DOCSLIB.ORG

Out-Of-Domain Detection for Low-Resource Text Classification

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Out-Of-Domain Detection for Low-Resource Text Classification Out-of-Domain Detection for Low-Resource Text Classification Tasks Ming Tan ∗y Yang Yu ∗† Haoyu Wang ∗† Dakuo Wang z Saloni Potdar y Shiyu Chang z Mo Yu z y IBM Watson z IBM Research Abstract Intent Label Example Help List List what you can help me Out-of-domain (OOD) detection for low- with. resource text classification is a realistic but Watson, I need your help understudied task. The goal is to detect the Schedule Appointment Can you book a cleaning with my dentist for me? OOD cases with limited in-domain (ID) train- Can you schedule my den- ing data, since we observe that training data tist’s appointment? is often insufficient in machine learning appli- End Meeting You can end the meeting cations. In this work, we propose an OOD- now resistant Prototypical Network to tackle this Meeting is over ··· ··· zero-shot OOD detection and few-shot ID classification task. Evaluation on real-world OOD utterances My birthday is coming! datasets show that the proposed solution out- blah blah... performs state-of-the-art methods in zero-shot OOD detection task, while maintaining a com- Table 1: A few-shot ID training set for a conversa- petitive performance on ID classification task. tion service for teleconference management, with OOD testing examples. 1 Introduction Text classification tasks in real-world applications son Assistant1. For example, Table1 shows some often consists of 2 components- In-Doman (ID) of the utterances a chat-bot builder provided for classification and Out-of-Domain (OOD) detec- training. Each class may only have less than 20 tion components (Liao et al., 2018; Kim and Kim, training utterances, due to the high cost of man- 2018; Shu et al., 2017; Shamekhi et al., 2018). ual labelling by domain experts. Meanwhile, the ID classification refers to classifying a user’s in- user also expects the service to effectively reject put with a label that exists in the training data, and irrelevant queries (as shown at the bottom of Table OOD detection refers to designate a special OOD 1). The challenge of OOD detection is reflected tag to the input when it does not belong to any by the undefined in-domain boundary. Although of the labels in the ID training dataset (Dai et al., one can provide a certain amount of OOD sam- 2007). Recent state-of-the-art deep learning (DL) ples to build a binary classifier for OOD detec- approaches for OOD detection and ID classifica- tion, such samples may not efficiently reflect the tion task often require massive amounts of ID or infinite OOD space. Recent approaches, such as OOD labeled data (Kim and Kim, 2018). In re- (Shu et al., 2017), make remarkable progress on ality, many applications have very limited ID la- OOD detection with only ID examples. However, beled data (i.e., few-shot learning) and no OOD such condition on ID data cannot be satisfied by labeled data (i.e., zero-shot learning). Thus, ex- the few-shot scenario presented in Table1. isting methods for OOD detection do not perform This work aims to build a model that can de- well in this setting. tect OOD inputs with limited ID data and zero One such application is the intent classification OOD training data, while classifying ID inputs for conversational AI services, such as IBM Wat- with a high accuracy. Learning similarities with ∗ Equal contributions from the corresponding authors: fmingtan,yu,[email protected]. 1https://www.ibm.com/cloud/watson-assistant/ 3566 Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 3566–3572, Hong Kong, China, November 3–7, 2019. c 2019 Association for Computational Linguistics the meta-learning strategy (Vinyals et al., 2016) Task !" Supporting Sets Prototypical Vectors Label-1 9"$ /,:*/; ℒ"$, ℒ%& example-1 has been proposed to deal with the problem of example-2 example-3 limited training examples for each label (few-shot ℒ''( Label-2 9"$ /,:*/< Prototypical Net- example-4 learning). In this line of work, example-5 works (Snell et al., 2017), which was originally 9"$ Label-3 /,:*/= introduced for few-shot image classification, has example-6 4$5'(*1 4(7) proven to be promising for few-shot ID text clas- 4$5'(*1 4 7 "$ '3& Random selection I) *+,-./* + OOD example +2 ,>% − .''/ " sification (Yu et al., 2018). However the usage of 01'- !" 