Learning to Weight for Text Classification

A novel term weighting scheme for text classification: TF-MONO

Although term weighting approach is typically used to improve the performance of text classification, this approach may not provide consistent results while imbalanced data distribution is available. This paper presents a probability based term weighting approach which addresses the different aspects of class imbalance problem in text classification. In this approach, we proposed two term evaluation functions called as PNF and \(PNF^2\) which can produce more influential weights by relying on the imbalanced data sets. These functions can determine the significance of a term in association with a particular category. This is a crucial point because in one hand a frequent term is more important than a rare term in a particular category according to feature selection approach, and on the other hand a rare term is no less important than a frequent term based on idf assumption of traditional term weighting approach. Incorporation of these two approaches at the same time is the main idea that make them superior to other weighting methods. The achieved results from experiments which were carried out on two popular benchmarks (Reuters-21578 and WebKB) demonstrate that the probability based term weighting approach yields more consistent results than the other methods on the imbalanced data sets.KeywordsText classificationClass imbalance problemTerm weighting approachMachine learning

The performance of text classification can be affected by the choice of appropriate term weighting scheme as well as other parameters. The terminology supervised term weighting scheme has become popular in recent years, as it may provide discriminative representation in vector space for text documents belonging to different classes. A term weighting scheme generally consists of three factors, namely term frequency factor, collection frequency factor, and length normalization factor. The researchers mostly have been focused on developing new collection frequency factors in term weighting studies. However, the term frequency factor has an important role, especially in supervised term weighting. In this study, we extensively analyzed the effects of using different term frequency factors on seven supervised term weighting schemes. While six of these supervised term weighting schemes were applied in the previous studies in the literature, we derived one of them from an existing feature selection method and it was not used as a weighting method before. This analysis is performed using SVM and Roccio classifiers on two widely known benchmark datasets with different characteristics. Experimental results showed that modification of term frequency factor in supervised term weighting schemes increased the performance of almost all weighting schemes. Also, term weighting schemes using square root function-based term frequency factor (SQRT_TF) are more successful than the ones using term frequency (TF) and logarithmic function-based term frequency (LOG_TF) factors. TF term frequency factor seems as the least effective one among three different term frequency factors according to the experimental results and statistical analysis.

The construction of a text classifier usually involves (i) a phase of term selection, in which the most relevant terms for the classification task are identified, (ii) a phase of term weighting, in which document weights for the selected terms are computed, and (iii) a phase of classifier learning, in which a classifier is generated from the weighted representations of the training documents. This process involves an activity of learning, in which information on the membership of training documents in categories is used. Traditionally, learning enters only phases (i) and (iii). In this paper we propose instead that learning from training data should also affect phase (ii), i.e. that information on the membership of training documents to categories be used to determine term weights. We call this idea term weighting (STW). As an example, we propose a number of supervised variants of t f idf weighting, obtained by replacing the idf function with the function that has been used in phase (i) for term selection. We present experimental results obtained on the standard Reuters-21578 benchmark with one classifier learning method (support vector machines), three term selection functions (information gain, chi-square, and gain ratio), and both local and global term selection and weighting.

A fundamental task in Information Retrieval (IR) is term weighting. Early IR theory considered both the presence or absence of all terms in the lexicon for ranking and needed to weight them all. Yet, as the size of lexicons grew and models became too complex, common weighting models preferred to aggregate only the weights of the query terms that are matched in candidate documents. Thus, unmatched term contribution in these models is only considered indirectly, such as in probability smoothing with corpus distribution, or in weight normalization by document length. In this work we propose a novel term weighting model that directly assesses the weights of unmatched terms, and show its benefits. Specifically, we propose a Learning To Rank framework, in which features corresponding to matched terms are also "mirrored" in similar features that account only for unmatched terms. The relative importance of each feature is learned via a click-through query log. As a test case, we consider vertical search in Community-based Question Answering(CQA) sites from Web queries. Queries that result in viewing CQA content often contain fine grained information needs and benefit more from unmatched term weighting. We assess our model both via manual evaluation and via automatic evaluation over a clickthrough log. Our results show consistent improvement in retrieval when unmatched information is taken into account. This holds both when only identical terms are considered matched, and when related terms are matched via distributional similarity.

