Search results for: approximate SQL queries

Authors: Ibrahim Gomaa, Hoda M. O. Mokhtar

Abstract:

Although most of the existing skyline queries algorithms focused basically on querying static points through static databases; with the expanding number of sensors, wireless communications and mobile applications, the demand for continuous skyline queries has increased. Unlike traditional skyline queries which only consider static attributes, continuous skyline queries include dynamic attributes, as well as the static ones. However, as skyline queries computation is based on checking the domination of skyline points over all dimensions, considering both the static and dynamic attributes without separation is required. In this paper, we present an efficient algorithm for computing continuous skyline queries without discriminating between static and dynamic attributes. Our algorithm in brief proceeds as follows: First, it excludes the points which will not be in the initial skyline result; this pruning phase reduces the required number of comparisons. Second, the association between the spatial positions of data points is examined; this phase gives an idea of where changes in the result might occur and consequently enables us to efficiently update the skyline result (continuous update) rather than computing the skyline from scratch. Finally, experimental evaluation is provided which demonstrates the accuracy, performance and efficiency of our algorithm over other existing approaches.

Keywords: continuous query processing, dynamic database, moving object, skyline queries

Procedia PDF Downloads 188

Authors: Mohammad Hossein Ahmadi, Vahid Haghighatdoost

Abstract:

Spatial databases have been an active area of research over years. In this paper, we study how to answer the General Time-Dependent Sequenced Route queries. Given the origin and destination of a user over a time-dependent road network graph, an ordered list of categories of interests and a departure time interval, our goal is to find the minimum travel time path along with the best departure time that minimizes the total travel time from the source location to the given destination passing through a sequence of points of interests belonging to each of the specified categories of interest. The challenge of this problem is the added complexity to the optimal sequenced route queries, where we assume that first the road network is time dependent, and secondly the user defines a departure time interval instead of one single departure time instance. For processing general time-dependent sequenced route queries, we propose two solutions as Discrete-Time and Continuous-Time Sequenced Route approaches, finding approximate and exact solutions, respectively. Our proposed approaches traverse the road network based on A*-search paradigm equipped with an efficient heuristic function, for shrinking the search space. Extensive experiments are conducted to verify the efficiency of our proposed approaches.

Keywords: trip planning, time dependent, sequenced route query, road networks

Procedia PDF Downloads 283

Authors: Filip Buruiana, Alexander Hall, Reimar Hofmann, Thomas Hofmann, Silviu Ganceanu, Alexandru Tudorica

Abstract:

In-memory column-stores make interactive analysis feasible for many big data scenarios. PowerDrill is a system used internally at Google for exploration in logs data. Even though it is a highly parallelized column-store and uses in memory caching, interactive response times cannot be achieved for all datasets (note that it is common to analyze data with 50 billion records in PowerDrill). In this paper, we investigate two orthogonal approaches to optimize performance at the expense of an acceptable loss of accuracy. Both approaches can be implemented as outer wrappers around existing database engines and so they should be easily applicable to other systems. For the first optimization we show that memory is the limiting factor in executing queries at speed and therefore explore possibilities to improve memory efficiency. We adapt some of the theory behind data sketches to reduce the size of particularly expensive fields in our largest tables by a factor of 4.5 when compared to a standard compression algorithm. This saves 37% of the overall memory in PowerDrill and introduces a 0.4% relative error in the 90th percentile for results of queries with the expensive fields. We additionally evaluate the effects of using sampling on accuracy and propose a simple heuristic for annotating individual result-values as accurate (or not). Based on measurements of user behavior in our real production system, we show that these estimates are essential for interpreting intermediate results before final results are available. For a large set of queries this effectively brings down the 95th latency percentile from 30 to 4 seconds.

