A R Rodolfo Pazos

Natural Language Interfaces to Databases: An Analysis of the State of the Art

People constantly make decisions based on information, most of which is stored in databases. Accessing this information requires the use of query languages to databases such as SQL. In order to avoid the difficulty of using these languages for users who are not computing experts, Natural Language Interfaces for Databases (NLIDB) have been developed, which permit to query databases through queries formulated in natural language. Although since the 60s many NLIDBs have been developed, their performance has not been satisfactory, there still remain very difficult problems that have not been solved by NLIDB technology, and there does not yet exist a standardized method of evaluation that permits to compare the performance of different NLIDBs. This chapter presents an analysis of NLIDBs, which includes their classification, techniques, advantages, disadvantages, and a proposal for a proper evaluation of them.

Comparison and Selection of Exact and Heuristic Algorithms

The traditional approach for comparing heuristic algorithms uses well-known statistical tests for meaningfully relating the empirical performance of the algorithms and concludes that one outperforms the other. In contrast, the method presented in this paper, builds a predictive model of the algorithms behavior using functions that relate performance to problem size, in order to define dominance regions. This method generates first a representative sample of the algorithms performance, then using a common and simplified regression analysis determines performance functions, which are finally incorporated into an algorithm selection mechanism. For testing purposes, a set of same-class instances of the database distribution problem was solved using an exact algorithm (Branch&Bound) and a heuristic algorithm (Simulated Annealing). Experimental results show that problem size affects differently both algorithms, in such a way that there exist regions where one algorithm is more efficient than the other.

A Machine Learning Approach for Modeling Algorithm Performance Predictors

This paper deals with heuristic algorithm selection, which can be stated as follows: given a set of solved instances of a NP-hard problem, for a new instance to predict which algorithm solves it better. For this problem, there are two main selection approaches. The first one consists of developing functions to relate performance to problem size. In the second more characteristics are incorporated, however they are not defined formally, neither systematically, In contrast, we propose a methodology to model algorithm performance predictors that incorporate critical characteristics.. The relationship among performance and characteristics is learned from historical data using machine learning techniques. To validate our approach we carried out experiments using an extensive test set. In particular, for the classical bin packing problem, we developed predictors that incorporate the interrelation among five critical characteristics and the performance of seven heuristic algorithms. We obtained an accuracy of 81% in the selection of the best algorithm.This paper deals with heuristic algorithm selection, which can be stated as follows: given a set of solved instances of a NP-hard problem, for a new instance to predict which algorithm solves it better. For this problem, there are two main selection approaches. The first one consists of developing functions to relate performance to problem size. In the second more characteristics are incorporated, however they are not defined formally, neither systematically. In contrast, we propose a methodology to model algorithm performance predictors that incorporate critical characteristics. The relationship among performance and characteristics is learned from historical data using machine learning techniques. To validate our approach we carried out experiments using an extensive test set. In particular, for the classical bin packing problem, we developed predictors that incorporate the interrelation among five critical characteristics and the performance of seven heuristic algorithms. We obtained an accuracy of 81% in the selection of the best algorithm.

Semantic model for improving the performance of natural language interfaces to databases

Despite the fact that since the late 60s many Natural Language Interfaces to Databases (NLIDBs) have been developed, up to now many problems continue, which prevent the translation process from natural language to SQL to be totally successful. Some of the main problems that have been encountered relate to 1) achieving domain independence, 2) the use of words or phrases of different syntactic categories for referring to tables and columns, and 3) semantic ellipsis. This paper introduces a new method for modeling databases that includes relevant information for improving the performance of NLIDBs. This method will be useful for solving many problems found in the translation from natural language to SQL, using a database model that contains linguistic information that provides more semantic information than that found in conventional database models (such as the extended entity-relationship model) and those used in previous NLIDBs.Despite the fact that since the late 60s many Natural Language Interfaces to Databases (NLIDBs) have been developed, up to now many problems continue, which prevent the translation process from natural language to SQL to be totally successful. Some of the main problems that have been encountered relate to 1) achieving domain independence, 2) the use of words or phrases of different syntactic categories for referring to tables and columns, and 3) semantic ellipsis. This paper introduces a new method for modeling databases that includes relevant information for improving the performance of NLIDBs. This method will be useful for solving many problems found in the translation from natural language to SQL, using a database model that contains linguistic information that provides more semantic information than that found in conventional database models (such as the extended entity-relationship model) and those used in previous NLIDBs.

