Automatic Generation of Related Work Sections in Scientific Papers: An Optimization Approach

Size: px
Start display at page:

Download "Automatic Generation of Related Work Sections in Scientific Papers: An Optimization Approach"

Transcription

1 Autoatic Generation of Related Work Sections in Scientific Papers: An Optiization Approach Yue Hu and Xiaojun Wan Institute of Coputer Science and Technology The MOE Key Laboratory of Coputational Linguistics Peking University, Beijing, China Abstract In this paper, we investigate a challenging task of autoatic related work generation. Given ultiple reference papers as input, the task ais to generate a related work section for a target paper. The generated related work section can be used as a draft for the author to coplete his or her final related work section. We propose our Autoatic Related Work Generation syste called ARWG to address this task. It first exploits a PLSA odel to split the sentence set of the given papers into different topic-biased parts, and then applies regression odels to learn the iportance of the sentences. At last it eploys an optiization fraework to generate the related work section. Our evaluation results on a test set of 15 target papers along with their reference papers show that our proposed ARWG syste can generate related work sections with better quality. A user study is also perfored to show ARWG can achieve an iproveent over generic ulti-docuent suarization baselines. 1 Introduction The related work section is an iportant part of a paper. An author often needs to help readers to understand the context of his or her research proble and copare his or her current work with previous works. A related work section is often used for this purpose to show the differences and advantages of his or her work, copared with related research works. In this study, we attept to autoatically generate a related work section for a target acadeic paper with its reference papers. This kind of related work sections can be used as a basis to reduce the author s tie and effort when he or she wants to coplete his or her final related work section. Autoatic related work section generation is a very challenging task. It can be considered a topic-biased, ultiple-docuent suarization proble. The input is a target acadeic paper, which has no related work section, along with its reference papers. The goal is to create a related work section that describes the related works and addresses the relationship between the target paper and the reference papers. Here we assue that the set of reference papers has been given as part of the input. Existing works in the NLP and recoendation systes counities have already focused on the task of finding reference papers. For exaple, citation prediction (Nallapati et al., 28) ais at finding individual paper citation patterns. Generally speaking, autoatic related work section generation is a strikingly different proble and it is uch ore difficult in coparison with general ulti-docuent suarization tasks. For exaple, ulti-docuent suarization of news articles ais at synthesizing contents of siilar news and reoving the redundant inforation contained by the different news articles. However, each scientific paper has uch specific content to state its own work and contribution. Even for the papers that investigate the sae research topic, their contributions and contents can be totally different. The related work section generation task needs to find the specific contributions of individual papers and arrange the into one or several paragraphs. In this study, we focus on the proble of autoatic related work section generation and propose a novel syste called ARWG to address the 1624 Proceedings of the 214 Conference on Epirical Methods in Natural Language Processing (EMNLP), pages , October 25-29, 214, Doha, Qatar. c 214 Association for Coputational Linguistics

2 proble. For the target paper, we assue that the abstract and introduction sections have already been written by the author and they can be used to help generate the related work section. For the reference papers, we only consider and extract the abstract, introduction, related work and conclusion sections, because other sections like the ethod and evaluation sections always describe the extree details of the specific work and they are not suitable for this task. Then we generate the related work section using both sentence sets which are extracted fro the target paper and reference papers, respectively. Firstly, we use a PLSA odel to group both sentence sets of the target paper and its reference papers into different topic-biased clusters. Secondly, the iportance of each sentence in the target paper and the reference papers is learned by using two different Support Vector Regression (SVR) odels. At last, a global optiization fraework is proposed to generate the related work section by selecting sentences fro both the target paper and the reference papers. Meanwhile, the fraework selects sentences fro different topic-biased clusters globally. Experiental results on a test set of 15 target papers show our ethod can generate related work sections with better quality than those of several baseline ethods. With the ROUGE toolkit, the results indicate the related work sections generated by our syste can get higher ROUGE scores. Moreover, our related work sections can get higher rating scores based on a user study. Therefore, our related work sections can be uch ore suitable for the authors to prepare their final related work sections. 2 Related Work There are few studies to directly address autoatic related work generation. Hoang and Kan (21) proposed a related work suarization syste given the set of keywords arranged in a hierarchical fashion that describes the paper s topic. They used two different rule-based strategies to extract sentences for general topics as well as detailed ones. A few studies focus on ulti-docuent scientific article suarization. Agarwal et al., (211) introduced an unsupervised approach to the proble of ulti-docuent suarization. The input is a list of papers cited together within the sae source article. The key point of this approach is a topic based clustering of fragents extracted fro each co-cited article. They rank all the clusters using a query generated fro the context surrounding the co-cited list of papers. Yeloglu et al., (211) copared four different approaches for ulti-docuent scientific articles suarization: MEAD, MEAD with corpus specific vocabulary, LexRank and W3SS. Other studies investigate ainly on the singledocuent scientific article suarization. Early works including (Luhn 1958; Baxendale 1958; Eduundson 1969) tried to use various features specific to scientific text (e.g., sentence position, or rhetorical clues features). They have proved that these features are effective for the scientific article suarization. Citation inforation has been already shown effective in suarize the scientific articles. Works including (Mei and Zhai 28; Qazvinian and Radev 28; Schwartz and Hearst 26; Mohaad et al., 29) eployed citation inforation for the single scientific article suarization. Earlier work (Nakov et al., 24) indicated that citation sentences ay contain iportant concepts that can give useful descriptions of a paper. Various ethods have been proposed for news docuent suarization, including rule-based ethods (Barzilay and Elhadad 1997; Marcu and Daniel 1997), graph-based ethods (Mani and Bloedorn 2; Erkan and Radev 24; Michalcea and Tarau 25), learning-based ethods (Conroy et al., 21; Shen et al., 27; Ouyang et al., 27; Galanis et al., 28), optiizationbased ethods (McDonald 27; Gillick et al., 29; Xie et al., 29; Berg-Kirkpatrick et al., 211; Lei Huang et al., 211; Woodsend et al., 212; Galanis 212), etc. The ost relevant work is (Hoang and Kan, 21) as entioned above. They also assued the set of reference papers was given as part of the input. They also adopt the hierarchical topic tree that describes the topic structure in the target paper as an essential input for their syste. However, it is non-trivial to build the hierarchical topic tree. Moreover, they do not consider the content of the target paper to construct the related work section, which is actually crucial in the related work section. To the best of our knowledge, no previous works have used supervised learning and optiization fraework to deal with the ultiple scientific article suarization tasks. 3 Proble Analysis and Corpus 3.1 Proble Analysis 1625

