Development of Information Extraction System from Multi Source Unstructured Documents for Knowledge Base Expansion (지식베이스 확장을 위한 멀티소스 비정형 문서에서의 정보 추출 시스템의 개발)
-
- Journal of Intelligence and Information Systems
- /
- v.24 no.4
- /
- pp.111-136
- /
- 2018
In this paper, we propose a methodology to extract answer information about queries from various types of unstructured documents collected from multi-sources existing on web in order to expand knowledge base. The proposed methodology is divided into the following steps. 1) Collect relevant documents from Wikipedia, Naver encyclopedia, and Naver news sources for "subject-predicate" separated queries and classify the proper documents. 2) Determine whether the sentence is suitable for extracting information and derive the confidence. 3) Based on the predicate feature, extract the information in the proper sentence and derive the overall confidence of the information extraction result. In order to evaluate the performance of the information extraction system, we selected 400 queries from the artificial intelligence speaker of SK-Telecom. Compared with the baseline model, it is confirmed that it shows higher performance index than the existing model. The contribution of this study is that we develop a sequence tagging model based on bi-directional LSTM-CRF using the predicate feature of the query, with this we developed a robust model that can maintain high recall performance even in various types of unstructured documents collected from multiple sources. The problem of information extraction for knowledge base extension should take into account heterogeneous characteristics of source-specific document types. The proposed methodology proved to extract information effectively from various types of unstructured documents compared to the baseline model. There is a limitation in previous research that the performance is poor when extracting information about the document type that is different from the training data. In addition, this study can prevent unnecessary information extraction attempts from the documents that do not include the answer information through the process for predicting the suitability of information extraction of documents and sentences before the information extraction step. It is meaningful that we provided a method that precision performance can be maintained even in actual web environment. The information extraction problem for the knowledge base expansion has the characteristic that it can not guarantee whether the document includes the correct answer because it is aimed at the unstructured document existing in the real web. When the question answering is performed on a real web, previous machine reading comprehension studies has a limitation that it shows a low level of precision because it frequently attempts to extract an answer even in a document in which there is no correct answer. The policy that predicts the suitability of document and sentence information extraction is meaningful in that it contributes to maintaining the performance of information extraction even in real web environment. The limitations of this study and future research directions are as follows. First, it is a problem related to data preprocessing. In this study, the unit of knowledge extraction is classified through the morphological analysis based on the open source Konlpy python package, and the information extraction result can be improperly performed because morphological analysis is not performed properly. To enhance the performance of information extraction results, it is necessary to develop an advanced morpheme analyzer. Second, it is a problem of entity ambiguity. The information extraction system of this study can not distinguish the same name that has different intention. If several people with the same name appear in the news, the system may not extract information about the intended query. In future research, it is necessary to take measures to identify the person with the same name. Third, it is a problem of evaluation query data. In this study, we selected 400 of user queries collected from SK Telecom 's interactive artificial intelligent speaker to evaluate the performance of the information extraction system. n this study, we developed evaluation data set using 800 documents (400 questions * 7 articles per question (1 Wikipedia, 3 Naver encyclopedia, 3 Naver news) by judging whether a correct answer is included or not. To ensure the external validity of the study, it is desirable to use more queries to determine the performance of the system. This is a costly activity that must be done manually. Future research needs to evaluate the system for more queries. It is also necessary to develop a Korean benchmark data set of information extraction system for queries from multi-source web documents to build an environment that can evaluate the results more objectively.
