《电子商务 E-business》阅读文献：Proposing a charting recommender system for second-language nurses

Expert Systems with Applications 38(2011)9281-9286 Contents lists available at ScienceDirect Expert Systems with Applications ELSEVIER journalhomepagewww.elsevier.com/locate/eswa Proposing a charting recommender system for second-language nurses Mei-Hua hsu x Department of General Education, Chang Gung institute of Technology, 261, Wen-Hwa Ist Road, Kwei-Shan, Taoyuan, Taiwan, ROC ARTICLE FO A BSTRACT poses a recommender system to assist second-language nur student nurses recommender system charting. Using data mining association rules, the system at ovides optimal Nursing care ns for the nurse to choose from as she or he proceeds with ti tion. The system, ested on Taiwanese nursing students who needed to chart Association rule nd to be highly effective for the intended purpose. Documentation e 2011 Elsevier Ltd. All rights reserved. 1 Introduction problem remains that nurses often encounter difficulties charting Patient care charting is the documentation by a nursing care This study therefore aims to create a recommender system that provider to allow continuity of care as well as the protection of assists second-language nurses, especially student nurses, in patient the legal interests of both patient and nurse. Recorded in the chart care charting. Using data mining association rules, the system auto- are the patients basic information and physical and mental condi- matically provides the optimal words or terms for the nurse to tions, the care plan, communication between patient and nurse and choose from as she or he proceeds with the documentation. The between the nurse and other medical professionals that contributes ommender corpus is based on the data collected from dictionaries of le patients care, and the care as well as services provided for nursing and books on nursing documentation of which the most the patient. The chart is very important information that must be renowned, for instance, are Mosby's Clinical Nursing(Thompson, accurately, fully, and concisely recorded. (Van der meijden, Tange. McFarland, Hirsch, Tucker, 2002). Mosby,s Medical, Nursing, and Troost, Hasman, 2001) Allied Health Dictionary(Anderson, Anderson, glanze, 2008). Approaches to nursing care documentation vary. The most Charting: An Incredibly Easy! Pocket Guide(Springhouse, 2006). as de en Susan Lampe's focus charting(Lam and Mosby,'s Surefire Documentation(Mosby, 2006). 1985: Lampe Hitchcock, 1987). Other methods that were The proposed recommender system does not deal with com- proposed in the 1980s include charting by exception(Burke mon errors in writing, as there are already many useful devices Murphy, 1988: Murphy, Beglinger, Johnson, 1988), the PlE sys- on their detection and correction. Genthial and Courtin(1992) tem(Buckley-Womack Gidney, 1987: Siegrist, Dettor, Stocks, present a detailed architecture for computer-aided writing that 985), and the CoRe system(Montemuro, 1988). In the 1990s, features detection and correction. Their system contains modules there were the VIPS model(Ehnfors, Thorell-Ekstrand, Ehren- capable of handling typographic errors, spelling errors, lexical er erg, 1991: Ehrenberg, Ehnfors, Thorell-Ekstrand, 1996). the rors, syntactic agreement, etc. When an error is made, the system FACT tool (Warne McWeeney, 1991), and the health care focus automatically suggests one or more corrections to the user. Scott documentation model(Scoates, Fishman, McAdam, 1996). And and New(1994)propose a computer-aided analysis of the foreign this century has seen the proposal of the Invocation Technique language writing process; they analyze system log data and sug (Navuluri, 2000)and concept mapping(Taylor Wros, 2007). gest how this information can be used in the classroom. Sullivan In Taiwan, focus charting is the most widely used method, as it and Pratt(1996), after a comparison of two ESL writing environ- is a highly efficient, patient-centered approach to monitoring pa- ments, a computer-assisted classroom and a traditional classroom tient problems without repetitious documentation. However, since find that the use of networked computers have far more positive the common practice in Taiwanese hospitals is to document nurs- effects in writing classrooms. Park, Palmer, and Washburn (1997) ing care in English instead of Chinese, the native language, the offer an English grammar checker as a writing aid for ESL students Elizabeth New(1999)argues that the lack of explicit revision instructions and computer strategies impede the reviewing and *Tel:+88632118999;fax:+88632118866. reworking of texts. Chodorow and Leacock(2000) propose an E-mail address: mhsuemailcgit edu. tw unsupervised method that detects grammatical errors without 0957-4174/s- see front matter o 2011 Elsevier Ltd. All rights reserved doi:10.1016/eswa2011.01.010

Proposing a charting recommender system for second-language nurses Mei-Hua Hsu ⇑ Department of General Education, Chang Gung Institute of Technology, 261, Wen-Hwa 1st Road, Kwei-Shan, Taoyuan, Taiwan, ROC article info Keywords: Recommender system Nursing care Charting Data mining Association rule Documentation abstract This paper proposes a recommender system to assist second-language nurses, especially student nurses, in patient care charting. Using data mining association rules, the system automatically provides optimal words or terms for the nurse to choose from as she or he proceeds with the documentation. The system, having been tested on Taiwanese nursing students who needed to chart in their practicum at hospitals, has been found to be highly effective for the intended purpose. 2011 Elsevier Ltd. All rights reserved. 