3 There has been a substantial aount of research on autoatic taxonoy induction. As we entioned earlier, two ain approaches are pattern-based and clustering-based. Pattern-based approaches are the ain trend for autoatic taxonoy induction. Pattern-based approaches started fro and still pay a great deal of attention to the ost coon is-a relations. Clustering-based approaches usually represent word contexts as vectors and cluster words based on siilarities of the vectors (Brown et al., 1992; Lin, 1998). Many clustering-based approaches face the challenge of appropriately labeling non-leaf clusters. In this paper, we take an increental clustering approach,... The advantage of the increental approach is that it eliinates the trouble of inventing cluster labels and concentrates on placing ters in the correct positions in a taxonoy hierarchy. The work by Snow et al. (26) is the ost siilar to ours Moreover, our approach eploys heterogeneous features fro a wide range; while their approach only used syntactic dependency. We firstly analyze the structure of related work sections briefly. By using exaples for illustration, we can gain insight on how to generate related work sections. A specific related work exaple is shown in Figure 1. This related work section introduces previous related works for a paper on Autoatic Taxonoy Induction. Fro Figure 1, we can have a glance at the structure of related work sections. Related work sections usually discuss several different topics, such as pattern-based and cluster-based approaches shown in the Figure 1. Besides the knowledge of previous works, the author often copares his own work with the previous works. The differences and advantages are generally entioned. The exaple in Figure 1 also indicates this phenoenon. Therefore, we design our syste to generate related work sections according to the related work section structure entioned above. Our syste takes the target paper for which a related work section needs to be drafted besides its reference papers as input. The goal of our syste is to generate a related work section with the above structure. The generated related work section should have several topic-biased parts. The author's own work is also needed to be described and its difference with other works is needed to be ephasized on. 3.2 Corpus and Preprocessing We build a corpus that contains acadeic papers and their corresponding reference papers. The acadeic papers are selected fro the ACL Anthology 1. The ACL Anthology currently hosts 1 Two different topics Coparison with the author s work Figure 1: A saple related work section (Yang and Callan 29) over 24,5 papers fro ajor conferences such as ACL, EMNLP, COLING in the fields of coputational linguistics and natural language processing. We reove the papers that contain related work sections with very short length, and randoly select 15 target papers to construct our whole corpus. The papers are all in PDF forat. We extract their texts by using PDFlib 2 and detect their physical structures of paragraphs, subsections and sections by using ParsCit 3. For the target papers, the related work sections are directly extracted as the gold suaries. The references are also extracted. For the references that can be found in the ACL Anthology, we download the fro the ACL Anthology. The other reference papers are searched and downloaded by using Google Scholar. References to books and PhD theses are discarded, for their verbosity ay change the proble drastically (Mihalcea and Ceylan, 27). The input of our syste includes the abstract and introduction sections of the target paper, and the abstract, introduction, related work and conclusion sections of the reference papers. As entioned above, the ethod and evaluation sections in the reference papers are not used as input because these sections usually describe extree details of the ethods and evaluation results and they are not suitable for related work generation. Note that it is reasonable to ake use of the abstract and introduction sections of the target paper to help generate the related work section, because an author usually has already written the abstract and introduction sections before he or she wants to write the related work section for the target paper. Otherwise, we cannot get any inforation about the author s own work. All other sections in the target paper are not used. 4 Our Proposed Syste 4.1 Overview In this paper, we propose a syste called ARWG to autoatically generate a related work section for a given target paper. The architecture of our syste is shown in Figure 2. We take both the target paper and its reference papers as input and they are represented by several sections entioned in Section 3.2. After preprocessing, we extract the feature vectors for sentences in the target paper and the reference papers, respective

4 ly. The iportance scores for sentences in the target paper and the reference papers are assigned by using two SVR based sentence scoring odels. The two SVR odels are trained for sentences in the target paper and the reference papers, respectively. Meanwhile, a topic odel is applied to the whole set of sentences in both the target paper and reference papers. The sentences are grouped into several different topic-biased clusters. The sentences with iportance scores and topic cluster inforation are taken as the input for the global optiization fraework. The optiization fraework extracts sentences to describe both the author s own work and background knowledge. More details of each part will be discussed in the following sections. 4.2 Topic Model Learning Target paper Sentence Score Assessent(target) Preprocessing Sentence Score Assessent(reference) Optiization Fraework Postprocessing Reference papers Figure 2: Syste Architecture Topic Model Related Work Section As entioned in the previous section, the related work section usually addresses several different topics. The topics ay be different research thees or different aspects of a broad research thee. The related work section should describe the specific details for each topic, respectively. Therefore, we ai to discover the hidden topics of the input papers, and we use the Probabilistic latent seantic analysis (PLSA) (Hofann, 1999) to solve this proble. The PLSA approach odels each word in a docuent as a saple fro a ixture odel. The ixture coponents are ultinoial rando variables that can be viewed as representations of topics. Different words in a docuent ay be generated fro different topics. Each docuent is represented a list of ixing proportions for these ixture coponents and can be reduced to a probability distribution on a fixed set of topics. Considering that the sentences in one paper ay relate to different topics, we treat each sentence as a docuent d. We treat the noun phases in the sentences as the words w. In order to extract the noun phrases, chunking ipleented by the OpenNLP toolkit 4 is applied to the sentences. Noun phrases that contain words such as paper and data are discarded. Then the sentences with their corresponding noun phrases are taken as input into the PLSA odel. Here both the sentences in the target paper and the sentences in the reference papers are treated the sae in the odel. Finally, we can get the sentence set with topic inforation and use it in the subsequent steps. Each sentence has a topic weight t in each topic. 4.3 Sentence Iportant Assessent In our proposed syste, sentence iportance assessent ais to assign an iportance score to each sentence in the target paper and reference papers. The score of each sentence will be used in the subsequent optiization fraework. We propose to use the support vector regression odel to achieve this goal. In the above topic odel learning process, we do not distinguish the sentences in the target paper and reference papers. In contrast, we train two different support vector regression odels separately for the sentences in the target paper and the sentences in the reference papers. In the related work section, the sentences that describe the author s own work usually address the differences fro the related works, while the sentences that describe the related works often focus on the specific details. We think the two kinds of sentences should be treated differently. Scoring Method To construct training data based on the papers collected, we apply a siilarity scoring ethod to assign the iportance scores to the sentences in the papers. The ain hypothesis is that the sentences in the gold related work sections should suarize the target paper and reference papers as well. Thus the sentences in the papers which are ore siilar to the sentences in the gold related work sections should be considered ore iportant and suitable to be selected. Our scoring ethod should assign higher scores to the