01'- !2(not shown here) prototypical network for OOD detection is unex- Figure 1: Model overview: the model maximizes likeli- plored in this regard. hood of the ground-truth of ID example, minimizes dis- To the best of our knowledge, this work is the tance between ID example and ground-truth, and max- first one to adopt a meta-learning strategy to train imizes the distance of OOD example and all ID labels. an OOD-Resistant Prototypical Network for si- studies suggests that few-shot learning is promis- multaneously detecting OOD examples and classi- ing in the text domain, including text classifica- fying ID examples. The contributions of this work tion (Yu et al., 2018; Jiang et al., 2018), relation are two-fold: 1) Unified solution using a proto- extraction (Han et al., 2018), link prediction in typical network model which can detect OOD in- knowledge bases (Xiong et al., 2018) and fine- stances and classify ID instances in a real-world grained entity typing (Xiong et al., 2019), and we low-resource scenario. 2) Experiments and analy- put it to test with the OOD detection task. sis on two datasets prove that the proposed model outperforms previous work on the OOD detection 3 Approach task, while maintaining a state-of-the-art ID clas- sification performance. In this paper, we target solving the zero-shot OOD detection problem for a few-shot meta-test dataset 2 Related Work D = (Dtrain;Dtest) by training a transferable Out-of-Domain Detection Existing methods prototypical network model from large-scale in- often formulate the OOD task as a one-class clas- dependent source datasets T = fT1;T2; :::; TN g sification problem, then use appropriate methods for dynamic construction of the meta-train set. to solve it (e.g., one-class SVM (Scholkopf¨ et al., Each task Ti contains labeled training examples 2001) and one-class DL-based classifiers (Ruff (note that a test set is not required in meta- et al., 2018; Manevitz and Yousef, 2007). A group train). D is different from the traditional super- of researchers also proposed an auto-encoder- vised close-domain classification dataset from two folds: 1) Dtest contains OOD testing examples, based approach and its variation to tackle OOD train tasks (Ryu et al., 2017, 2018). Recently, a few pa- whereas D only includes labeled examples for the target domain. 2) The training size for each pers have investigated ID classification and OOD train detection simultaneously (Kim and Kim, 2018; label in D is limited (e.g. less than 100 exam- Shu et al., 2017), but they fail in a low resource ples). Such limitations prevent existing methods setting. from efficiently training a model for either ID clas- sification or OOD detection using Dtrain only. Few-Shot Learning While few-shot learning We propose an OOD-resistant prototypical net- approaches may help with this low-resource set- work for both OOD detection and few-shot ID ting, some recent work is promising in this regard. classification. We follow (Snell et al., 2017) in For example, (Vinyals et al., 2016; Bertinetto few-shot image classification by training a proto- et al., 2016; Snell et al., 2017) use metric learn- typical network on T and directly perform pre- ing by learning a good similarity metric between diction on D without additional training. But our input examples; some other methods adapt a method is different from the prior work in that dur- meta-learning framework, and train the model ing the meta-training, while we maximize the like- to quickly adapt to new tasks with gradients on lihood of the true label for an example in Ti, we small samples, e.g., learning the optimization step also sample an example from another meta-train sizes (Ravi and Larochelle) or model initializa- task Tj for the purpose of OOD training by max- tion (Finn et al., 2017). Though most of these ap- imizing the distance between the OOD instance proaches are explored for computer vision, recent and the prototypical vector of each ID label. 3567 3.1 General Framework 1)2 between the E(·)-encoded representations of As in Fig.1, on a large-scale source dataset T with x and the prototypical vector of a label. Our ex- the following steps: periments show this meta-learning approach is ef- 1. Sample a training task Ti from T (e.g., the ficient for ID classification, but is not good enough Book category of Amazon Review in Section for detecting the OOD examples. 4), and another task Tj from T − Ti (e.g. the We propose two more training losses in addi- Apps-for-Android category). tion to the Lin for OOD detection. The rationale in behind this addition is to adopt the examples from 2. Sample an ID training example xi from Ti, out other tasks as simulated OOD examples for the and a simulated OOD example xj from Tj. current meta-train tasks. Specifically, we first de- 3. Sample N labels (N=4) from Ti in addition out to the label of xin. For the ground-truth la- fine a hinge loss on xj and the closest ID sup- i in bel and N negative labels, we select K train- porting set in S , then we push the examples from ing examples for each label (K-shot learning, another task away from the prototypical vectors of we set K=20).