1. Introduction Patient care charting is the documentation by a nursing care provider to allow continuity of care as well as the protection of the legal interests of both patient and nurse. Recorded in the chart are the patient’s basic information and physical and mental condi￾tions, the care plan, communication between patient and nurse and between the nurse and other medical professionals that contributes to the patient’s care, and the care as well as services provided for the patient. The chart is very important information that must be accurately, fully, and concisely recorded. (Van der Meijden, Tange, Troost, & Hasman, 2001). Approaches to nursing care documentation vary. The most popular has been Susan Lampe’s focus charting (Lampe, 1984, 1985; Lampe & Hitchcock, 1987). Other methods that were proposed in the 1980s include charting by exception (Burke & Murphy, 1988; Murphy, Beglinger, & Johnson, 1988), the PIE sys￾tem (Buckley-Womack & Gidney, 1987; Siegrist, Dettor, & Stocks, 1985), and the CORE system (Montemuro, 1988). In the 1990s, there were the VIPS model (Ehnfors, Thorell-Ekstrand, & Ehren￾berg, 1991; Ehrenberg, Ehnfors, & Thorell-Ekstrand, 1996), the FACT tool (Warne & McWeeney, 1991), and the health care focus documentation model (Scoates, Fishman, & McAdam, 1996). And this century has seen the proposal of the Invocation Technique (Navuluri, 2000) and concept mapping (Taylor & Wros, 2007). In Taiwan, focus charting is the most widely used method, as it is a highly efficient, patient-centered approach to monitoring pa￾tient problems without repetitious documentation. However, since the common practice in Taiwanese hospitals is to document nurs￾ing care in English instead of Chinese, the native language, the problem remains that nurses often encounter difficulties charting in English on the computer. This study therefore aims to create a recommender system that assists second-language nurses, especially student nurses, in patient care charting. Using data mining association rules, the system auto￾matically provides the optimal words or terms for the nurse to choose from as she or he proceeds with the documentation. The rec￾ommender corpus is based on the data collected from dictionaries of nursing and books on nursing documentation of which the most renowned, for instance, are Mosby’s Clinical Nursing (Thompson, McFarland, Hirsch, & Tucker, 2002), Mosby’s Medical, Nursing, and Allied Health Dictionary (Anderson, Anderson, & Glanze, 2008), Charting: An Incredibly Easy! Pocket Guide (Springhouse., 2006), and Mosby’s Surefire Documentation (Mosby, 2006). The proposed recommender system does not deal with com￾mon errors in writing, as there are already many useful devices on their detection and correction. Genthial and Courtin (1992) present a detailed architecture for computer-aided writing that features detection and correction. Their system contains modules capable of handling typographic errors, spelling errors, lexical er￾rors, syntactic agreement, etc. When an error is made, the system automatically suggests one or more corrections to the user. Scott and New (1994) propose a computer-aided analysis of the foreign language writing process; they analyze system log data and sug￾gest how this information can be used in the classroom. Sullivan and Pratt (1996), after a comparison of two ESL writing environ￾ments, a computer-assisted classroom and a traditional classroom, find that the use of networked computers have far more positive effects in writing classrooms. Park, Palmer, and Washburn (1997) offer an English grammar checker as a writing aid for ESL students. Elizabeth New (1999) argues that the lack of explicit revision instructions and computer strategies impede the reviewing and reworking of texts. Chodorow and Leacock (2000) propose an unsupervised method that detects grammatical errors without 0957-4174/$ - see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2011.01.010 ⇑ Tel.: +886 3 2118999; fax: +886 3 2118866. E-mail address: mhsu@mail.cgit.edu.tw Expert Systems with Applications 38 (2011) 9281–9286 Contents lists available at ScienceDirect Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa

9282 M.-H. Hsu/Ex doing any parsing: Chodorow, Tetreault, and Han (2007) the detection of grammatical errors involving prepositions. 2. Preliminar 2.1. Association rules In data mining. the association analysis, or the m the numbers large item se actions, and Supp Support(a→b)over can discover high c minimum support value and the threshold values decided useful, the lift value must be based on some

doing any parsing; Chodorow, Tetreault, and Han (2007) focus on the detection of grammatical errors involving prepositions. 2. Preliminary 2.1. Association rules In data mining, the association analysis, or the market-basket analysis, is mainly used to find out the relationships between items or features that occur synchronously in the database—i.e., to learn rules there may be for groceries purchased during a trip to the shopping center. For instance, 80% of the people who buy milk also buy bread as well. With such information, a decision maker can implement new strategies such as changing the positions of rele￾vant counters or organizing related promotions. Therefore, the main purpose of implementing the association rule algorithm is to find out synchronous relationships by analyzing random data and to use this data as reference during decision making. The asso￾ciation rule is defined as follows: Make I = {i1, i2, ... , im} as the item set, in which each item repre￾sents a specific commodity. D stands for a trading database in which each transaction ‘‘T’’ represents an item set. That is, T # I. Each item set is a non-empty sub-item set of ‘‘I,’’ and the only iden￾tifying code is TID. Each item set X I has a measuring standard, called support, to evaluate the statistical importance of D. Support (X, D) denotes the rate of merchandising X in transaction D. The association rule is X ? Y, in which X, Y I, and X \ Y ¼ ;. The rule means that if X is purchased, Y can be bought at the same time. Each rule has a measuring standard called Confidence; i.e., Confidence(X ? Y) = Support(X [ Y, D)/Support(X, D). In this case, Confidence(X ? Y) denotes that if the merchandise includes X, the chance of buying Y is relatively high. Two steps are needed to determine the association rule. The first step is to detect the large item set and the second to generate the association rule, using the large item set. Such rules must satisfy two conditions: 1. Support(X [ Y, D) P Minsup. 