5 We define the iportance score of a sentence in the papers as below: score(s) = ax s i S (si(s, s i )) (1) where s is a sentence in the papers, S is the set of the sentences in the corresponding gold related work section. The standard cosine easure is eployed as the siilarity function. Considering the difference between the sentences that describe the author s work and the sentences that describe the related works, we split the set of sentences in the gold related work section into two parts: one discusses the author s own work and the other introduces the related works. We observe that sentences related to the author s own work often feature specific words or phrases (such as we, our work, in this paper etc.) in the related work section. So we check the sentences about whether they contain clue words or phrases (i.e., in this paper, our work and 18 other phrases). If the clue phrase check fails, the sentence belongs to the related work part. If not, it belongs the own work part. Thus for the sentences in the target paper, S is the set of sentences in the own work part of the gold related work section, while for the sentences in the reference papers, S is the set of sentences in the related work part of the gold related work section. Then we can use the scoring ethod to copute the target scores of the sentences in the training set. It is noteworthy that two SVR odels can be trained on the two parts of the training data, respectively. Feature Each sentence is represented by a set of features. The coon features used for the sentences of the target paper and reference papers are shown in Table 1. The additional features applied to the sentences of the target paper are introduced in Table 2. Here, s is a sentence that needs to extract features. th is paper title, section headings and subsection headings set of the reference papers or target paper for the two SVR odels, respectively. Each feature with * represent a feature set that contains siilar features. All the features are scaled into [, 1]. Thus we can learn SVR odels based on the features and iportance scores of the sentences, and then use the odels to predict an iportance score for each sentence in the test set. The SVR odels are trained and applied for the target paper and reference papers, respectively. Table 1: Coon features eployed in the SVR odels Feature Si(s, th) WS(s,th) SP(s) Description The siilarity between s and each title in th; Stop words are reoved and steing is eployed. Nuber of words shared by s and th. The position of s in its section or subsection PTI(s) The parse tree inforation of s, including the nuber of noun phrase and verb phrases, the depth of the parse tree, etc. IsHead(s) IsEnd(s) SWP(s) Length(s) Length_rw(s) SI(s) CluePhrase(s) Indicates whether s is the first sentence of the section or subsection Indicates whether s is the last sentence of the section or subsection The percentage of the stop words The length of sentence s The length of s after reoving stop words The section index of s that indicates which section s is fro. Indicates whether a clue phrase appears in s. the clue phrases include our work, propose and other 2 words. Each clue phrase corresponds to one feature. Table 2: Additional features for sentences in the target paper Feature Description HasCitation(s) PhraseForCp(s) Indicates whether s contains a citation Indicates whether s contains words or phrases used for coparison such as in contrast, instead and other 26 words. Each word or phrase corresponds to one feature. 4.4 A Global Optiization Fraework In the above steps, we can get the predicted iportance score and topic inforation for each sentence in the target paper and reference papers. Here, we introduce a global optiization fraework to generate the related work section. According to the structure of the related work section entioned above, the related work section usually discusses several topics. In each topic, the related works and their details are introduced. Besides, the author often copares his own work with these previous works. Therefore, we propose to forulate the generation as an optiization proble. Basically, we will be searching for a set of sentences to optiize the objective function. 1628

6 Table 3: Notations used in this section Sybol Description sr i /st i lr i /lt i wr i /wt i xr ij /xt ij nr/nt t ij B y i c bi L ax L j B B i Sr /St λ 1, λ 2, λ 3 the sentence in the reference/target paper the length of sentence sr i / st i the iportance score of sr i /st i indicates whether sr i /st i is selected into the part of topic j in the generated related work section the nuber of sentences in the reference/target papers the topic count the topic weight of sr i /st i in topic j fro the PLSA odel the set of unique bigras indicates whether bigra b i is included in the result the count of the occurrences of bigra b i in the both target paper and reference papers the axiu word count of the related work section the axiu word count of the part of topic j which depends on the percentage of sentences belong to topic j the total set of bigras in the whole paper set the set of bigras that sentence sr i /st i contains the set of sentences that include bigra b in the reference/target papers paraeters for tuning To design the objective function, three aspects should be considered: 1) First, the related work section we generate should introduce the previous works well. In our assuption, sentences with higher iportance scores are better to be selected. In addition, very short sentences should be penalized. So we introduce the first part of our objective function below: nr i=1 (lr i wr i j=1 t ij xr ij ) (2) We add the sentence length as a ultiplication factor in order to penalize the very short sentences, or the objective function tends to select ore and shorter sentences. At the sae tie, the objective function does not tend to select the very long sentences. The total length of the sentences selected is fixed. So if the objective function tends to select the longer sentences, the fewer sentences can be selected. A tradeoff needs to be ade between the nuber and the average length of the sentences selected. The constraints introduced below ensure that the sentence can only be selected into one topic and the topic weight is used to easure the degree that the sentence is relevant to the specific topic. 2) Second, siilar to the first part, we should consider the own work part of the related work section. Thus the second part of our objective function is shown as follows: nt i=1 (lt i wt i j=1 t ij xt ij ) (3) 3) At last, redundancy reduction should be considered in the objective function. The last part of the objective function is shown below: B i=1 c bi y i (4) The intuition is that the ore unique bigras the related work section contains, the less redundancy the related work section has. We add c bi as the weight of the bigra in order to include ore iportant bigras. By cobing all the parts defined above, we have the following full objective function: ax λ nr 1 ( lr i i=1 wr xr,xt αl i j=1 t ij xr ij ) + ax nt lt λ 2 ( i i=1 wt (1 α)l i j=1 t ij xt ij ) + ax B λ 3 c b i y i i=1 (5) B Subject to: nr nt i=1 lr i xr ij + i=1 lt i xt ij < L j, for j = 1,, (6) nr i=1 j=1 lr i xr ij < αl ax (7) nt i=1 j=1 lt i xt ij < (1 α)l ax (8) j=1 xr ij 1, for i = 1,, nr (9) j=1 xt ij 1, for i = 1,, nt (1) bk B i y k B i j=1 xr ij, for i = 1,, nr (11) bk B i y k B i j=1 xt ij, for i = 1,, nt (12) sr i Sr k j=1 xr ij + st i St k j=1 xt ij y k, k = 1, B (13) xr ij, xt ij, y i {,1} (14) All the three parts in the objective function are noralized to [, 1] by using the axiu length L ax and the total nuber of bigras B. λ 1, λ 2 and λ 3 are paraeters for tuning the three parts and we set λ 1 +λ 2 +λ 3 = 1. We explain the constraints as follows: Constraint (6): It ensures that the total word count of the part of topic j does not exceed L j. Constraints (7), (8): The two constraints try to balance the lengths of the previous works part and the own work part, respectively. α is set to 2/3. Constraints (9), (1): These two constraints guarantee that the sentence can only be included into one topic. 1629

