Information Retrieval Components Research Articles

Framework for syntactic string similarity measures

Similarity measure is an essential component of information retrieval, document clustering, text summarization, and question answering, among others. In this paper, we introduce a general framework of syntactic similarity measures for matching short text. We thoroughly analyze the measures by dividing them into three components: character-level similarity, string segmentation, and matching technique. Soft variants of the measures are also introduced. With the help of two existing toolkits (SecondString and SimMetric), we provide an open-source Java toolkit of the proposed framework, which integrates the individual components together so that completely new combinations can be created. Experimental results reveal that the performance of the similarity measures depends on the type of the dataset. For well-maintained dataset, using a token-level measure is important but the basic (crisp) variant is usually enough. For uncontrolled dataset where typing errors are expected, the soft variants of the token-level measures are necessary. Among all tested measures, a soft token-level measure that combines set matching and q-grams at the character level perform best. A gap between human perception and syntactic measures still remains due to lacking semantic analysis.

R<sup>3</sup>: Reinforced Ranker-Reader for Open-Domain Question Answering

In recent years researchers have achieved considerable success applying neural network methods to question answering (QA). These approaches have achieved state of the art results in simplified closed-domain settings such as the SQuAD (Rajpurkar et al. 2016) dataset, which provides a pre-selected passage, from which the answer to a given question may be extracted. More recently, researchers have begun to tackle open-domain QA, in which the model is given a question and access to a large corpus (e.g., wikipedia) instead of a pre-selected passage (Chen et al. 2017a). This setting is more complex as it requires large-scale search for relevant passages by an information retrieval component, combined with a reading comprehension model that “reads” the passages to generate an answer to the question. Performance in this setting lags well behind closed-domain performance. In this paper, we present a novel open-domain QA system called Reinforced Ranker-Reader (R3), based on two algorithmic innovations. First, we propose a new pipeline for open-domain QA with a Ranker component, which learns to rank retrieved passages in terms of likelihood of extracting the ground-truth answer to a given question. Second, we propose a novel method that jointly trains the Ranker along with an answer-extraction Reader model, based on reinforcement learning. We report extensive experimental results showing that our method significantly improves on the state of the art for multiple open-domain QA datasets.

Unsupervised Latent Dirichlet Allocation for supervised question classification

Question answering systems assist users in satisfying their information needs more precisely by providing focused responses to their questions. Among the various systems developed for such a purpose, community-based question answering has recently received researchers’ attention due to the large amount of user-generated questions and answers in social question-and-answer platforms. Reusing such data sources requires an accurate information retrieval component enhanced by a question classifier. The question classification gives the system the possibility to have information about question categories to focus on questions and answers from relevant categories to the input question. In this paper, we propose a new method based on unsupervised Latent Dirichlet Allocation for classifying questions in community-based question answering. Our method first uses unsupervised topic modeling to extract topics from a large amount of unlabeled data. The learned topics are then used in the training phase to find their association with the available category labels in the training data. The category mixture of topics is finally used to predict the label of unseen data.

The influence of image interactivity upon user engagement when using mobile touch screens

Touch screens are a key component of consumer mobile devices such as smartphones and tablets, as well as an increasingly common self-service component of information retrieval on fixed screens and mobile devices in-store. The ubiquity of touch screens in daily life increases consumer accessibility and extended use for shopping, whilst software innovations have increased the functionality of touch screens, for example the extent to which images respond to fingertip control. This study examines how users engage with interactive visual rotation and tactile simulation features while browsing fashion clothing products on touch screen devices and thus contributes to retail touch screen research that previously focused on in-store kiosks and window displays. Findings show that three dimensions of user engagement (endurability, novelty and felt involvement) are positively influenced by both forms of manipulation. In order to examine the extent to which touch screen user engagement varies with individual preferences for an in-store experience, the paper also examines whether user engagement outcomes are mediated by an individual's need for physical touch. Findings indicate that the need for touch does not explain the variance between individuals. We conclude that touch screen technology complements the physical retail environment.