2. Confidence(X ? Y) P Minconf. The Minsup and Minconf are both set by the users. In general, the numbers of the transactions that comprise X is called the sup￾port of X, denoted by sx. Make Minsup the minimum value of sup￾port. If the support of X meets the condition, sx P Minsup, X is the large item set. As for the exploration of association rules, many researchers usually use the Apriori algorithm (Agrawal, Imielinski, & Swami, 1993). The judging standard is called lift. When all of the data employed can be classified, the simplest way to compare them is to use the lift judgment. In order to focus on a single target group from all trans￾actions, we can use rule X ? Y and find Y, applying the lift judg￾ment. Y would be less likely to be selected if we applied any other model than the lift. For example, if we use rule X ? Y to determine Support(Y), i.e., a customer’s probability of purchasing Y, the prod￾uct purchase multiple for Y, which is the lift target, is defined as: Lift = Confidence(X ? Y)/Support(Y) Support(A) denotes the probability of A that appears in all trans￾actions, and Support(A ? B) the probability of A that appears when B appears. Confidence(A ? B) denotes the condition probability of Support(A ? B) over Support(B). Therefore, the association analysis can discover high correlation rules from massive transactions. The minimum support value and the minimum confidence value are the threshold values decided by the user. To decide which rule is useful, the lift value must be bigger than one. Agrawal et al. (1993) first propose a calculation method to learn association rules. The Apriori’s algorithm mainly finds the large itemsets in a database and then combines the itemsets which sat￾isfy the minimum support. After the rules are found, Apriori will discard those that do not satisfy the minimum confidence. Park, Chen, and Yu (1995) present the DHP (Direct Hashing and Pruning) algorithm, using a hashing table to reduce both the space of item￾sets and computing complexity. This requires a little effort to build the hashing table at the beginning. Savasere, Omiecinski, and Navathe (1995) suggest the Partition algorithm, which divides the database into two parts and thus requires scanning the database only twice: the first scan builds the itemsets, and the second deter￾mines the support of large itemsets. Toivonen (1996) advises a sam￾pling algorithm, deeming that sampling will become necessary as the database increases. This method also requires scanning the database only twice. Brin, Motwani, Ullman, and Tsur (1997) propose the DIC (Dynamic Itemset Count) algorithm, which dynam￾ically evaluate the support value of itemsets. Similar to the Partition algorithm, the DIC divides database into many little parts to reduce search time. Han, Pei, & Yin (2000) offers the FP-Tree algorithm, using the tree structure to store itemsets; however, this device has been little used due to the complexity of the data structure. 2.2. Recommender system In recent years, particularly in the information technology field, the recommender system has attracted a growing amount of atten￾tion because of its success in many applications (Wang, Chuang, Hsu, & Keh, 2004). A recommender system is regarded as a heuris￾tic system that recommends useful information and can be applied in different domains. A simple example is Amazon.com online book store that, when no results are found for a query, suggests alternate books or queries that may achieve better results. Many successful recommender systems have been used in the fields of e-com￾mence, movies, music, books, and Web pages. Ma, Liu, Wong, Yu, and Lee (2000) find the right students using data mining. Chang (2000) discovers learning patterns from Web logs by concept transformation analysis. Hsu (2008) first clusters users with data mining, and then proceeds with a hybrid technique to develop an online personalized English learning recommender system capable of providing students with reading lessons that suit their different interests and, therefore, increase the motivation to learn. The idea of recommender systems comes from personalized information delivery. In ‘‘Personalized Information Delivery: An Analysis of Information Filtering Methods,’’ Foltz and Dumais (1992) present results of an experiment aimed at determining the effectiveness of four information-filtering methods in the do￾main of technical reports. Goldberg, Nichols, Oki, and Terry (1992) use collaborative filtering to weave an information tapestry. They describe an experimental system that manages an incoming stream of electronic documents, including Netnews, e-mails, news￾wire stories, and Netnews articles. Resnick and Varian (1997) sug￾gest the idea of using content-based and collaborative filtering methods in developing recommender systems. Sarwar, Kerypis, Konstan, and Riedl (2001) state that recom￾mender systems apply knowledge discovery techniques to the problem of making product recommendations during a live cus￾tomer interaction, and that one successful recommender system technology is collaborative filtering, which works by matching customer preferences to those of other customers in making recommendations. In the collaborative filtering approach, one identifies users whose tastes are similar to those of other users and recommends items such other users have chosen. Given a set of items, users can express their ratings of items they have tried before. The recommender can then compare the user’s ratings to those of other users to find the most similar users based on some 9282 M.-H. Hsu / Expert Systems with Applications 38 (2011) 9281–9286

M.-H. Hsu/ Expert Systems with Applications 38(2011)9281-9286 criteria of similarity and recommend items that similar users have 3. Proposing a charting recommender system liked in the past. The scores for unseen items are predicted based on a combination of the scores known from the nearest neighbors. le architecture of tI The other technique of recommender systems is the content-based shown in Fig. 1. First, we input our collected approach to recommending particular objects to users. Wang et al. system. Second the system segments the text = tences and (2004)report that with the content-based approach, one tries to then each sentence into words. Third with each sentence treate recommend the items similar to those that given users have liked as a transaction in the application of the association rules, the sys- in the past. It is based on a col between their content and a tem performs weighted sorting of the words. Finally, the system user profile makes word or term recommendations as the nurse does the Today, recommender systems are widely used in many different domains. Schafer, Konstan, and Riedl (1999) stress that recom- mender systems are achieving widespread success in e-commerce, 3. 