Load more

Recommended publications
  • An Evaluation of Machine Learning Approaches to Natural Language Processing for Legal Text Classification
    Imperial College London Department of Computing An Evaluation of Machine Learning Approaches to Natural Language Processing for Legal Text Classification Supervisors: Author: Prof Alessandra Russo Clavance Lim Nuri Cingillioglu Submitted in partial fulfillment of the requirements for the MSc degree in Computing Science of Imperial College London September 2019 Contents Abstract 1 Acknowledgements 2 1 Introduction 3 1.1 Motivation .................................. 3 1.2 Aims and objectives ............................ 4 1.3 Outline .................................... 5 2 Background 6 2.1 Overview ................................... 6 2.1.1 Text classification .......................... 6 2.1.2 Training, validation and test sets ................. 6 2.1.3 Cross validation ........................... 7 2.1.4 Hyperparameter optimization ................... 8 2.1.5 Evaluation metrics ......................... 9 2.2 Text classification pipeline ......................... 14 2.3 Feature extraction ............................. 15 2.3.1 Count vectorizer .......................... 15 2.3.2 TF-IDF vectorizer ......................... 16 2.3.3 Word embeddings .......................... 17 2.4 Classifiers .................................. 18 2.4.1 Naive Bayes classifier ........................ 18 2.4.2 Decision tree ............................ 20 2.4.3 Random forest ........................... 21 2.4.4 Logistic regression ......................... 21 2.4.5 Support vector machines ...................... 22 2.4.6 k-Nearest Neighbours .......................
    [Show full text]
  • Fasttext.Zip: Compressing Text Classification Models
    Under review as a conference paper at ICLR 2017 FASTTEXT.ZIP: COMPRESSING TEXT CLASSIFICATION MODELS Armand Joulin, Edouard Grave, Piotr Bojanowski, Matthijs Douze, Herve´ Jegou´ & Tomas Mikolov Facebook AI Research fajoulin,egrave,bojanowski,matthijs,rvj,[email protected] ABSTRACT We consider the problem of producing compact architectures for text classifica- tion, such that the full model fits in a limited amount of memory. After consid- ering different solutions inspired by the hashing literature, we propose a method built upon product quantization to store word embeddings. While the original technique leads to a loss in accuracy, we adapt this method to circumvent quan- tization artefacts. Combined with simple approaches specifically adapted to text classification, our approach derived from fastText requires, at test time, only a fraction of the memory compared to the original FastText, without noticeably sacrificing quality in terms of classification accuracy. Our experiments carried out on several benchmarks show that our approach typically requires two orders of magnitude less memory than fastText while being only slightly inferior with respect to accuracy. As a result, it outperforms the state of the art by a good margin in terms of the compromise between memory usage and accuracy. 1 INTRODUCTION Text classification is an important problem in Natural Language Processing (NLP). Real world use- cases include spam filtering or e-mail categorization. It is a core component in more complex sys- tems such as search and ranking. Recently, deep learning techniques based on neural networks have achieved state of the art results in various NLP applications. One of the main successes of deep learning is due to the effectiveness of recurrent networks for language modeling and their application to speech recognition and machine translation (Mikolov, 2012).