7 Constraints (11), (12): When these two constraints hold, all bigras that s i has are selected if s i is selected. Constraint (13): This constraint akes sure that at least one sentence in Sr or St is selected if bigra b is selected. Therefore, we transfor our optiization proble into a linear prograing proble. We solve this linear prograing proble by using the IBM CPLEX optiizer 5. It generally takes tens of seconds to solve the proble and it is very efficient. Finally, ARWG post-processes sentences to iprove readability, including replacing agentive fors with a citation to the specific article (e.g., our work (Hoang and Kan, 21) ) for the sentences extracted fro reference papers. The sentences belonging to different topics are placed separately. 5 Evaluation 5.1 Evaluation Setup To set up our experients, we divide our dataset which contains 15 target papers and their reference papers into two parts: 7 target papers for training, 15 papers for test and the other 2 papers for validation. The PLSA topic odel is applied to the whole dataset. We train two SVR regression odels based on the own work part and the previous work part of the training data and apply the odels to the test data. The global optiization fraework is used to generate the related work sections. We set the axiu word count of the generated related work section to be equal to that of the gold related work section. The paraeter values of λ 1, λ 2 and λ 3 are set to.3,.1 and.6, respectively. The paraeter values are tuned on the validation data. We copare our syste with five baseline systes: MEAD-WT, LexRank-WT, ARWG-WT, MEAD and LexRank. MEAD 6 (Radev et al., 24) is an open-source extractive ultidocuent suarizer. LexRank 7 (Eran and Radev, 24) is a ulti-docuent suarization syste which is based on a rando walk on the siilarity graph of sentences. We also ipleent the MEAD, LexRank baselines and our ethod 5 www-1.ib.co/software/integration/optiization/cplexoptiizer/ 6 7 In our experients, LexRank perfors uch better than the ore coplex variant - C-LexRank (Qazvinian and Radev, 28), and thus we choose LexRank, rather than C- LexRank, to represent graph-based suarization ethods for coparison in this paper. with only the reference papers (i.e. the target paper s content is not considered). Those ethods are signed by -WT. To evaluate the effectiveness of the SVR odels we eploy, we ipleent a baseline syste RWGOF that uses the rando walk scores as the iportant scores of the sentences and take the scores as inputs for the sae global optiization fraework as our syste to generate the related work section. The rando walk scores are coputed for the sentences in the reference papers and the target paper, respectively. We use the ROUGE toolkit to evaluate the content quality of the generated related work sections. ROUGE (Lin, 24) is a widely used autoatic suarization evaluation ethod based on n-gra coparison. Here, we use the F- Measure scores of ROUGE-1, ROUGE-2 and ROUGE-SU4. The odel texts are set as the gold related work sections extracted fro the target papers, and word steing is utilized. ROUGE-N is an n-gra based easure between a candidate text and a reference text. The recall oriented score, the precision oriented score and the F-easure score for ROUGE-N are coputed as follows: ROUGE N Recall = S {Reference Text} gra n Count atch (gra n ) / S {Reference Text} gra n Count(gra n ) (15) ROUGE N Precision = S {Reference Text} gra n Count atch (gra n ) / S {Candidate Text} gra n Count(gra n ) (16) ROUGE N F easure = 2 ROUGE N Recall ROUGE N Precision / ROUGE N Recall + ROUGE N Precision (17) where n stands for the length of the n-gra gra n, and Count atch (gra n ) is the axiu nuber of n-gras co-occurring in a candidate text and a reference text. In addition, we conducted a user study to subjectively evaluate the related work sections to get ore evidences. We selected the related work sections generated by different ethods for 15 rando target papers in the test set. We asked three huan judges to follow an evaluation guideline we design and evaluate these related work sections. The huan judges are graduate students in the coputer science field and they did not know the identities of the evaluated related work sections. They were asked to give a rating on a scale of 1 (very poor) to 5 (very good) for the correctness, readability and usefulness of the related work sections, respectively: 163