1. The corpus of nursing documentation especially with the advent of the Internet. The Group Lens recom- mender system helps users wade through articles in Usenet news The recommender corpus is based on the text data from dictio- ( Konstan et al, 1997: Resnick, lacovo, Suchak, Bergstrom, Riedl, naries of nursing and books on nursing documentation Of these 1994). Ringo, a Net-based music recommendation service, allow sources, the most renowned, for instance, are Mosby s Clinical other music fans(Shardanand Maes, 1995). Gauch, Gauch, Bouix, Allied Health Dictionary(Anderson et al., 2008), Charting: An and Zhu(1999)recommend a real-time video scene detection and Incredibly Easy! Pocket Guide(Springhouse, 2006), and Mosby,'s classification. Cheng and Yang (1999)put forth the idea of a new Surefire Documentation(Mosby, 2006) content-based access method for video databases. Mooney and Roy(2000) describe a content-based book recommending system that utilizes information extraction and a machine-learning algo- 3. 2. Text and sentence segmentation rithm for text categorization. Kim and Choi(2002)introduce a con- The system first segments text data into sentences by checking Fleischman and Hovy(2003) present a natural language processing the dot signs. Then it proceeds with a lexical analysis, with each pproach for recommendation without user preferences. Carenin Smith, and Poole(2003)find a set of techniques to intelligently se- are predetermined to be stop words(including function words, lect what information to elicit from the user. In fact, numerous numbers, punctuation marks, pronouns, etc. )being automatically other recommender systems are successfully applied by online removed during the processing. This analysis thus identifies the food stores(Svensson, Laaksolahti, Hook, Waern, 2000)and mu- sic suggestion services( Chen Chen, 2001). Wang, Zhang, and Zhu (2004)suggest adaptive music emotion recognition by the var 3.3. Applying association rules and scoring tions of tonality, beat, tempo and interval. Tiemann and Pauws (2007)suggest ensemble learning for hybrid music recommenda In this phase, the system mines and runs the association rules. tion. Adomavicius and Tuzhilin(2005)offer an effective solution utilizing the concept of Agrawal et al. (1993)and treating each sen- by recommendation to searching information over the Internet. tence as a transaction. The initial recommender score of each word Salter and Antonopoulos(2006)propose the CinemaScreen system is calculated as zero. Then, the association rules is applied to deter for movie recommendations. Das, Datar, Garg, and Rajaram(2007) mine rules A-B, where B=B1, B n≥1, and the initial use scalable online collaborative filtering for Google news person- recommender score is adjust ed using the formula below: alization. Paliouras, Mouzakidis, Moustakas, and Skouras(2008) present a personalized news aggregator on the Web. Table 1 Text segmented into sentences. Text data input female head injury patient has transferred from another hospital this morning, with headache, vomiting, and diarrhea Corpora of nursing On admission, she was having stabbing left abdominal pain and high fever documentation Tert and Sentence In early aftemoon, she began shivering and was getting weaker Finding Association Rules Table 2 A sample of transactions database Scoring Transaction Items Recommendations dmission, have, stabbing, left, abdominal, pain, high, fever Early, afternoon, begin, shiver, get, weak Fig. 1. A simple architecture of the charting recommender system vomiting. and diarrhea. On bdominal pain and high fever. In early afternoon was getting weaker. Fig. 2. A part of a nursing record

criteria of similarity and recommend items that similar users have liked in the past. The scores for unseen items are predicted based on a combination of the scores known from the nearest neighbors. The other technique of recommender systems is the content-based approach to recommending particular objects to users. Wang et al. (2004) report that with the content-based approach, one tries to recommend the items similar to those that given users have liked in the past. It is based on a comparison between their content and a user profile. Today, recommender systems are widely used in many different domains. Schafer, Konstan, and Riedl (1999) stress that recom￾mender systems are achieving widespread success in e-commerce, especially with the advent of the Internet. The Group Lens recom￾mender system helps users wade through articles in Usenet news (Konstan et al., 1997; Resnick, Iacovou, Suchak, Bergstrom, & Riedl, 1994). Ringo, a Net-based music recommendation service, allows users to get music recommendations online and connect with other music fans (Shardanand & Maes, 1995). Gauch, Gauch, Bouix, and Zhu (1999) recommend a real-time video scene detection and classification. Cheng and Yang (1999) put forth the idea of a new content-based access method for video databases. Mooney and Roy (2000) describe a content-based book recommending system that utilizes information extraction and a machine-learning algo￾rithm for text categorization. Kim and Choi (2002) introduce a con￾tent-based video code translation in the compressed domain. Fleischman and Hovy (2003) present a natural language processing approach for recommendation without user preferences. Carenini, Smith, and Poole (2003) find a set of techniques to intelligently se￾lect what information to elicit from the user. In fact, numerous other recommender systems are successfully applied by online food stores (Svensson, Laaksolahti, Höök, & Waern, 2000) and mu￾sic suggestion services (Chen & Chen, 2001). Wang, Zhang, and Zhu (2004) suggest adaptive music emotion recognition by the varia￾tions of tonality, beat, tempo and interval. Tiemann and Pauws (2007) suggest ensemble learning for hybrid music recommenda￾tion. Adomavicius and Tuzhilin (2005) offer an effective solution by recommendation to searching information over the Internet. Salter and Antonopoulos (2006) propose the CinemaScreen system for movie recommendations. Das, Datar, Garg, and Rajaram (2007) use scalable online collaborative filtering for Google news person￾alization. Paliouras, Mouzakidis, Moustakas, and Skourlas (2008) present a personalized news aggregator on the Web. 3. Proposing a charting recommender system A simple architecture of the proposed recommender system is shown in Fig. 1. First, we input our collected text data into the system. Second, the system segments the text into sentences and then each sentence into words. Third, with each sentence treated as a transaction in the application of the association rules, the sys￾tem performs weighted sorting of the words. Finally, the system makes word or term recommendations as the nurse does the charting. 