Keywords: big data, in-memory column-store, high-performance SQL queries, approximate SQL queries

Procedia PDF Downloads 228

Authors: Gemma García López

Abstract:

The goal of this study is to analyze if search queries carried out in search engines such as Google, can offer emotional information about the user that performs them. Knowing the emotional state in which the Internet user is located can be a key to achieve the maximum personalization of content and the detection of worrying behaviors. For this, two studies were carried out using tools with advanced natural language processing techniques. The first study determines if a query can be classified as positive, negative or neutral, while the second study extracts emotional content from words and applies the categorical and dimensional models for the representation of emotions. In addition, we use search queries in Spanish and English to establish similarities and differences between two languages. The results revealed that text search queries performed by users on the Internet can be classified emotionally. This allows us to better understand the emotional state of the user at the time of the search, which could involve adapting the technology and personalizing the responses to different emotional states.

Keywords: emotion classification, text search queries, emotional analysis, sentiment analysis in text, natural language processing

Procedia PDF Downloads 113

Authors: Elaheh Vaezpour

Abstract:

Data sensing in wireless sensor networks is done by query deceleration the network by the users. In many applications of the wireless sensor networks, many users send queries to the network simultaneously. If the queries are processed separately, the network’s energy consumption will increase significantly. Therefore, it is very important to aggregate the queries before sending them to the network. In this paper, we propose a multiple query optimization framework based on sensors physical and temporal correlation. In the proposed method, queries are merged and sent to network by considering correlation among the sensors in order to reduce the communication cost between the sensors and the base station.

Keywords: wireless sensor networks, multiple query optimization, data correlation, reducing energy consumption

Procedia PDF Downloads 306

Authors: Zhanfeng Ma, Zhiping Xiong, Hu Yin, Zhengwei She, Aditya P. Gurajada, Tianlun Chen, Ying Li

Abstract:

Many RDBMS vendors use B-tree index to achieve high performance for point queries and range queries, and some of them also employ hash index to further enhance the performance as hash table is more efficient for point queries. However, there are extra overheads to maintain a separate hash index, for example, hash mapping for all data records must always be maintained, which results in more memory space consumption; locking, logging and other mechanisms are needed to guarantee ACID, which affects the concurrency and scalability of the system. To relieve the overheads, Hash Cached B-tree (HCB) index is proposed in this paper, which consists of a standard disk-based B-tree index and an additional in-memory lock-free hash table. Initially, only the B-tree index is constructed for all data records, the hash table is built on the fly based on runtime workload, only data records accessed by point queries are indexed using hash table, this helps reduce the memory footprint. Changes to hash table are done using compare-and-swap (CAS) without performing locking and logging, this helps improve the concurrency and avoid contention. The hash table is also optimized to be cache conscious. HCB index is implemented in SAP ASE database, compared with the standard B-tree index, early experiments and customer adoptions show significant performance improvement. This paper provides an overview of the design of HCB index and reports the experimental results.

Keywords: B-tree, compare-and-swap, lock-free hash table, point queries, range queries, SAP ASE database

Procedia PDF Downloads 254

Authors: Hamza Zafar, Quratulain Rajput

Abstract:

Salah (نماز ) is one of five pillars of Islam and obligatory for every Muslims. However, due to the lack of Islamic knowledge it might be very difficult for a layperson to perform it correctly. This paper presents an ontology based representation of Islamic rules to perform Salah. The Salah ontology has been built under the guidance of domain expert in light of Quran and Hadith. The ontology consists of basic concepts as well as relationship among concepts and constraints on them. The basic concepts include cleanness, body cover, Salah timing and steps to perform Salah. The SWRL rule language has been used to represent rule to determine whether the Salah performed correctly or it should be repeated. Finally, we evaluate the use of the Salat ontology through user’s example queries using SPARQL queries.

Keywords: prayer, salah, ontology, SPARQL queries, reasoning

Procedia PDF Downloads 383

Authors: Laurent Thiry, Michel Hassenforder

Abstract:

This paper presents a general approach to implement efficient queries’ interpreters in a functional programming language. Indeed, most of the standard tools actually available use an imperative and/or object-oriented language for the implementation (e.g. Java for Jena-Fuseki) but other paradigms are possible with, maybe, better performances. To proceed, the paper first explains how to model data structures and queries in a functional point of view. Then, it proposes a general methodology to get performances (i.e. number of computation steps to answer a query) then it explains how to integrate some optimization techniques (short-cut fusion and, more important, data transformations). It then compares the functional server proposed to a standard tool (Fuseki) demonstrating that the first one can be twice to ten times faster to answer queries.