    [Show full text]
  • Supervised and Unsupervised Word Sense Disambiguation on Word Embedding Vectors of Unambiguous Synonyms
    Supervised and Unsupervised Word Sense Disambiguation on Word Embedding Vectors of Unambiguous Synonyms Aleksander Wawer Agnieszka Mykowiecka Institute of Computer Science Institute of Computer Science PAS PAS Jana Kazimierza 5 Jana Kazimierza 5 01-248 Warsaw, Poland 01-248 Warsaw, Poland [email protected] [email protected] Abstract Unsupervised WSD algorithms aim at resolv- ing word ambiguity without the use of annotated This paper compares two approaches to corpora. There are two popular categories of word sense disambiguation using word knowledge-based algorithms. The first one orig- embeddings trained on unambiguous syn- inates from the Lesk (1986) algorithm, and ex- onyms. The first one is an unsupervised ploit the number of common words in two sense method based on computing log proba- definitions (glosses) to select the proper meaning bility from sequences of word embedding in a context. Lesk algorithm relies on the set of vectors, taking into account ambiguous dictionary entries and the information about the word senses and guessing correct sense context in which the word occurs. In (Basile et from context. The second method is super- al., 2014) the concept of overlap is replaced by vised. We use a multilayer neural network similarity represented by a DSM model. The au- model to learn a context-sensitive transfor- thors compute the overlap between the gloss of mation that maps an input vector of am- the meaning and the context as a similarity mea- biguous word into an output vector repre- sure between their corresponding vector represen- senting its sense. We evaluate both meth- tations in a semantic space.
    [Show full text]
  • Investigating Classification for Natural Language Processing Tasks
    UCAM-CL-TR-721 Technical Report ISSN 1476-2986 Number 721 Computer Laboratory Investigating classification for natural language processing tasks Ben W. Medlock June 2008 15 JJ Thomson Avenue Cambridge CB3 0FD United Kingdom phone +44 1223 763500 http://www.cl.cam.ac.uk/ c 2008 Ben W. Medlock This technical report is based on a dissertation submitted September 2007 by the author for the degree of Doctor of Philosophy to the University of Cambridge, Fitzwilliam College. Technical reports published by the University of Cambridge Computer Laboratory are freely available via the Internet: http://www.cl.cam.ac.uk/techreports/ ISSN 1476-2986 Abstract This report investigates the application of classification techniques to four natural lan- guage processing (NLP) tasks. The classification paradigm falls within the family of statistical and machine learning (ML) methods and consists of a framework within which a mechanical `learner' induces a functional mapping between elements drawn from a par- ticular sample space and a set of designated target classes. It is applicable to a wide range of NLP problems and has met with a great deal of success due to its flexibility and firm theoretical foundations. The first task we investigate, topic classification, is firmly established within the NLP/ML communities as a benchmark application for classification research. Our aim is to arrive at a deeper understanding of how class granularity affects classification accuracy and to assess the impact of representational issues on different classification models. Our second task, content-based spam filtering, is a highly topical application for classification techniques due to the ever-worsening problem of unsolicited email.
    [Show full text]
  • Linked Data Triples Enhance Document Relevance Classification
    applied sciences Article Linked Data Triples Enhance Document Relevance Classification Dinesh Nagumothu * , Peter W. Eklund , Bahadorreza Ofoghi and Mohamed Reda Bouadjenek School of Information Technology, Deakin University, Geelong, VIC 3220, Australia; [email protected] (P.W.E.); [email protected] (B.O.); [email protected] (M.R.B.) * Correspondence: [email protected] Abstract: Standardized approaches to relevance classification in information retrieval use generative statistical models to identify the presence or absence of certain topics that might make a document relevant to the searcher. These approaches have been used to better predict relevance on the basis of what the document is “about”, rather than a simple-minded analysis of the bag of words contained within the document. In more recent times, this idea has been extended by using pre-trained deep learning models and text representations, such as GloVe or BERT. These use an external corpus as a knowledge-base that conditions the model to help predict what a document is about. This paper adopts a hybrid approach that leverages the structure of knowledge embedded in a corpus. In particular, the paper reports on experiments where linked data triples (subject-predicate-object), constructed from natural language elements are derived from deep learning. These are evaluated as additional latent semantic features for a relevant document classifier in a customized news- feed website. The research is a synthesis of current thinking in deep learning models in NLP and information retrieval and the predicate structure used in semantic web research. Our experiments Citation: Nagumothu, D.; Eklund, indicate that linked data triples increased the F-score of the baseline GloVe representations by 6% P.W.; Ofoghi, B.; Bouadjenek, M.R.