For a long time, many studies have been conducted on predicting the success of campaigns for customers in academia, and prediction models applying various techniques are still being studied. Recently, as campaign channels have been expanded in various ways due to the rapid revitalization of online, various types of campaigns are being carried out by companies at a level that cannot be compared to the past. However, customers tend to perceive it as spam as the fatigue of campaigns due to duplicate exposure increases. Also, from a corporate standpoint, there is a problem that the effectiveness of the campaign itself is decreasing, such as increasing the cost of investing in the campaign, which leads to the low actual campaign success rate. Accordingly, various studies are ongoing to improve the effectiveness of the campaign in practice. This campaign system has the ultimate purpose to increase the success rate of various campaigns by collecting and analyzing various data related to customers and using them for campaigns. In particular, recent attempts to make various predictions related to the response of campaigns using machine learning have been made. It is very important to select appropriate features due to the various features of campaign data. If all of the input data are used in the process of classifying a large amount of data, it takes a lot of learning time as the classification class expands, so the minimum input data set must be extracted and used from the entire data. In addition, when a trained model is generated by using too many features, prediction accuracy may be degraded due to overfitting or correlation between features. Therefore, in order to improve accuracy, a feature selection technique that removes features close to noise should be applied, and feature selection is a necessary process in order to analyze a high-dimensional data set. Among the greedy algorithms, SFS (Sequential Forward Selection), SBS (Sequential Backward Selection), SFFS (Sequential Floating Forward Selection), etc. are widely used as traditional feature selection techniques. It is also true that if there are many risks and many features, there is a limitation in that the performance for classification prediction is poor and it takes a lot of learning time. Therefore, in this study, we propose an improved feature selection algorithm to enhance the effectiveness of the existing campaign. The purpose of this study is to improve the existing SFFS sequential method in the process of searching for feature subsets that are the basis for improving machine learning model performance using statistical characteristics of the data to be processed in the campaign system. Through this, features that have a lot of influence on performance are first derived, features that have a negative effect are removed, and then the sequential method is applied to increase the efficiency for search performance and to apply an improved algorithm to enable generalized prediction. Through this, it was confirmed that the proposed model showed better search and prediction performance than the traditional greed algorithm. Compared with the original data set, greed algorithm, genetic algorithm (GA), and recursive feature elimination (RFE), the campaign success prediction was higher. In addition, when performing campaign success prediction, the improved feature selection algorithm was found to be helpful in analyzing and interpreting the prediction results by providing the importance of the derived features. This is important features such as age, customer rating, and sales, which were previously known statistically. Unlike the previous campaign planners, features such as the combined product name, average 3-month data consumption rate, and the last 3-month wireless data usage were unexpectedly selected as important features for the campaign response, which they rarely used to select campaign targets. It was confirmed that base attributes can also be very important features depending on the type of campaign. Through this, it is possible to analyze and understand the important characteristics of each campaign type.
Laboratory-scale experiments were performed to identify the As removal mechanism of the residual slag generated after the mineral carbonation process. The residual slags were manufactured from the steelmaking slag (blast oxygen furnace slag: BOF) through direct and indirect carbonation process. RDBOF (residual BOF after the direct carbonation) and RIBOF (residual BOF after the indirect carbonation) showed different physicochemical-structural characteristics compared with raw BOF such as chemical-mineralogical properties, the pH level of leachate and forming micropores on the surface of the slag. In batch experiment, 0.1 g of residual slag was added to 10 mL of As-solution (initial concentration: 203.6 mg/L) titrated at various pH levels. The RDBOF showed 99.3% of As removal efficiency at initial pH 1, while it sharply decreased with the increase of initial pH. As the initial pH of solution decreased, the dissolution of carbonate minerals covering the surface was accelerated, increasing the exposed area of Fe-oxide and promoting the adsorption of As-oxyanions on the RDBOF surface. Whereas, the As removal efficiency of RIBOF increased with the increase of initial pH levels, and it reached up to 70% at initial pH 10. Considering the PZC (point of zero charge) of the RIBOF (pH 4.5), it was hardly expected that the electrical adsorption of As-oxyanion on surface of the RIBOF at initial pH of 4-10. Nevertheless it was observed that As-oxyanion was linked to the Fe-oxide on the RIBOF surface by the cation bridge effect of divalent cations such as Ca2+, Mn2+, and Fe2+. The surface of RIBOF became stronger negatively charged, the cation bridge effect was more strictly enforced, and more As can be fixed on the RIBOF surface. However, the Ca-products start to precipitate on the surface at pH 10-11 or higher and they even prevent the surface adsorption of As-oxyanion by Fe-oxide. The TCLP test was performed to evaluate the stability of As fixed on the surface of the residual slag after the batch experiment. Results supported that RDBOF and RIBOF firmly fixed As over the wide pH levels, by considering their As desorption rate of less than 2%. From the results of this study, it was proved that both residual slags can be used as an eco-friendly and low-cost As remover with high As removal efficiency and high stability and they also overcome the pH increase in solution, which is the disadvantage of existing steelmaking slag as an As remover.