Keywords: data transformation, functional programming, information server, optimization

Procedia PDF Downloads 121

Authors: Simon Keller, Rainer Mueller

Abstract:

Continuous range queries are a common means to handle mobile clients in high-density areas. Most existing approaches focus on settings in which the range queries for location-based services are more or less static, whereas the mobile clients in the ranges move. We focus on a category called dynamic real-time range queries (DRRQ), assuming that both, clients requested by the query and the inquirers, are mobile. In consequence, the query parameters and the query results continuously change. This leads to two requirements: the ability to deal with an arbitrarily high number of mobile nodes (scalability) and the real-time delivery of range query results. In this paper, we present the highly decentralized solution adaptive quad streaming (AQS) for the requirements of DRRQs. AQS approximates the query results in favor of a controlled real-time delivery and guaranteed scalability. While prior works commonly optimize data structures on the involved servers, we use AQS to focus on a highly distributed cell structure without data structures automatically adapting to changing client distributions. Instead of the commonly used request-response approach, we apply a lightweight streaming method in which no bidirectional communication and no storage or maintenance of queries are required at all.

Keywords: approximation of client distributions, continuous spatial range queries, mobile objects, streaming-based decentralization in spatial mobile environments

Procedia PDF Downloads 111

Authors: Elimhan N. Mahmudov

Abstract:

The paper concerns the necessary and sufficient conditions of optimality for Cauchy problem of fourth order discrete (PD) and discrete-approximate (PDA) inclusions. The main problem is formulation of the fourth order adjoint discrete and discrete-approximate inclusions and transversality conditions, which are peculiar to problems including fourth order derivatives and approximate derivatives. Thus the necessary and sufficient conditions of optimality are obtained incorporating the Euler-Lagrange and Hamiltonian forms of inclusions. Derivation of optimality conditions are based on the apparatus of locally adjoint mapping (LAM). Moreover in the application of these results we consider the fourth order linear discrete and discrete-approximate inclusions.

Keywords: difference, optimization, fourth, approximation, transversality

Procedia PDF Downloads 343

Authors: Anandhi Santhosh

Abstract:

In order to describe various real-world problems in physical and engineering sciences subject to abrupt changes at certain instants during the evolution process, impulsive differential equations has been used to describe the system model. In this article, the problem of approximate controllability for nonlinear impulsive integrodifferential equations with state-dependent delay is investigated. We study the approximate controllability for nonlinear impulsive integrodifferential system under the assumption that the corresponding linear control system is approximately controllable. Using methods of functional analysis and semigroup theory, sufficient conditions are formulated and proved. Finally, an example is provided to illustrate the proposed theory.

Keywords: approximate controllability, impulsive differential system, fixed point theorem, state-dependent delay

Procedia PDF Downloads 353

Authors: Prasanna Haddela

Abstract:

Document analysis is a well matured yet still active research field, partly as a result of the intricate nature of building computational tools but also due to the inherent problems arising from the variety and complexity of human languages. Breaking down language barriers is vital in enabling access to a number of recent technologies. This paper investigates the application of document classification methods to new Sinhalese datasets. This language is geographically isolated and rich with many of its own unique features. We will examine the interpretability of the classification models with a particular focus on the use of evolved Lucene search queries generated using a Genetic Algorithm (GA) as a method of document classification. We will compare the accuracy and interpretability of these search queries with other popular classifiers. The results are promising and are roughly in line with previous work on English language datasets.