    [Show full text]
  • Natural Language Processing Security- and Defense-Related Lessons Learned
    July 2021 Perspective EXPERT INSIGHTS ON A TIMELY POLICY ISSUE PETER SCHIRMER, AMBER JAYCOCKS, SEAN MANN, WILLIAM MARCELLINO, LUKE J. MATTHEWS, JOHN DAVID PARSONS, DAVID SCHULKER Natural Language Processing Security- and Defense-Related Lessons Learned his Perspective offers a collection of lessons learned from RAND Corporation projects that employed natural language processing (NLP) tools and methods. It is written as a reference document for the practitioner Tand is not intended to be a primer on concepts, algorithms, or applications, nor does it purport to be a systematic inventory of all lessons relevant to NLP or data analytics. It is based on a convenience sample of NLP practitioners who spend or spent a majority of their time at RAND working on projects related to national defense, national intelligence, international security, or homeland security; thus, the lessons learned are drawn largely from projects in these areas. Although few of the lessons are applicable exclusively to the U.S. Department of Defense (DoD) and its NLP tasks, many may prove particularly salient for DoD, because its terminology is very domain-specific and full of jargon, much of its data are classified or sensitive, its computing environment is more restricted, and its information systems are gen- erally not designed to support large-scale analysis. This Perspective addresses each C O R P O R A T I O N of these issues and many more. The presentation prioritizes • identifying studies conducting health service readability over literary grace. research and primary care research that were sup- We use NLP as an umbrella term for the range of tools ported by federal agencies.
    [Show full text]
  • Using Wordnet to Disambiguate Word Senses for Text Classification
    Using WordNet to Disambiguate Word Senses for Text Classification Ying Liu1, Peter Scheuermann2, Xingsen Li1, and Xingquan Zhu1 1 Data Technology and Knowledge Economy Research Center, Chinese Academy of Sciences Graduate University of Chinese Academy of Sciences 100080, Beijing, China [email protected], [email protected], [email protected] 2 Department of Electrical and Computer Engineering Northwestern University, Evanston, Illinois, USA, 60208 [email protected] Abstract. In this paper, we propose an automatic text classification method based on word sense disambiguation. We use “hood” algorithm to remove the word ambiguity so that each word is replaced by its sense in the context. The nearest ancestors of the senses of all the non-stopwords in a give document are selected as the classes for the given document. We apply our algorithm to Brown Corpus. The effectiveness is evaluated by comparing the classification results with the classification results using manual disambiguation offered by Princeton University. Keywords: disambiguation, word sense, text classification, WordNet. 1 Introduction Text classification aims at automatically assigning a document to a pre-defined topic category. A number of machine learning algorithms have been investigated for text classification, such as K-Nearest Neighbor (KNN) [1], Centroid Classifier [2], Naïve Bayes [3], Decision Trees [4], Support Vector Machines (SVM) [4]. In such classifiers, each document is represented by a n-dimensional vector in a feature space, where each feature is a keyword in the given document. Then traditional classification algorithms can be applied to generate a classification model. To classify an unseen document, a feature vector is constructed by using the same set of n features and then passed to the model as the input.