Multi-Tier Attention Network using Term-weighted Question Features for Visual Question Answering

With the rapid development of the internet technology, a large amount of internet text data can be obtained. The text classification (TC) technology plays a very important role in processing massive text data, but the accuracy of classification is directly affected by the performance of term weighting in TC. Due to the original design of information retrieval (IR), term frequency-inverse document frequency (TF-IDF) is not effective enough for TC, especially for processing text data with unbalanced distributions in internet media reports. Therefore, the variance between the DF value of a particular term and the average of all DFsDF¯, namely, the document frequency variance (ADF), is proposed to enhance the ability in processing text data with unbalanced distribution. Then, the normal TF-IDF is modified by the proposed ADF for processing unbalanced text collection in four different ways, namely, TF-IADF, TF-IADF+, TF-IADFnorm, and TF-IADF+norm. As a result, an effective model can be established for the TC task of internet media reports. A series of simulations have been carried out to evaluate the performance of the proposed methods. Compared with TF-IDF on state-of-the-art classification algorithms, the effectiveness and feasibility of the proposed methods are confirmed by simulation results.

Now finding relevant information is absolutely needed by users of the information. Actually, finding relevant information is more difficult than just looking for information. The presence information retrieval (IR) system is a system for automatically searching the relevant information to the user needs, which is expressed in the required query, and this query will be used as an input and processed by IR system and then will get relevant documents to the required query. One of methods of finding relevant information by query is term weighting by both query and document. The commonly used methods are the term weighting of local and global weighting, while the local term weighting algorithms that are used such as TF, TF logarithmic, binary TF and augmented TF and global weighting is IDF, IDFP, IDFB. In this research, besides of comparison of term weighting algorithms, it is also proposed new improvement in term weighting algorithms by adding global weight value that changes base 10 logarithmic to base 2 logarithmic and adds the number 1 in the global term weighting IDFP. Measurements result of this research are Precision, Recall and NIAP. In This research, the best result for precision and NAIP have been performed by new propose term weighting algorithms. After compare and analyze the result of the weighting algorithms. It is necessary to combine the weighting with semantic algorithms so that it will get better result.

Natural Language Processing, specifically text classification or text categorization, has become a trend in computer science. Commonly, text classification is used to categorize large amounts of data to allocate less time to retrieve information. Students, as well as research advisers and panelists, take extra effort and time in classifying research documents. To solve this problem, the researchers used state-of-the-art supervised term weighting schemes, namely: TF-MONO and SQRTF-MONO and its application to machine learning algorithms: K-Nearest Neighbor, Linear Support Vector, Naive Bayes Classifiers, creating a total of six classifier models to ascertain which of them performs optimally in classifying research documents while utilizing Optical Character Recognition for text extraction. The results showed that among all classification models trained, SQRTF-MONO and Linear SVC outperformed all other models with an F1 score of 0.94 both in the abstract and the background of the study datasets. In conclusion, the developed classification model and application prototype can be a tool to help researchers, advisers, and panelists to lessen the time spent in classifying research documents.

This is the first workshop on the emerging interdisciplinary research area of applying axiomatic thinking to information retrieval (IR) and related tasks. The workshop aims to help foster collaboration of researchers working on different perspectives of axiomatic thinking and encourage discussion and research on general methodological issues related to applying axiomatic thinking to IR and related tasks.

A machine learning-based sentiment analysis of online product reviews with a novel term weighting and feature selection approach

Learning a high-dimensional dense representation for vocabulary terms, also known as a word embedding, has recently attracted much attention in natural language processing and information retrieval tasks. The embedding vectors are typically learned based on term proximity in a large corpus. This means that the objective in well-known word embedding algorithms, e.g., word2vec, is to accurately predict adjacent word(s) for a given word or context. However, this objective is not necessarily equivalent to the goal of many information retrieval (IR) tasks. The primary objective in various IR tasks is to capture relevance instead of term proximity, syntactic, or even semantic similarity. This is the motivation for developing unsupervised relevance-based word embedding models that learn word representations based on query-document relevance information. In this paper, we propose two learning models with different objective functions; one learns a relevance distribution over the vocabulary set for each query, and the other classifies each term as belonging to the relevant or non-relevant class for each query. To train our models, we used over six million unique queries and the top ranked documents retrieved in response to each query, which are assumed to be relevant to the query. We extrinsically evaluate our learned word representation models using two IR tasks: query expansion and query classification. Both query expansion experiments on four TREC collections and query classification experiments on the KDD Cup 2005 dataset suggest that the relevance-based word embedding models significantly outperform state-of-the-art proximity-based embedding models, such as word2vec and GloVe.