Keywords: evolved search queries, Sinhala document classification, Lucene Sinhala analyzer, interpretable text classification, genetic algorithm

Procedia PDF Downloads 86

Authors: Zhijie Ma, Qinglin Zhao, Hongning Dai, Huan Zhang

Abstract:

This paper proposes an APPLE scheme that aims at providing absolute and proportional throughput guarantees, and maximizing system throughput simultaneously for wireless LANs with homogeneous and heterogenous traffic. We formulate our objectives as an optimization problem, present its exact and approximate solutions, and prove the existence and uniqueness of the approximate solution. Simulations validate that APPLE scheme is accurate, and the approximate solution can well achieve the desired objectives already.

Keywords: IEEE 802.11e, throughput guarantee, priority, WLANs

Procedia PDF Downloads 325

Authors: Yu Liang, Wei Wei

Abstract:

Spatial representation of numbers and numerical magnitude are important for preschoolers’ mathematical ability. Mental number line, a typical index to measure numbers spatial representation, and numerical comparison are both related to arithmetic obviously. However, they seem to rely on different mechanisms and probably influence arithmetic through different mechanisms. In line with this idea, preschool children were trained with two tasks to investigate which one is more important for approximate arithmetic. The training of numerical processing and number line estimation were proved to be effective. They both improved the ability of approximate arithmetic. When the difficulty of approximate arithmetic was taken into account, the performance in number line training group was not significantly different among three levels. However, two harder levels achieved significance in numerical comparison training group. Thus, comparing spatial representation ability, symbolic approximation arithmetic relies more on numerical magnitude. Educational implications of the study were discussed.

Keywords: approximate arithmetic, mental number line, numerical magnitude, preschooler

Procedia PDF Downloads 216

Authors: Abdelaziz Fellah, Allaoua Maamir

Abstract:

We propose a domain-independent merging-cluster filter approach complemented with a set of algorithms for identifying approximate duplicate entities efficiently and accurately within a single and across multiple data sources. The near-optimal merging-cluster filter (MCF) approach is based on the Monge-Elkan well-tuned algorithm and extended with an affine variant of the Smith-Waterman similarity measure. Then we present constant, variable, and function threshold algorithms that work conceptually in a divide-merge filtering fashion for detecting near duplicates as hierarchical clusters along with their corresponding representatives. The algorithms take recursive refinement approaches in the spirit of filtering, merging, and updating, cluster representatives to detect approximate duplicates at each level of the cluster tree. Experiments show a high effectiveness and accuracy of the MCF approach in detecting approximate duplicates by outperforming the seminal Monge-Elkan’s algorithm on several real-world benchmarks and generated datasets.

Keywords: data mining, data cleaning, approximate duplicates, near-duplicates detection, data mining applications and discovery

Procedia PDF Downloads 355

Authors: Subha D. Puthankattil, Paul K. Joseph

Abstract:

Analyzing brain signals of the patients suffering from the state of depression may lead to interesting observations in the signal parameters that is quite different from a normal control. The present study adopts two different methods: Time frequency domain and nonlinear method for the analysis of EEG signals acquired from depression patients and age and sex matched normal controls. The time frequency domain analysis is realized using wavelet entropy and approximate entropy is employed for the nonlinear method of analysis. The ability of the signal processing technique and the nonlinear method in differentiating the physiological aspects of the brain state are revealed using Wavelet entropy and Approximate entropy.

Keywords: EEG, depression, wavelet entropy, approximate entropy, relative wavelet energy, multiresolution decomposition

Procedia PDF Downloads 295

Authors: Manana Chumburidze, David Lekveishvili

Abstract:

We have considered the harmonic oscillations and general dynamic (pseudo oscillations) systems of theory generalized Green-Lindsay of couple-stress thermo-elasticity for isotropic, homogeneous elastic media. Approximate solution of some mixed boundary value problems for finite domain, bounded by the some closed surface are constructed.