A library’s information retrieval system (In)effectiveness: case study

Purpose– The purpose of this paper is to evaluate the effectiveness of the information retrieval component of a daily newspaper publisher’s integrated library system (ILS) in comparison with the open source alternatives and observe the impact of the scale of metadata, generated daily by library administrators, on retrieved result sets.Design/methodology/approach– In Experiment 1, the authors compared the result sets of the information retrieval system (IRS) component of the publisher’s current ILS and the result sets of proposed ones with human-assessed relevance judgment set. In Experiment 2, the authors compared the performance of proposed IRS components with the publisher’s current production IRS, using result sets of current IRS classified as relevant. Both experiments were conducted using standard information retrieval (IR) evaluation methods: precision, recall, precision atk,F-measure, mean average precision and 11-point interpolated average precision.Findings– Results showed that: first, in Experiment 1, the publisher’s current production ILS ranked last of all participating IRSs when compared to a relevance document set classified by the senior library administrator; and second, in Experiment 2, the tested IR components’ request handlers that used only automatically generated metadata performed slightly better than request handlers that used all of the metadata fields. Therefore, regarding the effectiveness of IR, the daily human effort of generating the publisher’s current set of metadata attributes is unjustified.Research limitations/implications– The experiments’ collections contained Slovene language with large number of variations of the forms of nouns, verbs and adjectives. The results could be different if the experiments’ collections contained languages with different grammatical properties.Practical implications– The authors have confirmed, using standard IR methods, that the IR component used in the publisher’s current ILS, could be adequately replaced with an open source component. Based on the research, the publisher could incorporate the suggested open source IR components in practice. In the research, the authors have described the methods that can be used by libraries for evaluating the effectiveness of the IR of their ILSs.Originality/value– The paper provides a framework for the evaluation of an ILS’s IR effectiveness for libraries. Based on the evaluation results, the libraries could replace the IR components if their current information system setup allows it.

Research and Improvement on Content-Based Web Search Engine

There is a lot of information contained in the World Wide Web. It has become a research focus to obtain the required related resources quickly and accurately from the web through the content-based search engines. Most current tools of full text web search engine, such as Lucene which is a widely used open source retrieval library in information retrieval field, are purely keyword based. This may not sufficient for users to retrieve in the web. In this paper, we employ a method to overcome the limitations of current full text search engines in represent of Lucene. We propose a Query Expansion and Information Retrieval approach which can help users to acquire more accurate contents from the web. The Query Expansion component finds expanded candidate words of the query word through WordNet which contains synonyms in several different senses; In the Information Retrieval component, the query word and its candidate words are used together as the input of the search module to get the result items. Furthermore, we can put the result items into different classes based on the expansion. Some experiments and the results are described in the late part of this paper.

Machine learning for query formulation in question answering

AbstractResearch on question answering dates back to the 1960s but has more recently been revisited as part of TREC's evaluation campaigns, where question answering is addressed as a subarea of information retrieval that focuses on specific answers to a user's information need. Whereas document retrieval systems aim to return the documents that are most relevant to a user's query, question answering systems aim to return actual answers to a users question. Despite this difference, question answering systems rely on information retrieval components to identify documents that contain an answer to a user's question. The computationally more expensive answer extraction methods are then applied only to this subset of documents that are likely to contain an answer. As information retrieval methods are used to filter the documents in the collection, the performance of this component is critical as documents that are not retrieved are not analyzed by the answer extraction component. The formulation of queries that are used for retrieving those documents has a strong impact on the effectiveness of the retrieval component. In this paper, we focus on predicting the importance of terms from the original question. We use model tree machine learning techniques in order to assign weights to query terms according to their usefulness for identifying documents that contain an answer. Term weights are learned by inspecting a large number of query formulation variations and their respective accuracy in identifying documents containing an answer. Several linguistic features are used for building the models, including part-of-speech tags, degree of connectivity in the dependency parse tree of the question, and ontological information. All of these features are extracted automatically by using several natural language processing tools. Incorporating the learned weights into a state-of-the-art retrieval system results in statistically significant improvements in identifying answer-bearing documents.

