Answer Retrieval Research Articles

Overcoming low-utility facets for complex answer retrieval

Many questions cannot be answered simply; their answers must include numerous nuanced details and additional context. Complex Answer Retrieval (CAR) is the retrieval of answers to such questions. In their simplest form, these questions are constructed from a topic entity (e.g., `cheese') and a facet (e.g., `health effects'). While topic matching has been thoroughly explored, we observe that some facets use general language that is unlikely to appear verbatim in answers. We call these low-utility facets. In this work, we present an approach to CAR that identifies and addresses low-utility facets. We propose two estimators of facet utility. These include exploiting the hierarchical structure of CAR queries and using facet frequency information from training data. To improve the retrieval performance on low-utility headings, we also include entity similarity scores using knowledge graph embeddings. We apply our approaches to a leading neural ranking technique, and evaluate using the TREC CAR dataset. We find that our approach perform significantly better than the unmodified neural ranker and other leading CAR techniques. We also provide a detailed analysis of our results, and verify that low-utility facets are indeed more difficult to match, and that our approach improves the performance for these difficult queries.

Cross-modal Retrieval of Chinese-CQA Based on CCA Algorithm

With the development of Chinese Q&A community, there are a large number of questionanswer pairs has being accumulated. For this question-answer pairs may contain text, pictures, audio, video and other multi-modal data. And the key question for the Chinese Q&A community platform becomes how to match the questions with the most appropriate answers by using cross-modal information such as text and images. In this paper, we propose a question and answer retrieval model based on CCA cross-modal retrieval algorithm. Firstly, the LDA is used to represent Chinese text features, and then the image features are extracted using a convolutional neural network and the Kmeans clustering method is used to obtain image features. Finally, the Canonical Correlation Analysis (CCA) method is used to retrieve between the image and text, CCA method crosses the heterogeneous problem of the underlying multimedia data, and retains the correlation of the variables, then get crossmodel search results of questions and answers. After Clear the correlation between the two models, the image and text features are mapped to the same feature space, and the similarity of the feature vectors can be directly measured, multimodal retrieval with document retrieval map is implemented. The experimental results show that the cross-modal retrieval method based on CCA in Chinese community can improve the accuracy of answer retrieval.

A Tensor Neural Network with Layerwise Pretraining: Towards Effective Answer Retrieval

In this paper we address the answer retrieval problem in community-based question answering. To fully capture the interactions between question-answer pairs, we propose an original tensor neural network to model the relevance between them. The question and candidate answers are separately embedded into different latent semantic spaces, and a 3-way tensor is then utilized to model the interactions between latent semantics. To initialize the network layers properly, we propose a novel algorithm called denoising tensor autoencoder (DTAE), and then implement a layerwise pretraining strategy using denoising autoencoders (DAE) on word embedding layers and DTAE on the tensor layer. The experimental results show that our tensor neural network outperforms various baselines with other competitive neural network methods, and our pretraining DTAE strategy improves the system’s performance and robustness.

Using Cognitive Interviewing to Better Assess Young Adult E-cigarette Use.

Characteristics of electronic nicotine delivery systems (ENDS) make assessment of their use a challenge for researchers. Cognitive interviews are a way of gaining insight into participants' interpretations of survey questions and the methods they use in answering them, to improve survey tools. We used cognitive interviews to modify a young adult survey and improve assessment of quantity and frequency of ENDS use, as well as reasons for initiation and use of ENDS products. Twenty-five college students between the ages of 18 and 32 participated in individual cognitive interviews, which assessed question comprehension, answer estimation, retrieval processes, and answer response processes. Comprehension issues arose discerning between ENDS device types (eg, cigalikes vs. vape pens), and answer estimation issues arose regarding ENDS use as drug delivery systems. These issues appeared to improve when pictures were added specifying the device in question, as well as when specific language naming nicotine as the ENDS product content was added to survey questions. Regarding answer retrieval, this sample of users had problems reporting their frequency of ENDS use, as well as quantifying the amount of ENDS products consumed (eg, volume of e-juice, number of cartridges, nicotine concentration). Accurate assessment of ENDS products proved challenging, but cognitive interviews provided valuable insight into survey interpretation that was otherwise inaccessible to researchers. Future research that explores how to assess the wide array of ENDS devices, as well as possible population differences among specific device-type users would be valuable to public health researchers and professionals. This study extends the current literature by using cognitive interviews to test ENDS assessment questions in a sample of young adults, a population at elevated risk for ENDS use. Problems encountered when answering ENDS use questions underscore the need to develop easily understood ENDS questions that allow for quantification of ENDS use. Future research examining the nature of ENDS product types and different levels of user experience will yield valuable assessment tools for researchers and tobacco control professionals.