Keywords: the couple-stress thermoelasticity, boundary value problems, dynamic problems, approximate solution

Procedia PDF Downloads 473

Authors: Basant K. Jha, M. L. Kaurangini

Abstract:

An approximate analytical solution is presented for convective flow in a horizontal channel filled with porous material. The Brinkman-Forchheimer extension of Darcy equation is utilized to model the fluid flow while the energy equation is utilized to model temperature distribution in the channel. The solutions were obtained utilizing the newly suggested technique and compared with those obtained from an implicit finite-difference solution.

Keywords: approximate analytical, convective flow, porous material, Brinkman-Forchiemer

Procedia PDF Downloads 348

Authors: Ramdas Sonawane, Mahaveer Gadiya

Abstract:

The control problem associated with nuclear dynamics is represented by nonlinear integro-differential equation with additive controls. To control chain reaction, certain amount of neutrons is added into (or withdrawn out of) chamber as and when required. It is not realistic. So, we can think of controlling the reactor dynamics by bilinear control, which enters the system as coefficient of state. In this paper, we study the approximate and exact controllability of parabolic integro-differential equation controlled by bilinear control with non-homogeneous boundary conditions in bounded domain. We prove the existence of control and propose an explicit control strategy.

Keywords: approximate control, exact control, bilinear control, nuclear dynamics, integro-differential equations

Procedia PDF Downloads 406

Authors: Montaceur Zaghdoud, Mohamed Moussaoui, Jalel Akaichi

Abstract:

Frequent subgraph mining refers usually to graph matching and it is widely used in when analyzing big data with large graphs. A lot of research works dealt with structural exact or inexact graph matching but a little attention is paid to semantic matching when graph vertices and/or edges are attributed and typed. Therefore, it seems very interesting to integrate background knowledge into the analysis and that extracted frequent subgraphs should become more pruned by applying a new semantic filter instead of using only structural similarity in graph matching process. Consequently, this paper focuses on developing a new hybrid approximate structuralsemantic graph matching to discover a set of frequent subgraphs. It uses simultaneously an approximate structural similarity function based on graph edit distance function and a possibilistic vertices similarity function based on affinity function. Both structural and semantic filters contribute together to prune extracted frequent set. Indeed, new hybrid structural-semantic frequent subgraph mining approach searches will be suitable to be applied to several application such as community detection in social networks.

Keywords: approximate graph matching, hybrid frequent subgraph mining, graph mining, possibility theory

Procedia PDF Downloads 364

Authors: Meaghan Bowman

Abstract:

Soldiers are taught about the importance of effective communication with repetition of the phrase, “Communication is key.” They receive training in preparing for, and carrying out, interactions between foreign and domestic leaders to gain crucial information about a mission. These interactions are known as Key Leader Engagements (KLEs). For the training of KLEs, doctrine mandates the skills needed to conduct these “engagements” such as how to: behave appropriately, identify key leaders, and employ effective strategies. Army officers in training learn how to confront leaders, what information to gain, and how to ask questions respectfully. Unfortunately, soldiers rarely learn how to formulate questions optimally. Since less complex questions are easier to understand, we hypothesize that semantic complexity affects content understanding, and that age and education levels may have an effect on one’s ability to form paraphrases and judge their quality. In this study, we looked at paraphrases of queries as well as judgments of both the paraphrases’ naturalness and their semantic similarity to the query. Queries were divided into three complexity categories based on the number of relations (the first number) and the number of knowledge graph edges (the second number). Two crowd-sourced tasks were completed by Amazon volunteer participants, also known as turkers, to answer the research questions: (i) Are more complex queries harder to paraphrase and judge and (ii) Do age and education level affect the ability to understand complex queries. We ran statistical tests as follows: MANOVA for query understanding and two-way ANOVA to understand the relationship between query complexity and education and age. A probe of the number of given-level queries selected for paraphrasing by crowd-sourced workers in seven age ranges yielded promising results. We found significant evidence that age plays a role and marginally significant evidence that education level plays a role. These preliminary tests, with output p-values of 0.0002 and 0.068, respectively, suggest the importance of content understanding in a communication skill set. This basic ability to communicate, which may differ by age and education, permits reproduction and quality assessment and is crucial in training soldiers for effective participation in KLEs.