Dialogue manager for a NLIDB for solving the semantic ellipsis problem in query formulation

A query written in natural language (NL) may involve several linguistic problems that cause a query not being interpreted or translated correctly into SQL. One of these problems is implicit information or semantic ellipsis, which can be understood as the omission of important words in the wording of a query written in NL. An exhaustive survey on NLIDB works has revealed that most of these works has not systematically dealt with semantic ellipsis. In experiments conducted on commercial NLIDBs, very poor results have been obtained (7% to 16.9%) when dealing with query corpora that involve semantic ellipsis. In this paper we propose a dialogue manager (DM) for a NLIDB for solving semantic ellipsis problems. The operation of this DM is based on a typification of elliptical problems found in queries, which permits to systematically deal with this problem. Additionally, the typification has two important characteristics: domain independence, which permits the typification to be applied to queries of different databases, and generality, which means that it holds for different languages such as English, French, Italian, Spanish, etc. These characteristics are inherited to the dialogue processes implemented in the DM, since they are based on this typification. In experiments conducted with this DM and a NLIDB on a corpus of elliptical queries, an increase of correctly answered queries of 30-35% was attained.

Adaptive Allocation of Data-Objects in the Web Using Neural Networks

In this paper we address the problem of allocation scheme design of large database-objects in the Web environment, which may suffer significant changes in usage and access patterns and scaling of data. In these circumstances, if the design is not adjusted to new changes, the system can undergo severe degradations in data access costs and response time. Since this problem is NP-complete, obtaining optimal solutions for large problem instances requires applying approximate methods. We present a mathematical model to generate a new object allocation scheme and propose a new method to solve it. The method uses a Hopfield neural network with the mean field annealing (MFA) variant. The experimental results and a comparative study with other two methods are presented. The new method has a similar capacity to solve large problem instances, regular level of solution quality and excellent execution time with respect to other methods.

Integer Linear Programming Formulation and Exact Algorithm for Computing Pathwidth

Computing the Pathwidth of a graph is the problem of finding a linear ordering of the vertices such that the width of its corresponding path decomposition is minimized. This problem has been proven to be NP-hard. Currently, some of the best exact methods for generic graphs can be found in the mathematical software project called SageMath. This project provides an integer linear programming model (IPSAGE) and an enumerative algorithm (EASAGE), which is exponential in time and space. The algorithm EASAGE uses an array whose size grows exponentially with respect to the size of the problem. The purpose of this array is to improve the performance of the algorithm. In this chapter we propose two exact methods for computing pathwidth. More precisely, we propose a new integer linear programming formulation (IPPW) and a new enumerative algorithm (BBPW). The formulation IPPW generates a smaller number of variables and constraints than IPSAGE. The algorithm BBPW overcomes the exponential space requirement by using a last-in-first-out stack. The experimental results showed that, in average, IPPW reduced the number of variables by 33.3 % and the number of constraints by 64.3 % with respect to IPSAGE. This reduction of variables and constraints allowed IPPW to save approximately 14.9 % of the computing time of IPSAGE. The results also revealed that BBPW achieved a remarkable use of memory with respect to EASAGE. In average, BBPW required 2073 times less amount of memory than EASAGE for solving the same set of instances.

Survey of polynomial transformations between NP-complete problems

This paper aims at being a guide to understand polynomial transformations and polynomial reductions between NP-complete problems by presenting the methodologies for polynomial reductions/transformations and the differences between reductions and transformations. To this end the article shows examples of polynomial reductions/transformations and the restrictions to reduce/transform between NP-complete problems. Finally, this paper includes a digraph with the historical reductions/transformations between instances of NP-complete problems and introduces the term family of polynomial transformations.

Modeling the Nonlinear Nature of Response Time in the Vertical Fragmentation Design of Distributed Databases

The generalized use of the Internet has facilitated the implementation of distributed database (DDB) systems, which are becoming increasingly common-place. Unfortunately, though there exist many models for optimizing the design of DDBs (i.e., the distribution of data), they usually seek to optimize the transmission and processing costs of queries and overlook the delays incurred by their transmission and processing, which can be a major concern for Internet-based systems. In this paper a mathematical model is presented, which describes the behavior of a DDB with vertical fragmentation and permits to optimize its design taking into account the roundtrip response time (query transmission time, query processing time, and response transmission time).

Dynamic Allocation of Data-Objects in the Web, Using Self-tuning Genetic Algorithms

In this paper, a new mechanism for automatically obtaining some control parameter values for Genetic Algorithms is presented, which is independent of problem domain and size. This approach differs from the traditional methods which require knowing the problem domain first, and then knowing how to select the parameter values for solving specific problem instances. The proposed method uses a sample of problem instances, whose solution allows to characterize the problem and to obtain the parameter values. To test the method, a combinatorial optimization model for data-object allocation in the Web (known as DFAR) was solved using Genetic Algorithms. We show how the proposed mechanism allows to develop a set of mathematical expressions that relates the problem instance size to the control parameters of the algorithm. The expressions are then used, in on-line process, to control the parameter values. We show the last experimental results with the self-tuning mechanism applied to solve a sample of random instances that simulates a typical Web workload. We consider that the proposed method principles must be extended to the self-tuning of control parameters for other heuristic algorithms.