3.1. The corpus of nursing documentation The recommender corpus is based on the text data from dictio￾naries of nursing and books on nursing documentation. Of these sources, the most renowned, for instance, are Mosby’s Clinical Nursing (Thompson et al., 2002), Mosby’s Medical, Nursing, and Allied Health Dictionary (Anderson et al., 2008), Charting: An Incredibly Easy! Pocket Guide (Springhouse, 2006), and Mosby’s Surefire Documentation (Mosby, 2006). 3.2. Text and sentence segmentation The system first segments text data into sentences by checking the dot signs. Then it proceeds with a lexical analysis, with each word defined as that between two blanks and with those that are predetermined to be stop words (including function words, numbers, punctuation marks, pronouns, etc.) being automatically removed during the processing. This analysis thus identifies the content words and stores them in the database. 3.3. Applying association rules and scoring In this phase, the system mines and runs the association rules, utilizing the concept of Agrawal et al. (1993) and treating each sen￾tence as a transaction. The initial recommender score of each word is calculated as zero. Then, the association rules is applied to deter￾mine rules A ? B, where B = B1, B2, ... , Bn, n P 1, and the initial recommender score is adjusted using the formula below: Text and Sentence Segmentation Finding Association Rules Scoring Recommendations Text data input Corpora of nursing documentation Fig. 1. A simple architecture of the charting recommender system. A female head injury patient has transferred from another hospital this morning, with headache, vomiting, and diarrhea. On admission, she was having stabbing left abdominal pain and high fever. In early afternoon, she began shivering and was getting weaker. Fig. 2. A part of a nursing record. Table 2 A sample of transactions database. Transaction no. Items 1 Female, head, injury, patient, transfer, hospital, morning, headache, vomiting, diarrhea 2 Admission, have, stabbing, left, abdominal, pain, high, fever 3 Early, afternoon, begin, shiver, get, weak Table 1 Text segmented into sentences. A female head injury patient has transferred from another hospital this morning, with headache, vomiting, and diarrhea On admission, she was having stabbing left abdominal pain and high fever In early afternoon, she began shivering and was getting weaker M.-H. Hsu / Expert Systems with Applications 38 (2011) 9281–9286 9283

9284 M.-H. Hsu/ Expert Systems with Applications 38(2011)9281-9286 0.0001 +1000000 ==>[1447 100.0000 ==>I1448 0.0001 +100000143]+527]++529 >1519 0.000I +1000000[14 ==>I1465 143]+[1332H+|1466 100.0000 +1000000143]+[454 =>Ⅱ30 0.0001 00.0000 +1000000 143] =>[396] 0.000 +1000000 [143] 0.0001 100.0000 +1000000 00000 143] ==>I1352 0.0001 00.0000 +1000000 143] ==>B372] Table 3 4. Experimental result The recommender score values. Score The system is written with Visual C++ on an IBM PC. 4. 1. Text input has just typed) Fig. 2 is a part of a nursing record that has been input into the 100 syster 4.2. Segmentation The input text is first segmented into three sentences, as shown in Table 1. Then, after removing stop words and performing stemming and lemmatization, these sentences are segmented into content words. as shown in Table 2. Score(B)=Confidence(A-B)+|BI (3.1) 4.3. Association and scores where Score(B)is the weight value of B, with A being any partic ular word or term, B any other particular word or term that appears Fig 3 is a sample result after the system has run the association immediately after A, and b the frequency of this appearance. rules. According to Eq (3.1). the system will score the recommen- dation words 3.4. offering reco 4. 4. Recommendations Now, when the user has typed A, the system can automatie recommend Bn in descending order of their respective weight val- When a user types a word or term the system will, based on ues. Therefore, the system has a high probability of success in Fig 3, automatically sort recommender scores and suggest a word offering needed recommendations to the user. list in descending order of association, as shown in Table 3. Confidence(%) Ty Word A Word B 0.3396 100.0000 +294.4641 I143 ==>47 0.3323 100.0000 +300.9329 43] ==>[448 0.3323 96.0000 +2888956143]+{527++1529]=>[519 0.3758 95.2400 +253.4327 ==> 0.3758 86.9600 +231.3997143]+1332+1466]=>4531 0.3323 85.7100 2579296 43]+I454 ==>1310 0.3323 85.710 +257.9296 143 0.3591 +235.2548 ==> 826100 +230.0473 I143 ==>[449 0.4 4362 76.6700 +175.768 43] [1352 71.8800 +164.7868 ==>B372 Fig. 4. Sample results of association and scores-with the corpus categorized

ScoreðBÞ ¼ ConfidenceðA ! BÞþ j B j ð3:1Þ where Score(B) is the weight value of B, with A being any partic￾ular word or term, B any other particular word or term that appears immediately after A, and |B| the frequency of this appearance. 3.4. Offering recommendations Now, when the user has typed A, the system can automatically recommend Bn in descending order of their respective weight val￾ues. Therefore, the system has a high probability of success in offering needed recommendations to the user. 4. Experimental result The system is written with Visual C++ on an IBM PC. 4.1. Text input Fig. 2 is a part of a nursing record that has been input into the system. 4.2. Segmentation The input text is first segmented into three sentences, as shown in Table 1. Then, after removing stop words and performing stemming and lemmatization, these sentences are segmented into content words, as shown in Table 2. 4.3. Association and scores Fig. 3 is a sample result after the system has run the association rules. According to Eq. (3.1), the system will score the recommen￾dation words. 4.4. Recommendations When a user types a word or term, the system will, based on Fig. 3, automatically sort recommender scores and suggest a word list in descending order of association, as shown in Table 3. Support % Confidence % Type Lift Word/Term A Word /Term B 0.0001 100.0000 + 1000000 [143] ==> [1447] 0.0001 100.0000 + 1000000 [143] ==> [1448] 0.0001 100.0000 + 1000000 [143]+[527]++[529] ==> [519] 0.0001 100.0000 + 1000000 [143] ==> [1465] 0.0001 100.0000 + 1000000 [143]+[1332]+[1466] ==> [1453] 0.0001 100.0000 + 1000000 [143]+[1454] ==> [1310] 0.0001 100.0000 + 1000000 [143] ==> [396] 0.0001 100.0000 + 1000000 [143] ==> [521] 0.0001 100.0000 + 1000000 [143] ==> [449] 0.0001 100.0000 + 1000000 [143] ==> [1352] 0.0001 100.0000 + 1000000 [143] ==> [372] : : :: : : : : : :: : : : Fig. 3. Sample results of association. Table 3 The recommender score values. Recommended word/term no. Score Word/term A (any particular word or term user has just typed) [1447] 102 [1448] 101 [519] 100 [1465] 100 [1453] 100 [1310] 100 [396] 100 [521] 100 [449] 100 [1352] 100 [372] 100 : : : : Support % Confidence % Type Lift Word A Word B 0.3396 100.0000 + 294.4641 [143] ==> [1447] 0.3323 100.0000 + 300.9329 [143] ==> [1448] 0.3323 96.0000 + 288.8956 [143] +[527]+ +[529] ==> [519] 0.3758 95.2400 + 253.4327 [143] ==> [1465] 0.3758 86.9600 + 231.3997 [143] +[1332]+[1466] ==> [1453] 0.3323 85.7100 + 257.9296 [143]+[1454] ==> [1310] 0.3323 85.7100 + 257.9296 [143] ==> [396] 0.3591 84.4800 + 235.2548 [143] ==> [521] 0.3591 82.6100 + 230.0473 [143] ==> [449] 0.4362 76.6700 + 175.768 [143] ==> [1352] 0.4362 71.8800 + 164.7868 [143] ==> [372] : : :: : : : : : : : : :: Fig. 4. Sample results of association and scores—with the corpus categorized. 9284 M.-H. Hsu / Expert Systems with Applications 38 (2011) 9281–9286