8 λ1 ROUGE-1 ROUGE-2 ROUGE-SU4 λ1 ROUGE-1 ROUGE-2 ROUGE-SU λ2.6 λ λ λ Figure 3: Paraeter influences (horizontal, vertical axis are λ 1, λ 2, respectively, λ 3 = 1 λ 1 λ 2 ) 1) Correctness: Is the related work section actually related to the target paper? 2) Readability: Is the related work section easy for the readers to read and grasp the key content? 3) Usefulness: Is the related work section useful for the author to prepare their final related work section? Paired T-Tests are applied to both the ROUGE scores and rating scores for coparing ARWG and baselines and coparing the systes with WT and without WT. 5.2 Results and Discussion Table 4: ROUGE F-easure coparison results Method ROUGE-1 ROUGE-2 ROUGE- Mead- WT LexRank- WT ARWG- WT SU {1,2}.9987 {1,2} {1}#{2} Mead.4112 {1}.9642 {1} {1} LexRank {2}.19 {2}.1767 {2} ARWG {1 5} {1 5} {1 5} ( * represents pairwise t-test value p <.1; # represents p <.5; the nubers in the brackets represent the indices of the ethods copared, e.g. 1 for MEAD-WT, 2 for LexRank-WT, etc.) Table 5: Average rating scores of judges Method Correctness Readability Usefulness Mead LexRank ARWG # 3.42 # # ( *# represents pairwise t-test value p <.1, copared with Mead and LexRank, respectively.) Table 6: ROUGE F-easure coparison of different sentence iportance scores Method ROUGE-1 ROUGE-2 ROUGE-SU4 RWGOF ARWG The evaluation results over ROUGE etrics are presented in Table 4. It shows that our proposed syste can get higher ROUGE scores, i.e., better content quality. In our syste, we split the sentence set into different topic-biased parts, and the iportance scores of sentences in the target paper and reference papers are learned differently. So the obtained iportance scores of the sentences are ore reliable. The global optiization fraework considers the extraction of both the previous work part and the own work part. We can see the iportance of the own work part by coparing the results of the ethods with or without considering the own work part. MEAD, LexRank and our ethod all get a significant iproveent after considering the own work part by extracting sentences fro the target paper. The results also prove our assuption about the related work section structure. Figure 3 presents the fluctuation of ROUGE scores when tuning the paraeters λ 1, λ 2 and λ 3. We can see our ethod generally perfors better than the baselines. All the three parts in the objective function are useful to generate related work sections with good quality. The average scores rated by huan judges for each ethod are showed in Table 5. We can see that the related work sections generated by our syste are ore related to the target papers. Moreover, because of the good structure of our generated related work sections, our generated related work sections are considered ore readable and ore useful for the author to prepare the final related work sections. T-test results show that the perforance iproveents of our ethod over baselines are statistically significant on both autoatic and anual evaluations. Most of p-values for t-test are far saller than.1. Overall, the results indicate that our ethod can generate uch better related work sections 1631

9 than the baselines on both autoatic and huan evaluations. Table 6 shows the coparison results between ARWG and RWGOF. We can see ARWG perfors better than RWGOF. It proves that the SVR odels can better estiate the iportance scores of the sentences. For the SVR odels are trained fro the large dataset, the sentence scores predicted by the SVR odels can be ore reliable to be used in the global optiization fraework. 6 Conclusion and Future Work This paper proposes a novel syste called ARWG to generate related work sections for acadeic papers. It first exploits a PLSA odel to split the sentence set of the given papers into different topic-biased parts, and then applies regression odels to learn the iportance scores of the sentences. At last an optiization fraework is proposed to generate the related work section. Evaluation results show that our syste can generate uch better related work sections than the baseline ethods. In future work, we will ake use of citation sentences to iprove our syste. Citation sentences are the sentences that contains an explicit reference to another paper and they usually highlight the ost iportant aspects of the cited papers. So citation sentences are likely to contain iportant and rich inforation for generating related work sections. Acknowledgents The work was supported by National Natural Science Foundation of China ( , ), Beijing Nova Progra (28B3) and National Hi-Tech Research and Developent Progra (863 Progra) of China (212AA1111). We also thank the anonyous reviewers for very helpful coents. The corresponding author of this paper, according to the eaning given to this role by Peking University, is Xiaojun Wan. Reference Nitin Agarwal, Kiran Gvr, Ravi Shankar Reddy, and Carolyn Penstein Rosé Towards ultidocuent suarization of scientific articles: aking interesting coparisons with SciSu. In Proceedings of the Workshop on Autoatic Suarization for Different Genres, Media, and Languages, pp Association for Coputational Linguistics. Phyllis B. Baxendale Machine-ade index for technical literature: an experient. IBM Journal of Research and Developent 2, no. 4: Taylor Berg-Kirkpatrick, Dan Gillick, and Dan Klein Jointly learning to extract and copress. In Proceedings of the 49th Annual Meeting of the Association for Coputational Linguistics: Huan Language Technologies-Volue 1, pp Association for Coputational Linguistics. Chih-Chung Chang, and Chih-Jen Lin LIBSVM: a library for support vector achines. ACM Transactions on Intelligent Systes and Technology (TIST) 2, no. 3: 27. John M. Conroy, and Dianne P. O'leary. 21. Text suarization via hidden arkov odels. In Proceedings of the 24th annual international ACM SIGIR conference on Research and developent in inforation retrieval, pp ACM. Harold P. Edundson New ethods in autoatic extracting. Journal of the ACM (JACM) 16, no. 2: Günes Erkan, and Dragoir R. Radev. 24. LexPageRank: Prestige in Multi-Docuent Text Suarization. In EMNLP, vol. 4, pp Günes Erkan, and Dragoir R. Radev. 24. LexRank: Graph-based lexical centrality as salience in text suarization. J. Artif. Intell. Res.(JAIR) 22, no. 1: Diitrios Galanis, Gerasios Lapouras, and Ion Androutsopoulos Extractive Multi- Docuent Suarization with Integer Linear Prograing and Support Vector Regression. In COLING, pp Diitrios Galanis, and Prodroos Malakasiotis. 28. Aueb at tac 28. InProceedings of the TAC 28 Workshop. Dan Gillick, and Benoit Favre. 29. A scalable global odel for suarization. InProceedings of the Workshop on Integer Linear Prograing for Natural Langauge Processing, pp Association for Coputational Linguistics. Cong Duy Vu Hoang, and Min-Yen Kan. 21. Towards autoated related work suarization. In Proceedings of the 23rd International Conference on Coputational Linguistics: Posters, pp Association for Coputational Linguistics. Lei Huang, Yanxiang He, Furu Wei, and Wenjie Li. 21. Modeling docuent suarization as ultiobjective optiization. In Intelligent Inforation Technology and Security Inforatics (IITSI), 21 Third International Syposiu on, pp IEEE. 1632