Query expansion is a classical topic in the field of information retrieval, which is proposed to bridge the gap between searchers' information intents and their queries. Previous researches usually expand queries based on document collections, or some external resources such as WordNet and Wikipedia [1, 2, 3, 4, 5]. However, it seems that independently using one of these resources has some defects, document collections lack semantic information of words, while WordNet and Wikipedia may not include domain-specific knowledge in certain document collection. Our work aims to combine these two kinds of resources to establish an expansion model which represents not only domain-specific information but also semantic information. In our preliminary experiments, we construct a two-layer word graph and use Random-Walk algorithm to calculate the weights of each term in pseudo-relevance feedback documents, then select the highest weighted term to expand original query. The first layer of the word graph contains terms in related documents, while the second layer contains semantic senses corresponding to these terms. These terms and semantic senses are treated as vertices of the graph and connected with each other by all possible relationships, such as mutual information and semantic similarities. We utilized mutual information, semantic similarity and uniform distribution as the weight of term-term relation, sense-sense relation and word-sense relation respectively. Though these experiments show that our expansion outperform original queries, we are troubled with some difficult problems. Given the framework of semantic graph model, we need more effort to find out an optimal graph to represent the relationships between terms and their semantic senses. We utilized a two-layer graph model in our preliminary research, where terms from different documents are treated equally. Maybe we can introduce the document as a third layer in future work, where we can differ the same terms in different documents according to document relevance and context. Then we need appropriately represent initial weights of this words, senses and relationships. Various measures for weights of terms and term relations have been proved effective in other information retrieval tasks, such as TFIDF, mutual information (MI), but there is little research on weights for semantic senses and their relations. For polysemous words, we add all of their semantic senses to the graph and assume that these senses are uniformly distributed. Actually, it is not precise for a word in a special document and query. As we know, a polysemous word may have only one or two senses in a document, and they are not uniformly distributed. Give a word, what we should do is to determine its word senses in a relevant document and estimate the distribution of these senses. Word sense disambiguation may help us in this problem. Then, there are many methods to compute word similarity according to WordNet, which we use to represent the weights of relationships between word senses. Varelas et al implemented some popular methods to compute semantic similarity by mapping terms to an ontology and examining their relationships in that ontology [4]. We also need to know which algorithm for semantic similarity is most suitable for our model. Additional, WordNet is suitable to calculate word similarity but not suitable to measure word relevance. The inner hyperlinks of Wikipedia could help us to calculate word relevance. We wish to find an effective way to combine the similarity measure from WordNet and relevance measure from Wikipedia, which may completely reflect word relationships.

Stemming is a fundamental step in processing textual data preceding the tasks of text mining, Information Retrieval (IR), and natural language processing (NLP). The common goal of stemming is to standardize words by reducing a word to its base (root or stem), thus can be also considered a feature reduction technique. This paper aims at presenting a new dictionary free, content-based Arabic stemmer and adopts it as a feature reduction (selection) mechanism to study its contribution in improving Arabic text categorization. We employed three stemming mechanisms (root-based, light, and our stemming technique and assessed their performance in text classification exercises for an Arabic corpus to compare and contrast the text mining effectiveness of these Arabic stemming algorithms. The experiments were conducted on a corpus consisting of 2,966 Arabic documents that fall into three categories: cultural, social, and general. The experiment results showed that our stemmer significantly improved text classification accuracy.

Preprocessing is an important task and critical step in information retrieval and text mining. The objective of this study is to analyze the effect of preprocessing methods in text classification on Turkish texts. We compiled two large datasets from Turkish newspapers using a crawler. On these compiled data sets and using two additional datasets, we perform a detailed analysis of preprocessing methods such as stemming, stopword filtering and word weighting for Turkish text classification on several different Turkish datasets. We report the results of extensive experiments.