M.-H. Hsu/ Expert Systems with Applications 38(2011)9281-9286 85 45. Evaluation Fleischman, M.& Hovy, E(2003). mendations without We have made a simple satisfaction investigation by asking 50 Foltz, P. w, Dumais, S. T.(1992). Personalized information delivery ond-language student nurses who used our system during their sis of information filtering methods. of the aCm practicum at hospitals. Most of them believe that the system is helpful, with their head ward nurses more satisfied with their Gauch, J M- Gauch, S, Bouix, S,& Zhu, X (1999). Real time video scene detection improved charting and other documentation. Genthial, D,& Courtin, J.(1992). From detection/ correction to compu 4.6 Discussion dberg, D Nichols, D, Oki, B M,& Terry, D (1992) Using collaborative filtering As shown in Fig 3, the Support values are close to zero and Han, J- Pei,J,& Yin, Y.(200 Proceedings of the 2000 ACM SIGMOD international conference Confidence values are close to 100. Such results are not surprising, management of data, ACM (pp 1-12) for the systems corpus of nursing documentation is not catego- Hsu, M. H(2008) A personalized English learning recommender system for ESL rized. If at first the corpus were categorized into gastrointestinal, Kim, T.Y. Choi, I. S. (2002). Content-based video transcoding in compressed cardiovascular, respiratory, neurological, orthopedic, etc, the re- sults as shown in Fig 4 would nevertheless be very similar to those onstan, IA Miller, B N Maltz, D, Herlocker, J. L Gordon, L R, Riedl, (1997 in Fig 3, with the same recommendations produced ng collaborative filtering to Usenet news Communications of ACM. 40 Lampe, S.(1984). Focus charting. Minneapolis, Minnesosta: Creative Nursing 5. Conclusion and future research Lampe, S.(1985). Focus charting: Streamlining documentation. Nursing This proposed recommender system aims to assist second charting. In A McLane(Ed h Classificaumenting nursing diagnosis using focus Nursing Diagnosis: proceedings of ing. Using data mining association rules, the process automatically Ma. Y. Liu gowee ce pp 37 s st Louis. Mo: Mosby Year Book. language nurses, especially student nurses. in patient care chart using data mining In Proceedings of the international geting the right students conference on knowledge provides optimal words or terms for the nurse to choose from as gACM(pp.457-464 she or he proceeds with the documentation. The system, having Montemuro, M.(1988) coRE documentation: A complete system for charting been tested on Taiwanese nursing students who needed to chart Mooney. R]. Roy, L(2000) Content-based book recommending using learning in their practicum at hospitals, has been found to be very effective for text categorization. In Proceedings of the ACM conference on digital libraries. for the intended purpose. However, a more complete corpus might still be desired if the system is to offer recommendations with Mosby(2006). Mosby 's surefire documentation: how, what, and when nurses need to more assuredness Murphy. Beglinger, E, Johnson, B (1988). Charting by exception: Meeting the challenge of cost containment. Nursing Management, 19(2)56-72. Navuluri, R B.(2000). The six honest servants of good documentation. Nursing. References New, E.(1999). Computer-aided writing in French as a foreign language: A Adomavicius, G,& Tuzhilin, A.(2005). Towards the next generation of of the state-of-the-art and possible Language journal 8300 81-97. 0oK as the process or revision. nsions. IEEE Transactions on Knowledge and Data Engineering. 17(6). Pa Moustakas. v. s.C.(2008)PNs:a lized news aggregator on the Web, intelligent interactive systems in Agrawal, R, Imielinski, T,& Swami, A (1993). Mining association rules between rou &L. Jain (Eds Studies in computational intelligence(VoL. 104, pp 175-197). Springer-Verla sets of items in large databases. In Proceedings of the SIGMOD 93. Washington, DC Park, J. C. Palmer, M,& Washburn, G(1997). An English grammar checker as a Anderson, K N, Anderson, LE,& Glanze, W. D(2008). Mosby,s medical, nursing, and writing aid for students of english age. In Proceedings of the Brin, S, Motwani, R. Ulm. New York: I Tsur, S(1997). Dynam Park, J. S, Chen, M.S,& Yu, P.S.(1995)- rules for market basket data. In Proceedings of the 1997 ACM SIGMOD mining association rules. In Proceedings of the SiGMoD 1995 intemational Buckley-Womack C.& Gidney, B(1987).A n in documenting: The Res 4) GroupLens: An ience Nursing. 19(5) 256-26 pen architecture for collaborative filtering of Netnews. In Proceedings of the urke, L, Murphy, J (1988) Charting by exception. New York: John Wiley nference on computer supported cooperative work, ACM SIGCHI bulletin(Vol. arian, H. R (1997). Recommender nference on intelligent user interfaces, ACM(pp 12-18 Chang. C. K.(2000)- Discovering learning patterns from Web logs by concep Salter, ], Antonopoulus, N (2006). CinemaScreen recommender agent: Cor transformation analysis. In Proceedings of the annual conference on intellige (pp.186-187 Sarwar. B. Ker stan, J. A,& Riedl, J. (2001) Item-based collab nternational conference on information and knowledge management, ACM(pp. Savasere. A. Omiecinski, E R,& Navathe,SB(1995).An efficient 231-238) Cheng, P.& Yang W. P (1999). A new content-based access method for video lining association rules in large databases. international conference on very large data bases(VLDB95) Morgan Kaufmann Schafer, J B, Konstan, J.A.& Riedl, ] (1999). Recommender systems in e-commerce. of the association for computational linguistics(NAACL 00) In Proceedings of the first ACM conference on electronic commerce, ACM (pp. 158- involving prepositions. In Proceedings of the fourth ACL Das, A S, Datar, M, Garg, A,& Rajaram, S(2007) Scott, V M, New, E (1994) Computer aided analysis of foreign language writing online collaborates the 16th international urnaL11(3).5- laes, P.(1995). Social in Ehnfors, M, Thorell-Ekstrand, L,& Ehrenberg, A(1991). Towards ace on human factors in information in patient records. Vard I Norden, 21, 12-3 computing systems 1995, ACMAddison-Wesley(pp 210-217). patient records: experience of the use of the VIPS-model. Journal of Advanced Springhouse(2006) Charting: An incredibly easy! Pocket guide. Spring nout. r a Ehrenberg. A Ehnfors, M.,& Thorell-Ekstrand, L (1996). Nursing documentation in ttor, R E,& Stocks, B (1985) The PIE system: Complete plannin uality Review Bulletin, 11(6), 186-189 Nursing24(4,853-867