Keywords: engagement, key leader, paraphrasing, query complexity, understanding

Procedia PDF Downloads 126

Authors: Sarisa Pinkham, Kanyarat Bussaban

Abstract:

The research aims to approximate the amount of daily rainfall by using a pixel value data approach. The daily rainfall maps from the Thailand Meteorological Department in period of time from January to December 2013 were the data used in this study. The results showed that this approach can approximate the amount of daily rainfall with RMSE=3.343.

Keywords: daily rainfall, image processing, approximation, pixel value data

Procedia PDF Downloads 348

Authors: J. Suneetha, Vijayalaxmi

Abstract:

Sequential Pattern Mining involves applying data mining methods to large data repositories to extract usage patterns. Sequential pattern mining methodologies used to analyze the data and identify patterns. The patterns have been used to implement efficient systems can recommend on previously observed patterns, in making predictions, improve usability of systems, detecting events, and in general help in making strategic product decisions. In this paper, identified performance of approximate sequential pattern mining defines as identifying patterns approximately shared with many sequences. Approximate sequential patterns can effectively summarize and represent the databases by identifying the underlying trends in the data. Conducting an extensive and systematic performance over synthetic and real data. The results demonstrate that ApproxMAP effective and scalable in mining large sequences databases with long patterns.

Keywords: multiple data, performance analysis, sequential pattern, sequence database scalability

Procedia PDF Downloads 302

Authors: A. Elbyed

Abstract:

Nowadays, ontologies are used for achieving a common understanding within a user community and for sharing domain knowledge. However, the de-centralized nature of the web makes indeed inevitable that small communities will use their own ontologies to describe their data and to index their own resources. Certainly, accessing to resources from various ontologies created independently is an important challenge for answering end user queries. Ontology mapping is thus required for combining ontologies. However, mapping complete ontologies at run time is a computationally expensive task. This paper proposes a system in which mappings between concepts may be generated dynamically as the concepts are encountered during user queries. In this way, the interaction itself defines the context in which small and relevant portions of ontologies are mapped. We illustrate application of the proposed system in the context of Technology Enhanced Learning (TEL) where learners need to access to learning resources covering specific concepts.

Keywords: resources query, ontologies, ontology mapping, similarity measures, semantic web, e-learning

Procedia PDF Downloads 277

Authors: Jian-Jun Shu

Abstract:

A cold, thin film of liquid impinging on an isothermal hot, horizontal surface has been investigated. An approximate solution for the velocity and temperature distributions in the flow along the horizontal surface is developed, which exploits the hydrodynamic similarity solution for thin film flow. The approximate solution may provide a valuable basis for assessing flow and heat transfer in more complex settings.

Keywords: flux, free impinging jet, solid-surface, uniform wall temperature

Procedia PDF Downloads 441

Authors: William Chung

Abstract:

This paper is to study the use of multiple subproblems in Dantzig-Wolfe decomposition of linear programming (DW-LP). Traditionally, the decomposed LP consists of one LP master problem and one LP subproblem. The master problem and the subproblem is solved alternatively by exchanging the dual prices of the master problem and the proposals of the subproblem until the LP is solved. It is well known that convergence is slow with a long tail of near-optimal solutions (asymptotic convergence). Hence, the performance of DW-LP highly depends upon the number of decomposition steps. If the decomposition steps can be greatly reduced, the performance of DW-LP can be improved significantly. To reduce the number of decomposition steps, one of the methods is to increase the number of proposals from the subproblem to the master problem. To do so, we propose to add a quadratic approximation function to the LP subproblem in order to develop a set of approximate-LP subproblems (multiple subproblems). Consequently, in each decomposition step, multiple subproblems are solved for providing multiple proposals to the master problem. The number of decomposition steps can be reduced greatly. Note that each approximate-LP subproblem is nonlinear programming, and solving the LP subproblem must faster than solving the nonlinear multiple subproblems. Hence, using multiple subproblems in DW-LP is the tradeoff between the number of approximate-LP subproblems being formed and the decomposition steps. In this paper, we derive the corresponding algorithms and provide some simple computational results. Some properties of the resulting algorithms are also given.