    [Show full text]
  • An Automatic Text Document Classification Using Modified Weight and Semantic Method
    International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-8 Issue-12, October 2019 An Automatic Text Document Classification using Modified Weight and Semantic Method K.Meena, R.Lawrance Abstract: Text mining is the process of transformation of Text analytics involves information transformation, useful information from the structured or unstructured sources. pattern recognition, information retrieval, association In text mining, feature extraction is one of the vital parts. This analysis, predictive analytics and visualization. The goal of paper analyses some of the feature extraction methods and text mining is to turn text data into useful data for analysis proposed the enhanced method for feature extraction. Term with the help of analytical methods and Natural Language Frequency-Inverse Document Frequency(TF-IDF) method only assigned weight to the term based on the occurrence of the term. Processing (NLP). The increased use of text management in Now, it is enlarged to increases the weight of the most important the internet has resulted in the enhanced research activities words and decreases the weight of the less important words. This in text mining. enlarged method is called as M-TF-IDF. This method does not Now-a-days, everyone has been using internet for surfing, consider the semantic similarity between the terms. Hence, Latent posting, creating blogs and for various purpose. Internet has Semantic Analysis(LSA) method is used for feature extraction massive collection of many text contents. The unstructured and dimensionality reduction. To analyze the performance of the text content size may be varying. For analyzing the text proposed feature extraction methods, two benchmark datasets contents of internet efficiently and effectively, it must be like Reuter-21578-R8 and 20 news group and two real time classified or clustered.
    [Show full text]
  • Natural Language Information Retrieval Text, Speech and Language Technology
    Natural Language Information Retrieval Text, Speech and Language Technology VOLUME7 Series Editors Nancy Ide, Vassar College, New York Jean Veronis, Universite de Provence and CNRS, France Editorial Board Harald Baayen, Max Planck Institute for Psycholinguistics, The Netherlands Kenneth W. Church, AT& T Bell Labs, New Jersey, USA Judith Klavans, Columbia University, New York, USA David T Barnard, University of Regina, Canada Dan Tufis, Romanian Academy of Sciences, Romania Joaquim Llisterri, Universitat Autonoma de Barcelona, Spain Stig Johansson, University of Oslo, Norway Joseph Mariani, LIMS/-CNRS, France The titles published in this series are listed at the end of this volume. Natural Language Information Retrieval Edited by Tomek Strzalkowski General Electric, Research & Development SPRINGER-SCIENCE+BUSINESS MEDIA, B.V. A C.I.P. Catalogue record for this book is available from the Library of Congress. ISBN 978-90-481-5209-4 ISBN 978-94-017-2388-6 (eBook) DOI 10.1007/978-94-017-2388-6 Printed on acid-free paper All Rights Reserved ©1999 Springer Science+Business Media Dordrecht Originally published by Kluwer Academic Publishers in 1999 No part of the material protected by this <:opyright notice may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without written permission from the copyright owner TABLE OF CONTENTS PREFACE xiii CONTRIBUTING AUTHORS xxiii 1 WHAT IS THE ROLE OF NLP IN TEXT RETRIEVAL? }(aren Sparck Jones 1 1 Introduction 0 0 0 0 0 0 0 0 0 0 0 0 1 2 Linguistically-motivated indexing 2 201 Basic Concepts 0 0 0 0 0 .
    [Show full text]
  • The Impact of NLP Techniques in the Multilabel Text Classification Problem
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Repositório Científico da Universidade de Évora The impact of NLP techniques in the multilabel text classification problem Teresa Gon¸calves and Paulo Quaresma Departamento de Inform´atica, Universidade de Ev´ ora, 7000 Ev´ ora, Portugal tcg j [email protected] Abstract. Support Vector Machines have been used successfully to classify text documents into sets of concepts. However, typically, linguistic information is not being used in the classification process or its use has not been fully evaluated. We apply and evaluate two basic linguistic procedures (stop-word removal and stemming/lemmatization) to the multilabel text classification problem. These procedures are applied to the Reuters dataset and to the Portuguese juridical documents from Supreme Courts and Attorney General's Office. 1 Introduction Automatic classification of documents is an important problem in many do- mains. Just as an example, it is needed by web search engines and information retrieval systems to organise text bases into sets of semantic categories. In order to develop better algorithms for document classification it is nec- essary to integrate research from several areas. In this paper, we evaluate the impact of using natural language processing (NLP) and information retrieval (IR) techniques in a machine learning algorithm. Support Vector Machines (SVM) was the chosen classification algorithm. Several approaches to automatically classify text documents using linguis- tic transformations and Machine Learning have been pursued. For example, there are approaches that present alternative ways of representing documents using the RIPPER algorithm [10], that use Na¨ıve Bayes to select features [5] and that generate of phrasal features using both [2].
