introduces a probabilistic framework, called sPLMap, for automatically learning schema mapping rules, based on given instances of both schemas. Different techniques, mostly from the IR and machine learning fields, are combined for finding suitable mapping candidates. Our approach gives a probabilistic interpretation of the prediction weights of the candidates, selects the rule set with highest matching probability, and outputs probabilistic rules which are capable to deal with the intrinsic uncertainty of the mapping process. Our approach with different variants has been evaluated on several test sets.     

Text classification from unlabeled documents with bootstrapping and feature projection techniques

Many machine learning algorithms have been applied to text classification tasks. In the machine learning paradigm, a general inductive process automatically builds a text classifier by learning, generally known as supervised learning. However, the supervised learning approaches have some problems. The most notable problem is that they require a large number of labeled training documents for accurate learning. While unlabeled documents are easily collected and plentiful, labeled documents are difficultly generated because a labeling task must be done by human developers. In this paper, we propose a new text classification method based on unsupervised or semi-supervised learning. The proposed method launches text classification tasks with only unlabeled documents and the title word of each category for learning, and then it automatically learns text classifier by using bootstrapping and feature projection techniques. The results of experiments showed that the proposed method achieved reasonably useful performance compared to a supervised method. If the proposed method is used in a text classification task, building text classification systems will become significantly faster and less expensive.     

面向网络虚假医疗信息的识别模型构建研究——一种基于预训练的BERT模型

【 目的/意义】解决获取虚假网络医疗信息数据集时专业知识不足的问题,帮助在小样本领域构建虚假网络 医疗信息识别模型。【方法/过程】本文提出一种基于权威辟谣信息转化提取构建网络虚假医疗信息数据集的思路, 并依次构建传统机器学习模型、CNN模型和BERT模型进行分类识别。【结果/结论】结果表明,基于辟谣信息能够 实现以较低成本、不依赖专家标注构建虚假医疗信息数据集。通过对比实验发现,基于微博数据预训练的 BERT 模型准确率为 95.91%,F1值为 94.57%,相比于传统机器学习模型和 CNN模型提升分别接近 6%和 4%,表明本文构 建的基于预训练的BERT模型在网络虚假医疗信息识别任务上取得了更好的效果。【创新/局限】本文提出的方法能 以较低成本建立专业领域的虚假信息数据集,所构建的BERT虚假医疗信息识别模型在小样本领域也具有实用价 值,但在数据集规模、深度学习模型对比、模型性能评价指标等方面还有待拓展与延伸。     

Information extraction with automatic knowledge expansion

POSIE (POSTECH Information Extraction System) is an information extraction system which uses multiple learning strategies, i.e., SmL, user-oriented learning, and separate-context learning, in a question answering framework. POSIE replaces laborious annotation with automatic instance extraction by the SmL from structured Web documents, and places the user at the end of the user-oriented learning cycle. Information extraction as question answering simplifies the extraction procedures for a set of slots. We introduce the techniques verified on the question answering framework, such as domain knowledge and instance rules, into an information extraction problem. To incrementally improve extraction performance, a sequence of the user-oriented learning and the separate-context learning produces context rules and generalizes them in both the learning and extraction phases. Experiments on the “continuing education” domain initially show that the F1-measure becomes 0.477 and recall 0.748 with no user training. However, as the size of the training documents grows, the F1-measure reaches beyond 0.75 with recall 0.772. We also obtain F-measure of about 0.9 for five out of seven slots on “job offering” domain.     

Automatic word spacing of erroneous sentences in mobile devices with limited hardware resources

With the rapid evolution of the mobile environment, the demand for natural language applications on mobile devices is increasing. This paper proposes an automatic word spacing system, the first step module of natural language processing (NLP) for many languages with their own word spacing rules, that is designed for mobile devices with limited hardware resources. The proposed system uses two stages. In the first stage, it preliminarily corrects word spacing errors by using a modified hidden Markov model based on character unigrams. In the second stage, the proposed system re-corrects the miscorrected word spaces by using lexical rules based on character bigrams or longer combinations. By using this hybrid method, the proposed system improves the robustness against unknown word patterns, reduces memory usage, and increases accuracy. To evaluate the proposed system in a realistic mobile environment, we constructed a mobile-style colloquial corpus using a simple simulation method. In experiments with a commercial mobile phone, the proposed system showed good performances (a response time of 0.20 s per sentence, a memory usage of 2.04 MB, and an accuracy of 92–95%) in the various evaluation measures.