10 Thoas Hofann Probabilistic latent seantic indexing. In Proceedings of the 22nd annual international ACM SIGIR conference on Research and developent in inforation retrieval, pp ACM. Chin-Yew Lin. 24. Rouge: A package for autoatic evaluation of suaries. InText Suarization Branches Out: Proceedings of the ACL-4 Workshop, pp Hans Peter Luhn The autoatic creation of literature abstracts. IBM Journal of research and developent 2, no. 2: Inderjeet Mani, and Eric Bloedorn Suarizing siilarities and differences aong related docuents. Inforation Retrieval 1, no. 1-2: Ryan McDonald. 27. A study of global inference algoriths in ulti-docuent suarization. Springer Berlin Heidelberg. Qiaozhu Mei, and ChengXiang Zhai. 28. Generating Ipact-Based Suaries for Scientific Literature. In ACL, vol. 8, pp Rada Mihalcea, and Paul Tarau. 25. A language independent algorith for single and ultiple docuent suarization. Rada Mihalcea, and Hakan Ceylan. 27. Explorations in Autoatic Book Suarization. In EMNLP-CoNLL, pp Saif Mohaad, Bonnie Dorr, Melissa Egan, Ahed Hassan, Pradeep Muthukrishan, Vahed Qazvinian, Dragoir Radev, and David Zajic. 29. Using citations to generate surveys of scientific paradigs. In Proceedings of Huan Language Technologies: The 29 Annual Conference of the North Aerican Chapter of the Association for Coputational Linguistics, pp Association for Coputational Linguistics. Preslav Nakov, Ariel Schwartz, and M. Hearst. 24. Citation sentences for seantic analysis of bioscience text. In Proceedings of the SIGIR'4 workshop on Search and Discovery in Bioinforatics. Raesh M. Nallapati, Ar Ahed, Eric P. Xing, and Willia W. Cohen. 28. Joint latent topic odels for text and citations. In Proceedings of the 14th ACM SIGKDD international conference on Knowledge discovery and data ining, pp ACM. Vahed Qazvinian, and Dragoir R. Radev. 28. Scientific paper suarization using citation suary networks. In Proceedings of the 22nd International Conference on Coputational Linguistics- Volue 1, pp Association for Coputational Linguistics. You Ouyang, Sujian Li, and Wenjie Li. 27. Developing learning strategies for topic-based suarization. In Proceedings of the sixteenth ACM conference on Conference on inforation and knowledge anageent, pp ACM. Dragoir Radev, Tiothy Allison, Sasha Blair- Goldensohn, John Blitzer, Arda Celebi, Stanko Diitrov, Elliott Drabek et al. 24. MEAD-a platfor for ultidocuent ultilingual text suarization. Proceedings of the 4th International Conference on Language Resources and Evaluation (LREC 24). Ariel S. Schwartz, and Marti Hearst. 26. Suarizing key concepts using citation sentences. In Proceedings of the Workshop on Linking Natural Language Processing and Biology: Towards Deeper Biological Literature Analysis, pp Association for Coputational Linguistics. Dou Shen, Jian-Tao Sun, Hua Li, Qiang Yang, and Zheng Chen. 27. Docuent Suarization Using Conditional Rando Fields. In IJCAI, vol. 7, pp Andreas Stolcke, Klaus Ries, Noah Coccaro, Elizabeth Shriberg, Rebecca Bates, Daniel Jurafsky, Paul Taylor, Rachel Martin, Carol Van Ess- Dykea, and Marie Meteer. 2. Dialogue act odeling for autoatic tagging and recognition of conversational speech. Coputational linguistics 26, no. 3: Kristian Woodsend, and Mirella Lapata Multiple aspect suarization using integer linear prograing. In Proceedings of the 212 Joint Conference on Epirical Methods in Natural Language Processing and Coputational Natural Language Learning, pp Association for Coputational Linguistics. Shasha Xie, Benoit Favre, Dilek Hakkani-Tür, and Yang Liu. 29. Leveraging sentence weights in a concept-based optiization fraework for extractive eeting suarization. In INTERSPEECH, pp Hui Yang, and Jaie Callan. 29. A etric-based fraework for autoatic taxonoy induction. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: Volue 1-Volue 1, pp Association for Coputational Linguistics. Ozge Yeloglu, Evangelos Milios, and Nur Zincir- Heywood Multi-docuent suarization of scientific corpora. In Proceedings of the 211 ACM Syposiu on Applied Coputing, pp ACM. 1633

Applying Multiple Neural Networks on Large Scale Data

Applying Multiple Neural Networks on Large Scale Data 0 International Conference on Inforation and Electronics Engineering IPCSIT vol6 (0) (0) IACSIT Press, Singapore Applying Multiple Neural Networks on Large Scale Data Kritsanatt Boonkiatpong and Sukree

More information

Software Quality Characteristics Tested For Mobile Application Development

Software Quality Characteristics Tested For Mobile Application Development Thesis no: MGSE-2015-02 Software Quality Characteristics Tested For Mobile Application Developent Literature Review and Epirical Survey WALEED ANWAR Faculty of Coputing Blekinge Institute of Technology

More information

PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO

PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO Bulletin of the Transilvania University of Braşov Series I: Engineering Sciences Vol. 4 (53) No. - 0 PERFORMANCE METRICS FOR THE IT SERVICES PORTFOLIO V. CAZACU I. SZÉKELY F. SANDU 3 T. BĂLAN Abstract:

More information

Fuzzy Sets in HR Management

Fuzzy Sets in HR Management Acta Polytechnica Hungarica Vol. 8, No. 3, 2011 Fuzzy Sets in HR Manageent Blanka Zeková AXIOM SW, s.r.o., 760 01 Zlín, Czech Republic blanka.zekova@sezna.cz Jana Talašová Faculty of Science, Palacký Univerzity,

More information

The Research of Measuring Approach and Energy Efficiency for Hadoop Periodic Jobs