4.5. Evaluation We have made a simple satisfaction investigation by asking 50 second-language student nurses who used our system during their practicum at hospitals. Most of them believe that the system is helpful, with their head ward nurses more satisfied with their improved charting and other documentation. 4.6. Discussion As shown in Fig. 3, the Support values are close to zero and Confidence values are close to 100. Such results are not surprising, for the system’s corpus of nursing documentation is not catego￾rized. If at first the corpus were categorized into gastrointestinal, cardiovascular, respiratory, neurological, orthopedic, etc., the re￾sults as shown in Fig. 4 would nevertheless be very similar to those in Fig. 3, with the same recommendations produced. 5. Conclusion and future research This proposed recommender system aims to assist second￾language nurses, especially student nurses, in patient care chart￾ing. Using data mining association rules, the process automatically provides optimal words or terms for the nurse to choose from as she or he proceeds with the documentation. The system, having been tested on Taiwanese nursing students who needed to chart in their practicum at hospitals, has been found to be very effective for the intended purpose. However, a more complete corpus might still be desired if the system is to offer recommendations with more assuredness. References Adomavicius, G., & Tuzhilin, A. (2005). Towards the next generation of recommender systems: a survey of the state-of-the-art and possible extensions. IEEE Transactions on Knowledge and Data Engineering, 17(6), 734–749. Agrawal, R., Imielinski, T., & Swami, A. (1993). Mining association rules between sets of items in large databases. In Proceedings of the SIGMOD’93, Washington, DC (pp. 207–216). Anderson, K. N., Anderson, L. E., & Glanze, W. D. (2008). Mosby’s medical, nursing, and allied health dictionary. New York: Mosby. Brin, S., Motwani, R., Ullman, J. D., & Tsur, S. (1997). Dynamic itemset counting and implication rules for market basket data. In Proceedings of the 1997 ACM SIGMOD international conference on management of data, ACM (pp. 255–264). Buckley-Womack, C., & Gidney, B. (1987). A new dimension in documenting: The PIE method. Journal of Neuroscience Nursing, 19(5), 256–260. Burke, L., & Murphy, J. (1988). Charting by exception. New York: John Wiley & Sons. Carenini, G., Smith, J., & Poole, D. (2003). Towards more conversational and collaborative recommender systems. In Proceedings of the 2003 international conference on intelligent user interfaces, ACM (pp. 12–18). Chang, C. K. (2000). Discovering learning patterns from Web logs by concept transformation analysis. In Proceedings of the annual conference on intelligent user interfaces, ACM (pp. 186–187). Chen, H. C., & Chen, A. L. P. (2001, November). A music recommendation system based on music data grouping and user interests. In Proceedings of ACM international conference on information and knowledge management, ACM (pp. 231–238). Cheng, P. J., & Yang, W. P. (1999). A new content-based access method for video databases. Information Sciences, 118, 37–73. Chodorow, M., & Leacock, C. (2000). An unsupervised method for detecting grammatical errors. In Proceedings of the 1st meeting of North American chapter of the association for computational linguistics (NAACL 00), Seattle (pp. 140–147). Chodorow, M., Tetreault, J. R. & Han, N. (2007). Detection of grammatical errors involving prepositions. In Proceedings of the fourth ACL-SIGSEM workshop on prepositions (pp. 25–30). Das, A. S., Datar, M., Garg, A., & Rajaram, S. (2007). Google news personalization: scalable online collaborative filtering. In Proceedings of the 16th international World Wide Web conference (WWW 2007) (pp. 271–280). Ehnfors, M., Thorell-Ekstrand, I., & Ehrenberg, A. (1991). Towards basic nursing information in patient records. Vard I Norden, 21, 12–31. Ehrenberg, A., Ehnfors, M., & Thorell-Ekstrand, I. (1996). Nursing documentation in patient records: experience of the use of the VIPS-model. Journal of Advanced Nursing, 24(4), 853–867. Fleischman, M., & Hovy, E. (2003). Recommendations without user preferences: A natural language processing approach. In Proceedings of the 2003 international conference on intelligent user interfaces, ACM (pp. 242–244). Foltz, P. W., & Dumais, S. T. (1992). Personalized information delivery: An analysis of information filtering methods. Communications of the ACM, 35, 51–60. Gauch, J. M., Gauch, S., Bouix, S., & Zhu, X. (1999). Real time video scene detection and classification. Information Processing and Management, 35, 381–400. Genthial, D., & Courtin, J. (1992). From detection/correction to computer aided writing. In Proceedings of the 14th conference on computational linguistics (Vol. 3, pp. 1013–1018). Goldberg, D., Nichols, D., Oki, B. M., & Terry, D. (1992). Using collaborative filtering to weave an information tapestry. Communications of the ACM, 35, 61–70. Han, J., Pei, J., & Yin, Y. (2000). Mining frequent patterns without candidate generation. In Proceedings of the 2000 ACM SIGMOD international conference on management of data, ACM (pp. 1–12). Hsu, M. H. (2008). A personalized English learning recommender system for ESL students. Expert Systems with Applications, 34, 683–688. Kim, T. Y., & Choi, J. S. (2002). Content-based video transcoding in compressed domain. Signal Processing: Image Communication, 17, 497–507. Konstan, J. A., Miller, B. N., Maltz, D., Herlocker, J. L., Gordon, L. R., & Riedl, J. (1997). GroupLens: Applying collaborative filtering to Usenet news. Communications of the ACM, 40, 77–87. Lampe, S. (1984). Focus charting. Minneapolis, Minnesosta: Creative Nursing Management. Lampe, S. (1985). Focus charting: Streamlining