Arabic Text Question Answering from an Answer Retrieval Point of View: a survey

Arabic Question Answering (QA) is gaining more importance due to the importance of the language and the dramatic increase in online Arabic content. The goal of this article is to review the state-of-the-art of Arabic QA methods, to classify them into different categories from an answer retrieval viewpoint and to present their applications, issues and new trends. The main components of question answering systems are also presented. Finally, this survey provides a comparative study of systems of each type of QA based on several criteria.

Preliminary report of a Web-based instrument to assess and teach knowledge and clinical thinking to medical student.

ObjectivesWe report the preliminary development of a unique Web-based instrument for assessing and teaching knowledge and developing clinical thinking called the “Sequential Questions and Answers” (SQA) test. Included in this feasibility report are physicians’ answers to the Sequential Questions and Answers pre- and posttests and their brief questionnaire replies.MethodsThe authors refined the SQA test case scenario for content, ease of modifications of case scenarios, test uploading and answer retrieval. Eleven geographically distant physicians evaluated the SQA test, taking the pretest and posttest within two weeks. These physicians completed a brief questionnaire about the SQA test.ResultsEleven physicians completed the SQA pre- and posttest; all answers were downloaded for analysis. They reported the ease of website login and navigating within the test module together with many helpful suggestions. Their average posttest score gain was 53% (p=0.012).ConclusionsWe report the successful launch of a unique Web-based instrument referred to as the Sequential Questions and Answers test. This distinctive test combines teaching organization of the clinical narrative into an assessment tool that promotes acquiring medical knowledge and clinical thinking. We successfully demonstrated the feasibility of geographically distant physicians to access the SQA instrument. The physicians’ helpful suggestions will be added to future SQA test versions. Medical schools might explore the integration of this multi-language-capable SQA assessment and teaching instrument into their undergraduate medical curriculum.

A Student Information System in Restricted Domains

Our approach is tried to achieve high accuracy in Student Information System (SIS) in restricted domain. The restriction leads to the question domains and extract answers. NLP queries get handled and get traversed under the various phases under SIS system for answer retrieval in appropriate defined manner. User query profiling, inference rules and framing are used to infer the answer.

The Electrophysiological Correlates of Scientific Innovation Induced by Heuristic Information

In this study, novel and old scientific innovations (NSI and OSI) were selected as materials to explore the electrophysiological correlates of scientific innovation induced by heuristic information. Using event-related brain potentials (ERPs) to do so, college students solved NSI problems (for which they did not know the answers) and OSI problems (for which they knew the answers). A new experimental paradigm (heuristic information learning–problems testing model) was adopted to make subjects actively find a solution. The results showed that the P3 amplitude was higher for OSI than for NSI between 360 and 430 ms after onset of the problem stimuli. This finding most likely reflects an automatic matching process based on the known answer retrieval, which would be easier for OSI than NSI problems. However, the N4 amplitude was higher for NSI than for OSI between 430 and 500 ms and a greater negativity in the NSI (in comparison with OSI) developed between 500 and 900 ms. This pattern could reflect the generation of novel solutions due to the application of heuristic information (retrieved from memory) during NSI problems solving.