The Research of Measuring Approach and Energy Efficiency for Hadoop Periodic Jobs Send Orders for Reprints to reprints@benthascience.ae 206 The Open Fuels & Energy Science Journal, 2015, 8, 206-210 Open Access The Research of Measuring Approach and Energy Efficiency for Hadoop Periodic

More information

CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS

CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS 641 CRM FACTORS ASSESSMENT USING ANALYTIC HIERARCHY PROCESS Marketa Zajarosova 1* *Ph.D. VSB - Technical University of Ostrava, THE CZECH REPUBLIC arketa.zajarosova@vsb.cz Abstract Custoer relationship

More information

Machine Learning Applications in Grid Computing

Machine Learning Applications in Grid Computing Machine Learning Applications in Grid Coputing George Cybenko, Guofei Jiang and Daniel Bilar Thayer School of Engineering Dartouth College Hanover, NH 03755, USA gvc@dartouth.edu, guofei.jiang@dartouth.edu

More information

High Performance Chinese/English Mixed OCR with Character Level Language Identification

High Performance Chinese/English Mixed OCR with Character Level Language Identification 2009 0th International Conference on Docuent Analysis and Recognition High Perforance Chinese/English Mixed OCR with Character Level Language Identification Kai Wang Institute of Machine Intelligence,

More information

Analyzing Spatiotemporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy

Analyzing Spatiotemporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy Vol. 9, No. 5 (2016), pp.303-312 http://dx.doi.org/10.14257/ijgdc.2016.9.5.26 Analyzing Spatioteporal Characteristics of Education Network Traffic with Flexible Multiscale Entropy Chen Yang, Renjie Zhou

More information

Online Bagging and Boosting

Online Bagging and Boosting Abstract Bagging and boosting are two of the ost well-known enseble learning ethods due to their theoretical perforance guarantees and strong experiental results. However, these algoriths have been used

More information

Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona Network

Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona Network 2013 European Control Conference (ECC) July 17-19, 2013, Zürich, Switzerland. Extended-Horizon Analysis of Pressure Sensitivities for Leak Detection in Water Distribution Networks: Application to the Barcelona

More information

INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE SYSTEMS

INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE SYSTEMS Artificial Intelligence Methods and Techniques for Business and Engineering Applications 210 INTEGRATED ENVIRONMENT FOR STORING AND HANDLING INFORMATION IN TASKS OF INDUCTIVE MODELLING FOR BUSINESS INTELLIGENCE

More information

An Integrated Approach for Monitoring Service Level Parameters of Software-Defined Networking

An Integrated Approach for Monitoring Service Level Parameters of Software-Defined Networking International Journal of Future Generation Counication and Networking Vol. 8, No. 6 (15), pp. 197-4 http://d.doi.org/1.1457/ijfgcn.15.8.6.19 An Integrated Approach for Monitoring Service Level Paraeters

More information

Factored Models for Probabilistic Modal Logic

Factored Models for Probabilistic Modal Logic Proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence (2008 Factored Models for Probabilistic Modal Logic Afsaneh Shirazi and Eyal Air Coputer Science Departent, University of Illinois

More information

Use of extrapolation to forecast the working capital in the mechanical engineering companies

Use of extrapolation to forecast the working capital in the mechanical engineering companies ECONTECHMOD. AN INTERNATIONAL QUARTERLY JOURNAL 2014. Vol. 1. No. 1. 23 28 Use of extrapolation to forecast the working capital in the echanical engineering copanies A. Cherep, Y. Shvets Departent of finance

More information

ASIC Design Project Management Supported by Multi Agent Simulation

ASIC Design Project Management Supported by Multi Agent Simulation ASIC Design Project Manageent Supported by Multi Agent Siulation Jana Blaschke, Christian Sebeke, Wolfgang Rosenstiel Abstract The coplexity of Application Specific Integrated Circuits (ASICs) is continuously

More information

Managing Complex Network Operation with Predictive Analytics

Managing Complex Network Operation with Predictive Analytics Managing Coplex Network Operation with Predictive Analytics Zhenyu Huang, Pak Chung Wong, Patrick Mackey, Yousu Chen, Jian Ma, Kevin Schneider, and Frank L. Greitzer Pacific Northwest National Laboratory

More information

An Innovate Dynamic Load Balancing Algorithm Based on Task

An Innovate Dynamic Load Balancing Algorithm Based on Task An Innovate Dynaic Load Balancing Algorith Based on Task Classification Hong-bin Wang,,a, Zhi-yi Fang, b, Guan-nan Qu,*,c, Xiao-dan Ren,d College of Coputer Science and Technology, Jilin University, Changchun

More information

Performance Evaluation of Machine Learning Techniques using Software Cost Drivers

Performance Evaluation of Machine Learning Techniques using Software Cost Drivers Perforance Evaluation of Machine Learning Techniques using Software Cost Drivers Manas Gaur Departent of Coputer Engineering, Delhi Technological University Delhi, India ABSTRACT There is a treendous rise

More information

Media Adaptation Framework in Biofeedback System for Stroke Patient Rehabilitation

Media Adaptation Framework in Biofeedback System for Stroke Patient Rehabilitation Media Adaptation Fraework in Biofeedback Syste for Stroke Patient Rehabilitation Yinpeng Chen, Weiwei Xu, Hari Sundara, Thanassis Rikakis, Sheng-Min Liu Arts, Media and Engineering Progra Arizona State

More information

REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES

REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES REQUIREMENTS FOR A COMPUTER SCIENCE CURRICULUM EMPHASIZING INFORMATION TECHNOLOGY SUBJECT AREA: CURRICULUM ISSUES Charles Reynolds Christopher Fox reynolds @cs.ju.edu fox@cs.ju.edu Departent of Coputer

More information

Local Area Network Management

Local Area Network Management Technology Guidelines for School Coputer-based Technologies Local Area Network Manageent Local Area Network Manageent Introduction This docuent discusses the tasks associated with anageent of Local Area

More information

Standards and Protocols for the Collection and Dissemination of Graduating Student Initial Career Outcomes Information For Undergraduates