Keywords: approximate subproblem, Dantzig-Wolfe decomposition, large-scale models, multiple subproblems

Procedia PDF Downloads 126

Authors: Wafa’ Alma'Aitah, Khaled Almakadmeh

Abstract:

this study presents an information retrieval system of using genetic algorithm to increase information retrieval efficiency. Using vector space model, information retrieval is based on the similarity measurement between query and documents. Documents with high similarity to query are judge more relevant to the query and should be retrieved first. Using genetic algorithms, each query is represented by a chromosome; these chromosomes are fed into genetic operator process: selection, crossover, and mutation until an optimized query chromosome is obtained for document retrieval. Results show that information retrieval with adaptive crossover probability and single point type crossover and roulette wheel as selection type give the highest recall. The proposed approach is verified using (242) proceedings abstracts collected from the Saudi Arabian national conference.

Keywords: genetic algorithm, information retrieval, optimal queries, crossover

Procedia PDF Downloads 260

Authors: Mahmood Otadi, Maryam Mosleh

Abstract:

The hybrid differential equations have a wide range of applications in science and engineering. In this paper, the homotopy analysis method (HAM) is applied to obtain the series solution of the hybrid differential equations. Using the homotopy analysis method, it is possible to find the exact solution or an approximate solution of the problem. Comparisons are made between improved predictor-corrector method, homotopy analysis method and the exact solution. Finally, we illustrate our approach by some numerical example.

Keywords: fuzzy number, fuzzy ODE, HAM, approximate method

Procedia PDF Downloads 476

Authors: Rui Yin, Haojie Wang, Xun Li

Abstract:

Recently, mainstream dense retrieval methods have obtained state-of-the-art results on some datasets and tasks. However, they require large amounts of training data, which is not available in most domains. The severe performance degradation of dense retrievers on new data domains has limited the use of dense retrieval methods to only a few domains with large training datasets. In this paper, we propose an unsupervised domain-adaptive approach based on query generation. First, a generative model is used to generate relevant queries for each passage in the target corpus, and then the generated queries are used for mining negative passages. Finally, the query-passage pairs are labeled with a cross-encoder and used to train a domain-adapted dense retriever. Experiments show that our approach is more robust than previous methods in target domains that require less unlabeled data.

Keywords: dense retrieval, query generation, unsupervised training, text retrieval

Procedia PDF Downloads 34

Authors: Hamdi Amroun, Yacine Benziani, Mehdi Ammi

Abstract:

In this paper, we propose an approach of detecting the behavior of the viewers of a TV program in a non-controlled environment. The experiment we propose is based on the use of three types of connected objects (smartphone, smart watch, and a connected remote control). 23 participants were observed while watching their TV programs during three phases: before, during and after watching a TV program. Their behaviors were detected using an approach based on The Dempster Shafer Theory (DST) in two phases. The first phase is to approximate dynamically the mass functions using an approach based on the correlation coefficient. The second phase is to calculate the approximate mass functions. To approximate the mass functions, two approaches have been tested: the first approach was to divide each features data space into cells; each one has a specific probability distribution over the behaviors. The probability distributions were computed statistically (estimated by empirical distribution). The second approach was to predict the TV-viewing behaviors through the use of classifiers algorithms and add uncertainty to the prediction based on the uncertainty of the model. Results showed that mixing the fusion rule with the computation of the initial approximate mass functions using a classifier led to an overall of 96%, 95% and 96% success rate for the first, second and third TV-viewing phase respectively. The results were also compared to those found in the literature. This study aims to anticipate certain actions in order to maintain the attention of TV viewers towards the proposed TV programs with usual connected objects, taking into account the various uncertainties that can be generated.

Keywords: Iot, TV-viewing behaviors identification, automatic classification, unconstrained environment

Procedia PDF Downloads 194