    [Show full text]
  • Word Embeddings for Multi-Label Document Classification
    Word Embeddings for Multi-label Document Classification Ladislav Lenc†‡ Pavel Kral´ †‡ ‡ NTIS – New Technologies † Department of Computer for the Information Society, Science and Engineering, University of West Bohemia, University of West Bohemia, Plzen,ˇ Czech Republic Plzen,ˇ Czech Republic [email protected] [email protected] Abstract proaches including computer vision and natural language processing. Therefore, in this work we In this paper, we analyze and evaluate use different approaches based on convolutional word embeddings for representation of neural nets (CNNs) which were already presented longer texts in the multi-label document in Kim(2014) and Lenc and Kr al´ (2017). classification scenario. The embeddings Usually, the pre-trained word vectors ob- are used in three convolutional neural net- tained by some semantic model (e.g. word2vec work topologies. The experiments are (w2v) (Mikolov et al., 2013a) or glove (Penning- ˇ realized on the Czech CTK and English ton et al., 2014)) are used for initialization of Reuters-21578 standard corpora. We com- the embedding layer of the particular neural net. pare the results of word2vec static and These vectors can then be progressively adapted trainable embeddings with randomly ini- during neural network training. It was shown in tialized word vectors. We conclude that many experiments that it is possible to obtain bet- initialization does not play an important ter results using these vectors compared to the ran- role for classification. However, learning domly initialized vectors. Moreover, it has been of word vectors is crucial to obtain good proven that even “static” vectors (initialized by results. pre-trained embeddings and fixed during the net- work training) usually bring better performance 1 Introduction than randomly initialized and trained ones.
    [Show full text]
  • Experiments with One-Class SVM and Word Embeddings for Document Classification
    Experiments with One-Class SVM and Word Embeddings for Document Classification Jingnan Bi1[0000−0001−6191−3151], Hyoungah Kim2[0000−0002−0744−5884], and Vito D'Orazio1[0000−0003−4249−0768] 1 University of Texas at Dallas School of Economic, Political, and Policy Sciences 2 Chung-Ang University Institute of Public Policy and Administration Abstract. Researchers often possess a document set that describes some concept of interest, but which has not been systematically collected. This means that the researcher's set of known, relevant documents, is of little use in machine learning applications where the training data is ideally sampled from the same set as the testing data. Here, we propose and test several methods designed to help solve this problem. The central idea is to combine a one-class classifier, here we use the OCC-SVM, with feature engineering methods that we expect to make the input data more generalizable. Specifically, we use two word embeddings approaches, and combine each with a topic model approach. Our experiments show that Word2Vec with Vector Averaging produces the best model. Furthermore, this model is able to maintain high levels of recall at moderate levels of precision. This is valuable for researchers who place a high cost on false negatives. Keywords: document classification · content analysis · one-class algo- rithms · word embeddings. 1 Introduction Text classification is widely used in the computational social sciences [5, 10, 12, 25]. Typically, the initial corpus is not labeled, but researchers wish to classify it to identify relevant information on some concept of interest. In an ideal setting, and to use standard supervised machine learning, a researcher would randomly sample from the corpus, label the sampled documents, and use that as training data.
    [Show full text]