An Improved Answer Retrieval System Taping the Linkage Structure for Noisy SMS Queries

advent of technology has not only tailored the scientific and research work but has also sufficed to the needs of a commoner. Although communication and online support has become quite contemporary yet it requires huge amount of human resource to meet the varied demands. In this paper a server application is proposed which provides automated support for the customer queries via Short Messaging Service (SMS). It facilitates a layman with an instant answer to any of his query. The system is highly capable of handling inherent noise present in the queries and also taps their syntactic and semantic structure. This approach performs two functions simultaneously. One of which is handling of noise and inferring the best possible question, which the user actually meant to ask. Second is to match the refined question with the existing database of questions and then provide a corresponding answer. The system exploits both the above elevated technologies. It is a server application which automates the answer retrieval task via SMS. The retrieval process is such that the user is provided with the five topmost relevant answers to his/her query instantly. The communication starts via SMS from the user. This SMS is received by the system which is installed on central server. It then processes the incoming message. Answer retrieval process deals with two major tasks. In the preliminary step, processing of the message is done in order to eliminate inherent noise. This noise is due to the presence of abbreviations, transliterations, etc. Subsequently, a noise free query is incurred. The refined query is then parsed using the natural language parser providing us with the Question focus. The Question focus derived is used to rank the corpus questions, thereby resulting in the precise answer (1).

Question answering using statistical language modelling

In this paper we present a statistical approach to question answering (QA). Our motivation is to build robust systems for many languages without the need for highly tuned linguistic modules. Consequently, word tokens and web data are used extensively but neither explicit linguistic knowledge nor annotated data is incorporated. A mathematical model for answer retrieval and answer classification is derived. Experiments are conducted by searching for answers in the AQUAINT corpus, as well as in web data. The redundancy inherent in web data outperforms retrieval from a fixed corpus, where there are typically relatively few answer occurrences for any given question. We participated with an implementation of this framework in the TREC 2006 QA evaluations, where we ranked 9th among 27 participants on the factoid task.

Toward Automatic Answers in User-Interactive Question Answering Systems

A strategy of automatic answer retrieval for repeated or similar questions in user-interactive systems by employing semantic question patterns is proposed in this paper. The used semantic question pattern is a generalized representation of a group of questions with both similar structure and relevant semantics. Specifically, it consists of semantic annotations (or constraints) for the variable components in the pattern and hence enhances the semantic representation and greatly reduces the ambiguity of a question instance when asked by a user using such pattern. The proposed method consists of four major steps: structure processing, similar pattern matching and filtering, automatic pattern generation, question similarity evaluation and answer retrieval. Preliminary experiments in a real question answering system show a precision of more than 90% of the method.

A comparison of answer retrieval through four evidence-based textbooks (ACP PIER, Essential Evidence Plus, First Consult, and UpToDate): A randomized controlled trial

Background: The efficacy of bedside information products has not been properly evaluated, particularly in developing countries.Aim: To compare four evidence-based textbooks by comparing efficacy of their use by clinical residents, as measured by the proportion of questions for which relevant answers could be obtained within 20 min, the time to reach the answer and user satisfaction.Methods: One hundred and twelve residents were taught information mastery basics and were randomly allocated to four groups to use: (1) ACP PIER, (2) Essential Evidence Plus (formerly InfoRetriever), (3) First Consult, and (4) UpToDate. Participants received 3 of 24 questions randomly to retrieve the answers from the assigned textbook. Retrieved answers and time-to-answers were recorded by special designed software, and the researchers determined if each recorded answer was relevant.Results: The rate of answer retrieval was 86% in UpToDate, 69% in First Consult, 49% in ACP PIER, and 45% in Essential Evidence Plus (p < 0.001). The mean time-to-answer was 14.6 min using UpToDate, 15.9 min using First Consult, 16.3 min using Essential Evidence Plus, and 17.3 min using ACP PIER (p < 0.001).Conclusion: UpToDate seems more comprehensive in content and also faster than the other three evidence-based textbooks. Thus, it may be considered as one of the best sources for answering clinicians’ questions at the point of care.