documentation. Nursing Management, 16(7), 43–46. Lampe, S., & Hitchcock, A. (1987). Documenting nursing diagnosis using focus charting. In A. McLane (Ed.), Classification of Nursing Diagnosis: proceedings of the 7th conference (pp. 377). St Louis, MO: Mosby Year Book. Ma, Y., Liu, B., Wong, C. K., Yu, P. S., & Lee, S. M. (2000). Targeting the right students using data mining. In Proceedings of the international conference on knowledge discovery and data mining, ACM (pp. 457–464). Montemuro, M. (1988). CORE documentation: A complete system for charting nursing care. Nursing Management, 19(8), 28–32. Mooney, R. J., & Roy, L. (2000). Content-based book recommending using learning for text categorization. In Proceedings of the ACM conference on digital libraries, ACM (pp. 195–204). Mosby (2006). Mosby’s surefire documentation: how, what, and when nurses need to document. New York: Mosby. Murphy, J., Beglinger, J. E., & Johnson, B. (1988). Charting by exception: Meeting the challenge of cost containment. Nursing Management, 19(2), 56–72. Navuluri, R. B. (2000). The six honest servants of good documentation. Nursing, 30(6), 22. New, E. (1999). Computer-aided writing in French as a foreign language: A qualitative and quantitative look as the process of revision. The Modern Language Journal, 83(1), 81–97. Paliouras, G., Mouzakidis, A., Moustakas, V., & Skourlas, C. (2008). PNS: a personalized news aggregator on the Web, intelligent interactive systems in knowledge-based environments. In M. Virvou & L. Jain (Eds.). Studies in computational intelligence (Vol. 104, pp. 175–197). Springer-Verlag. Park, J. C., Palmer, M., & Washburn, G. (1997). An English grammar checker as a writing aid for students of English as a second language. In Proceedings of the conference of applied natural language processing (ANLP) (p. 24). Park, J. S., Chen, M. S., & Yu, P. S. (1995). An effective hash based algorithm for mining association rules. In Proceedings of the SIGMOD 1995 international conference on management of data (SIGMOD’ 95), ACM (pp. 175–186). Resnick, P., Iacovou, N., Suchak, M., Bergstrom, P., & Riedl, J. (1994). GroupLens: An open architecture for collaborative filtering of Netnews. In Proceedings of the conference on computer supported cooperative work, ACM SIGCHI bulletin (Vol. 27(3), pp. 175–186). Resnick, P., & Varian, H. R. (1997). Recommender systems. Communications of the ACM, 40, 56–60. Salter, J., & Antonopoulus, N. (2006). CinemaScreen recommender agent: Combining collaborative filtering and content-based filtering. IEEE Intelligent Systems, 21(1), 35–41. Sarwar, B., Kerypis, G., Konstan, J. A., & Riedl, J. (2001). Item-based collaborative filtering recommendation algorithms. In Proceedings of the 10th international World Wide Web conference, ACM (pp. 285–295). Savasere, A., Omiecinski, E. R., & Navathe, S. B. (1995). An efficient algorithm for mining association rules in large databases. In Proceedings of the 21th international conference on very large data bases (VLDB’ 95), Morgan Kaufmann (pp. 432–444). Schafer, J. B., Konstan, J. A., & Riedl, J. (1999). Recommender systems in e-commerce. In Proceedings of the first ACM conference on electronic commerce, ACM (pp. 158– 166). Scoates, G. H., Fishman, M., & McAdam, B. (1996). Health care focus documentation. Nursing Management, 27(8), 30–32. Scott, V. M., & New, E. (1994). Computer aided analysis of foreign language writing process. CALIC Journal, 11(3), 5–18. Shardanand, U., & Maes, P. (1995). Social information filtering: Algorithms for automating word of mouth. In Proceedings of the conference on human factors in computing systems 1995, ACM/Addison-Wesley (pp. 210–217). Siegrist, L. M., Dettor, R. E., & Stocks, B. (1985). The PIE system: Complete planning and documentation of care. Quality Review Bulletin, 11(6), 186–189. Springhouse (2006). Charting: An incredibly easy! Pocket guide. Springhouse: PA. M.-H. Hsu / Expert Systems with Applications 38 (2011) 9281–9286 9285

M.-H. Hsu/ Expert Systems with Applications 38(2011)9281-9286 Pratt, E.(1996). A comparative study of two ESL writin environments: A computer-ass ani meplenentat ion ot an Er: How os encouage the Ase. interational romano interfaces, ACM(pp (CSP04) IEEE(pp. 1352 pping: A nursing model for care planning. Wang. Y. F. Chuang. Y. L Hsu, M. H,& Keh, H. C(2004).A Thompson,. M, McFarland, G K, Hirsch, J.., & Tucker, S M. (2002) Mosby's clinical W.pplications, 26, 427-434. A,& McWeeny, M. C (1991) Managing the cost of documentation: The Tiemann, M.,& Pauws, S(2007). Towards ensemble learning for hybrid music FACT charting system Nursing Economics, 9(3) 181-18 commendation. In Proceedings of the 2007 ACM conference on recommender Toivonen, H (1996).Sar the 22nd conference of very Large Databases(VLDB'96), Morgan Kaufmann(pp

Sullivan, N., & Pratt, E. (1996). A comparative study of two ESL writing environments: A computer-assisted classroom and a traditional oral classroom. System, 24(4), 491–501. Svensson, M., Laaksolahti, J., Höök, K., & Waern, A. (2000). A recipe based online food store. In Proceedings of the 2000 international conference on intelligent user interfaces, ACM (pp. 260–263). Taylor, J., & Wros, P. (2007). Concept mapping: A nursing model for care planning. Journal of Nursing Education, 46(5), 211–216. Thompson, J. M., McFarland, G. K., Hirsch, J. E., & Tucker, S. M. (2002). Mosby’s clinical nursing. New York: Mosby. Tiemann, M., & Pauws, S. (2007). Towards ensemble learning for hybrid music recommendation. In Proceedings of the 2007 ACM conference on recommender systems, ACM (pp. 177–178). Toivonen, H. (1996). Sampling large databases for association rules. In Proceedings of the 22nd conference of Very Large Databases (VLDB’ 96), Morgan Kaufmann (pp. 134–145). Van der Meijden, M. J., Tange, H., Troost, J., & Hasman, A. (2001). Development and implementation of an EPR: How to encourage the user. International Journal of Medical Informatics, 64, 173–185. Wang, M. Y., Zhang, N. Y., & Zhu, H. C. (2004). User-adaptive music emotion recognition. In Proceedings of the 7th international conference on signal processing (ICSP’04), IEEE (pp. 1352–1355). Wang, Y. F., Chuang, Y. L., Hsu, M. H., & Keh, H. C. (2004). A personalized recommender system for the cosmetic business. Expert Systems with Applications, 26, 427–434. Warne, M. A., & McWeeny, M. C. (1991). Managing the cost of documentation: The FACT charting system. Nursing Economic$, 9(3), 181–187. 9286 M.-H. Hsu / Expert Systems with Applications 38 (2011) 9281–9286