Recently strategic alliance between business firms has become prevalent to overcome increasing competitive threats and to supplement resource limitation of individual firms. As one of allianced sales promotion activities, a new type of discount program, so called "Alliance Card Discount", is introduced with the partnership of credit cards and loyalty cards. The program mainly pursues short-term sales growth by larger discount scheme while spends less through cost share among alliance partners. Thus this program can be regarded as cost efficient discount promotion. But because there is no solid evidence that it can really deliver profitable sales growth, an empirical study for its effects on sales and profit should be conducted. This study has two basic research questions concerning the effects of allianced discount program ; 1)the possibility of sales increase 2) the profitability of the discount driven sales. In F&B industry, sales increase mainly comes from increased guest count. Especially in family restaurants, to increase the number of guests we need to enlarge the size of visitor group (number of visitors for one group) because customers visit by group in a special occasion. And because they pay the bill by group(table), the increase of sales per table is a key measure for sales improvement. The past researches for price & discount sensitivity and reference discount rate explain that price sensitive consumers have narrow reference discount zone and make rational purchase decision. Differently from all time discount scheme of regular sales promotions, the alliance card discount program only provides the right to get discount like discount coupon. And because it is usually once a month opportunity given by the past month usage level, customers tend to perceive alliance card discount as a rare chance to get. So that we can expect customers try to maximize the discount effect when they use the limited discount opportunity. Considering group visiting practice and low visit frequency of family restaurants, the way to maximize discount effect should be the increase the size of visit group. And their sensitivity to discount and rational consumption behavior defer the additional spending for ordering high price menu, even though they get considerable amount of savings from the discount. From the analysis of sales data paid by alliance discount cards for four months, we found the below. 1) The relation between discount rate and number of guest per table is positive : 25% discount results one additional guest 2) The relation between discount rate and the spending per guest is negative. 3) However, total profit amount per table is increased when discount rate is increased. 4) Reward point accumulation & redemption did not show any significant relationship with the increase of number of guests. These results suggest that the allianced discount program substantially contributes to sales increase and profit improvement by increasing the number of guests per table. Though the spending per guest is decreased by discount rate increase, the total amount of profit per table is improved. It seems the incremental profit by increased guest count offsets the profit decrease. Additional intriguing finding is the point reward system does not have any significant impact on the increase of number of guest, even if the point accumulation & redemption of loyalty program are usually regarded as another saving offers by customers. In sum, because it is proved that allianced discount program with credit cards and loyalty cards is effective to both sales drive and profit increase, the alliance card program could be recommended as strategically buyable program.
An atmospheric correction algorithm based on the radiative transfer model is required to obtain remote-sensing reflectance (Rrs) from the Geostationary Ocean Color Imager-II (GOCI-II) observed at the top-of-atmosphere. This Rrs derived from the atmospheric correction is utilized to estimate various marine environmental parameters such as chlorophyll-a concentration, total suspended materials concentration, and absorption of dissolved organic matter. Therefore, an atmospheric correction is a fundamental algorithm as it significantly impacts the reliability of all other color products. However, in clear waters, for example, atmospheric path radiance exceeds more than ten times higher than the water-leaving radiance in the blue wavelengths. This implies atmospheric correction is a highly error-sensitive process with a 1% error in estimating atmospheric radiance in the atmospheric correction process can cause more than 10% errors. Therefore, the quality assessment of Rrs after the atmospheric correction is essential for ensuring reliable ocean environment analysis using ocean color satellite data. In this study, a Quality Assurance (QA) algorithm based on in-situ Rrs data, which has been archived into a database using Sea-viewing Wide Field-of-view Sensor (SeaWiFS) Bio-optical Archive and Storage System (SeaBASS), was applied and modified to consider the different spectral characteristics of GOCI-II. This method is officially employed in the National Oceanic and Atmospheric Administration (NOAA)'s ocean color satellite data processing system. It provides quality analysis scores for Rrs ranging from 0 to 1 and classifies the water types into 23 categories. When the QA algorithm is applied to the initial phase of GOCI-II data with less calibration, it shows the highest frequency at a relatively low score of 0.625. However, when the algorithm is applied to the improved GOCI-II atmospheric correction results with updated calibrations, it shows the highest frequency at a higher score of 0.875 compared to the previous results. The water types analysis using the QA algorithm indicated that parts of the East Sea, South Sea, and the Northwest Pacific Ocean are primarily characterized as relatively clear case-I waters, while the coastal areas of the Yellow Sea and the East China Sea are mainly classified as highly turbid case-II waters. We expect that the QA algorithm will support GOCI-II users in terms of not only statistically identifying Rrs resulted with significant errors but also more reliable calibration with quality assured data. The algorithm will be included in the level-2 flag data provided with GOCI-II atmospheric correction.