To Compare PubMed Clinical Queries and UpToDate in Teaching Information Mastery to Clinical Residents: A Crossover Randomized Controlled Trial

PurposeTo compare PubMed Clinical Queries and UpToDate regarding the amount and speed of information retrieval and users' satisfaction.MethodA cross-over randomized trial was conducted in February 2009 in Tehran University of Medical Sciences that included 44 year-one or two residents who participated in an information mastery workshop. A one-hour lecture on the principles of information mastery was organized followed by self learning slide shows before using each database. Subsequently, participants were randomly assigned to answer 2 clinical scenarios using either UpToDate or PubMed Clinical Queries then crossed to use the other database to answer 2 different clinical scenarios. The proportion of relevantly answered clinical scenarios, time to answer retrieval, and users' satisfaction were measured in each database.ResultsBased on intention-to-treat analysis, participants retrieved the answer of 67 (76%) questions using UpToDate and 38 (43%) questions using PubMed Clinical Queries (P<0.001). The median time to answer retrieval was 17 min (95% CI: 16 to 18) using UpToDate compared to 29 min (95% CI: 26 to 32) using PubMed Clinical Queries (P<0.001). The satisfaction with the accuracy of retrieved answers, interaction with UpToDate and also overall satisfaction were higher among UpToDate users compared to PubMed Clinical Queries users (P<0.001).ConclusionsFor first time users, using UpToDate compared to Pubmed Clinical Querries can lead to not only a higher proportion of relevant answer retrieval within a shorter time, but also a higher users' satisfaction. So, addition of tutoring pre-appraised sources such as UpToDate to the information mastery curricula seems to be highly efficient.

Situating math word problems: The story matters

In two experiments, we explored how the situation model of a math story problem impacts math problem performance. Participants completed multiplication story problems in which a set of objects was associated with or dissociated from a protagonist, making them more or less accessible in memory during answer retrieval. On the basis of previous findings that the sum of two numbers interferes with retrieval of their product, the number of objects in the math problem was either highly interfering ("9" for 4 x 5) or less interfering ("8" for 4 x 5) for multiplication retrieval in the problem. Participants made more errors in problem solving when highly interfering numerical content was associated with the protagonist and, thus, foregrounded. Moreover, the lower one's working memory, the bigger this effect. In sum, small changes in the situation model of a math story problem can harm performance. These changes shift the balance of factors that influence math performance away from math knowledge and toward individual differences in general cognitive capacity.

The impact of frame semantic annotation levels, frame‐alignment techniques, and fusion methods on factoid answer processing

AbstractThe impact of frame semantic enrichment of texts on the task of factoid question answering (QA) is studied in this paper. In particular, we consider different techniques for answer processing with frame semantics: the level of semantic class identification and role assignment to texts, and the fusion of frame semantic‐based answer‐processing approaches with other methods used in the Text REtrieval Conference (TREC). The impact of each of these aspects on the overall performance of a QA system is analyzed in this paper. The TREC 2004 and TREC 2006 factoid question sets were used for the experiments. These demonstrate that the exploitation of encapsulated frame semantics in FrameNet in a shallow semantic parsing process can enhance answer‐processing performance in factoid QA systems. This improvement is dependent on the level of semantic annotation, the frame semantic alignment method, and the method of fusing frame semantic‐based answer‐processing models with other existing models. A more comprehensively annotated environment with all different part‐of‐speech target predicates provides a higher chance of correct factoid answer retrieval where semantic alignment is based on both semantic classes and a relaxed set of semantic roles for answer span identification. Our experiments on fusion techniques of frame semantic‐based and entity‐based answer‐processing models show that merging answer lists with respect to their scores and redundancy by exploiting a fusion function leads to a more effective overall factoid QA system compared to the use of individual models.