Pattern-based knowledge workflow automation: concepts and issues

As the result of business process automation, more and more knowledge is codified and stored in knowledge repositories and scattered in employees’ computers across functionally and geographically separated business units. While several alternative mechanisms such as chat rooms, search engines, recommender systems exist to retrieve and access knowledge, satisfying a knowledge requirement often involves the coordination of multiple tasks and the use of several technologies. However, there is no existing technology for orchestrating various collaboration, communication and information retrieval components to satisfy the knowledge needs in corporations. In order to address this gap, we propose the approach of pattern-based knowledge workflow that can enable the automation of knowledge flows across an organization. In this paper, we present an overview of the pattern-based knowledge workflow approach and propose extensions to the conventional workflow paradigm. We demonstrate the feasibility of implementing knowledge workflows by means of a Business Process Execution Language (BPEL) specification for executing knowledge workflows in a business setting. We then discuss the engineering challenges and research issues that need to be addressed to further develop this approach.

Introducing a conceptual information retrieval (IR) framework.

This is the third in the series of the articles on an application of the systems analytic approach to evaluation of information retrieval (IR). Previously terminological and evaluation problems associated with IR were identified, and it was proposed that the systems analytic approach can provide solutions to these problems. Here, after the general discussion of a systems approach, different attempts to introduce this approach into IR are presented and critiqued: modelling of IR, identifying boundaries of a system under evaluation, identifying variables under investigation; and creating evaluation frameworks. On the basis of the critique of this work, the authors present their proposal for a solution: first the IR components, their boundaries and relationships are identified, and then the two models of IR are introduced. One of the models addresses components of the IR process and relationships between them, the second one is the process model of IR.

Bankruptcy support system: taking advantage of information retrieval and case-based reasoning

This paper is a presentation of an on-going work in which we attempt to take advantage of information retrieval (IR) and artificial intelligence techniques combined. It is an application of case-based reasoning (CBR) with an automatic indexing IR component in the legal domain of bankruptcy law. The model is based on our intuition of how lawyers go about doing their legal research and reasoning tasks in case law. We take advantage of the built-in knowledge contained in the carefully prepared statute text in a front-end processor and classification component to the CBR system. Our end result is an IR–CBR bankruptcy support system (BanXupport).

Classification-based reasoning

A representation formalism for N-ary relations, quantification, and definition of concepts is described. Three types of conditions are associated with the concepts: (1) necessary and sufficient properties, (2) contingent properties, and (3) necessary properties. Also explained is how complex chains of inferences can be accomplished by representing existentially quantified sentences, and concepts denoted by restrictive relative clauses as classification hierarchies. The representation structures that make possible the inferences are explained first, followed by the reasoning algorithms that draw the inferences from the knowledge structures. All the ideas explained have been implemented and are part of the information retrieval component of a program called Snowy. An appendix contains a brief session with the program. >

Linguistically based functions in information retrieval: PADOK and the German Patent Information System

This paper reports on methodological considerations and the results of the Information Retrieval (IR) project PADOK I and II. PADOK has been carried out by the Linguistic Information Science Group of the University of Regensburg (LIR) since November 1984 and has been sponsored by the German Ministry for Research and Technology. The long term objective is to integrate artificial intelligence topics and the methods of information retrieval research without neglecting traditional IR methodology. In PADOK we consider a type of mass data IR system which indexes its documents rather shallowly (freetext or morphological components) and adds an intelligent information retrieval component to this kernel system. So far we have obtained, on the basis of two large-scale retrieval tests of the German Patent Information System results which show how the linguistically based functions of an indexing system contribute to its performance, and indicate what is the most reasonable basic content analysis program for a German Patent Information System. This paper focusses on the general principles and aims of PADOK I and PADOK R and on the statistical evaluation of the retrieval tests. Christa Womser-Hacker has a Ph.D. in Linguistic Information Science. From 1985 until 1990 she was involved in several LIR-Projects concerning text processing, evaluation of the German Patent Information System, man-machine-interaction, intelligent interfaces for databases. Since May 1990 she has been an LIR staff member. She is interested in information retrieval, (statistical) evaluation methods of man-machine-interaction, intelligent interfaces. She has published Der PADOK-Retrieval-test (1989) and “Die statistische Auswertung des Retrievaltests” (1990).

Some research problems in automatic information retrieval

Information retrieval components are currently incorporated in several types of information systems, including bibliographic retrieval systems, data base management systems and question-answering systems. Some of the problems arising in the real-time environment in which these systems operate are briefly discussed. Certain recent advances in information retrieval research are then mentioned, including the formulation of new probabilistic retrieval models, and the development of automatic document analysis and Boolean query processing techniques.