Introduction As consumers' purchase behavior change into a rational and practical direction, the discount store industry came to have keen competition along with rapid external growth. Therefore as a solution, distribution businesses are concentrating on developing PB(Private Brand) which can realize differentiation and profitability at the same time. And as improvement in customer loyalty beyond customer satisfaction is effective in surviving in an environment with keen competition, PB is being used as a strategic tool to improve customer loyalty. To improve loyalty among PB users, it is necessary to develop PB by examining properties of a customer group, first of all, quality level perceived by consumers should be met to obtain customer satisfaction and customer trust and consequently induce customer loyalty. To provide results of systematic analysis on relations between antecedents influenced perceived quality and variables affecting customer loyalty, this study proposed a research model based on causal relations verified in prior researches and set 16 hypotheses about relations among 9 theoretical variables. Data was collected from 400 adult customers residing in Seoul and the Metropolitan area and using large scale discount stores, among them, 375 copies were analyzed using SPSS 15.0 and Amos 7.0. The findings of the present study followed as; We ascertained that the higher company reputation, brand reputation, product experience and brand familiarity, the higher perceived quality. The study also examined the higher perceived quality, the higher customer satisfaction, customer trust and customer loyalty. The findings showed that the higher customer satisfaction and customer trust, the higher customer loyalty. As for moderating effects between PB and NB in terms of influences of perceived quality factors on perceived quality, we can ascertain that PB was higher than NB in the influences of company reputation on perceived quality while NB was higher than PB in the influences of brand reputation and brand familiarity on perceived quality. These results of empirical analysis will be useful for those concerned to do marketing activities based on a clearer understanding of antecedents and consecutive factors influenced perceived quality. At last, discussions about academical and managerial implications in these results, we suggested the limitations of this study and the future research directions. Research Model and Hypotheses Test After analyzing if antecedent variables having influence on perceived quality shows any difference between PB and NB in terms of their influences on them, the relation between variables that have influence on customer loyalty was determined as Figure 1. We established 16 hypotheses to test and hypotheses are as follows; H1-1: Perceived price has a positive effect on perceived quality. H1-2: It is expected that PB and NB would have different influence in terms of perceived price on perceived quality. H2-1: Company reputation has a positive effect on perceived quality. H2-2: It is expected that PB and NB would have different influence in terms of company reputation on perceived quality. H3-1: Brand reputation has a positive effect on perceived quality. H3-2: It is expected that PB and NB would have different influence in terms of brand reputation on perceived quality. H4-1: Product experience has a positive effect on perceived quality. H4-2: It is expected that PB and NB would have different influence in terms of product experience on perceived quality. H5-1: Brand familiarity has a positive effect on perceived quality. H5-2: It is expected that PB and NB would have different influence in terms of brand familiarity on perceived quality. H6: Perceived quality has a positive effect on customer satisfaction. H7: Perceived quality has a positive effect on customer trust. H8: Perceived quality has a positive effect on customer loyalty. H9: Customer satisfaction has a positive effect on customer trust. H10: Customer satisfaction has a positive effect on customer loyalty. H11: Customer trust has a positive effect on customer loyalty. Results from analyzing main effects of research model is shown as