Table extraction for answer retrieval

The ability to find tables and extract information from them is a necessary component of many information retrieval tasks. Documents often contain tables in order to communicate densely packed, multi-dimensional information. Tables do this by employing layout patterns to efficiently indicate fields and records in two-dimensional form. Their rich combination of formatting and content presents difficulties for traditional retrieval techniques. This paper describes techniques for extracting tables from text and retrieving answers from the extracted information. We compare machine learning (especially, Conditional Random Fields) and heuristic methods for table extraction. To retrieve answers, our approach creates a cell document, which contains the cell and its metadata (headers, titles) for each table cell, and the retrieval model ranks the cells of the extracted tables using a language-modeling approach. Performance is tested using government statistical Web sites and news articles, and errors are analyzed in order to improve the system.

Verification of Multiplication Facts: An Investigation Using Retrospective Protocols

Retrospective verbal protocols collected throughout participants' performance of a multiplication verification task (e.g., "7 x 3 = 28, true or false?") documented a number of different strategies and changes in strategy use across different problem categories used for this common experimental task. Correct answer retrieval and comparison to the candidate answer was the modal but not the only strategy reported. Experiment 1 results supported the use of a calculation algorithm on some trials and the use of the difference between the candidate and correct answers (i.e., split) on others. Experiment 2 clearly demonstrated that participants sometimes bypassed retrieval by relying on the split information. Implications for mental arithmetic theories and the general efficacy of retrospective protocols are discussed.

Implementing clarification dialogues in open domain question answering

We examine the implementation of clarification dialogues, a mechanism for ensuring that question answering systems take into account user goals by allowing them to ask series of related questions either by refining or expanding on previous questions with follow-up questions, in the context of open domain Question Answering systems. We develop an algorithm for clarification dialogue recognition through the analysis of collected data on clarification dialogues and examine the importance of clarification dialogue recognition for question answering. The algorithm is evaluated and shown to successfully recognize the start and continuation of clarification dialogues in 94% of cases. We then show the usefulness of the algorithm by demonstrating how the recognition of clarification dialogues can simplify the task of answer retrieval.

ベクトル空間モデルに基づく特定領域向け中国語質問応答システムの構築

In order to meet the demand to acquire necessary information efficiently from large electronic text, the Question and Answering (QA) technology to show a clear reply automatically to a question asked in the user's natural language has widely attracted attention in recent years. Although the research of QA system in China is later than that in western countries and Japan, it has attracted more and more attention recently. In this paper, we propose a Question-Answering construction, which synthesizes the answer retrieval to the questions asked most frequently based on common knowledge, and the document retrieval concerning sightseeing information. In order to improve reply accuracy, one must consider the synthetic model based on statistic VSM and the shallow semantic analysis, and the domain is limited to sightseeing information. A Chinese QA system about sightseeing based on the proposed method has been built. The result is obtained by evaluation experiments, where high accuracy can be achieved when the results of retrieval were regarded as correct, if the correct answer appeared among those of the top three resemblance degree. The experiments proved the efficiency of our method and it is feasible to develop Question-Answering technology based on this method.

Interactive effects of numerical surface form and operand parity in cognitive arithmetic.

In Experiment 1, adults (n = 48) performed simple addition, multiplication, and parity (i.e., odd-even) comparisons on pairs of Arabic digits or English number words. For addition and comparison, but not multiplication, response time increased with the number of odd operands. For addition, but not comparison, this parity effect was greater for words than for digits. In Experiment 2, adults (n = 50) solved simple addition problems in digit and word format and reported their strategies (i.e., retrieval or procedures). Procedural strategies were used more for odd than even addends and much more for word than digit problems. The results indicate that problem encoding and answer retrieval processes for cognitive arithmetic are interactive rather than strictly additive stages.