Standards and Protocols for the Collection and Dissemination of Graduating Student Initial Career Outcomes Information For Undergraduates National Association of Colleges and Eployers Standards and Protocols for the Collection and Disseination of Graduating Student Initial Career Outcoes Inforation For Undergraduates Developed by the NACE

More information

Keyphrase Extraction for Scholarly Big Data

Keyphrase Extraction for Scholarly Big Data Keyphrase Extraction for Scholarly Big Data Cornelia Caragea Computer Science and Engineering University of North Texas July 10, 2015 Scholarly Big Data Large number of scholarly documents on the Web PubMed

More information

Evaluating Inventory Management Performance: a Preliminary Desk-Simulation Study Based on IOC Model

Evaluating Inventory Management Performance: a Preliminary Desk-Simulation Study Based on IOC Model Evaluating Inventory Manageent Perforance: a Preliinary Desk-Siulation Study Based on IOC Model Flora Bernardel, Roberto Panizzolo, and Davide Martinazzo Abstract The focus of this study is on preliinary

More information

arxiv:0805.1434v1 [math.pr] 9 May 2008

arxiv:0805.1434v1 [math.pr] 9 May 2008 Degree-distribution stability of scale-free networs Zhenting Hou, Xiangxing Kong, Dinghua Shi,2, and Guanrong Chen 3 School of Matheatics, Central South University, Changsha 40083, China 2 Departent of

More information

Image restoration for a rectangular poor-pixels detector

Image restoration for a rectangular poor-pixels detector Iage restoration for a rectangular poor-pixels detector Pengcheng Wen 1, Xiangjun Wang 1, Hong Wei 2 1 State Key Laboratory of Precision Measuring Technology and Instruents, Tianjin University, China 2

More information

Research Article Performance Evaluation of Human Resource Outsourcing in Food Processing Enterprises

Research Article Performance Evaluation of Human Resource Outsourcing in Food Processing Enterprises Advance Journal of Food Science and Technology 9(2): 964-969, 205 ISSN: 2042-4868; e-issn: 2042-4876 205 Maxwell Scientific Publication Corp. Subitted: August 0, 205 Accepted: Septeber 3, 205 Published:

More information

An improved TF-IDF approach for text classification *

An improved TF-IDF approach for text classification * Zhang et al. / J Zheiang Univ SCI 2005 6A(1:49-55 49 Journal of Zheiang University SCIECE ISS 1009-3095 http://www.zu.edu.cn/zus E-ail: zus@zu.edu.cn An iproved TF-IDF approach for text classification

More information

CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY

CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY CLOSED-LOOP SUPPLY CHAIN NETWORK OPTIMIZATION FOR HONG KONG CARTRIDGE RECYCLING INDUSTRY Y. T. Chen Departent of Industrial and Systes Engineering Hong Kong Polytechnic University, Hong Kong yongtong.chen@connect.polyu.hk

More information

A framework for performance monitoring, load balancing, adaptive timeouts and quality of service in digital libraries

A framework for performance monitoring, load balancing, adaptive timeouts and quality of service in digital libraries Int J Digit Libr (2000) 3: 9 35 INTERNATIONAL JOURNAL ON Digital Libraries Springer-Verlag 2000 A fraework for perforance onitoring, load balancing, adaptive tieouts and quality of service in digital libraries

More information

A Scalable Application Placement Controller for Enterprise Data Centers

A Scalable Application Placement Controller for Enterprise Data Centers W WWW 7 / Track: Perforance and Scalability A Scalable Application Placeent Controller for Enterprise Data Centers Chunqiang Tang, Malgorzata Steinder, Michael Spreitzer, and Giovanni Pacifici IBM T.J.

More information

An Approach to Combating Free-riding in Peer-to-Peer Networks

An Approach to Combating Free-riding in Peer-to-Peer Networks An Approach to Cobating Free-riding in Peer-to-Peer Networks Victor Ponce, Jie Wu, and Xiuqi Li Departent of Coputer Science and Engineering Florida Atlantic University Boca Raton, FL 33431 April 7, 2008

More information

Cooperative Caching for Adaptive Bit Rate Streaming in Content Delivery Networks

Cooperative Caching for Adaptive Bit Rate Streaming in Content Delivery Networks Cooperative Caching for Adaptive Bit Rate Streaing in Content Delivery Networs Phuong Luu Vo Departent of Coputer Science and Engineering, International University - VNUHCM, Vietna vtlphuong@hciu.edu.vn

More information

Dynamic Placement for Clustered Web Applications

Dynamic Placement for Clustered Web Applications Dynaic laceent for Clustered Web Applications A. Karve, T. Kibrel, G. acifici, M. Spreitzer, M. Steinder, M. Sviridenko, and A. Tantawi IBM T.J. Watson Research Center {karve,kibrel,giovanni,spreitz,steinder,sviri,tantawi}@us.ib.co

More information

Data Streaming Algorithms for Estimating Entropy of Network Traffic

Data Streaming Algorithms for Estimating Entropy of Network Traffic Data Streaing Algoriths for Estiating Entropy of Network Traffic Ashwin Lall University of Rochester Vyas Sekar Carnegie Mellon University Mitsunori Ogihara University of Rochester Jun (Ji) Xu Georgia

More information

Preference-based Search and Multi-criteria Optimization

Preference-based Search and Multi-criteria Optimization Fro: AAAI-02 Proceedings. Copyright 2002, AAAI (www.aaai.org). All rights reserved. Preference-based Search and Multi-criteria Optiization Ulrich Junker ILOG 1681, route des Dolines F-06560 Valbonne ujunker@ilog.fr

More information

Energy Proportionality for Disk Storage Using Replication

Energy Proportionality for Disk Storage Using Replication Energy Proportionality for Disk Storage Using Replication Jinoh Ki and Doron Rote Lawrence Berkeley National Laboratory University of California, Berkeley, CA 94720 {jinohki,d rote}@lbl.gov Abstract Energy

More information

Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects

Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects Evaluating the Effectiveness of Task Overlapping as a Risk Response Strategy in Engineering Projects Lucas Grèze Robert Pellerin Nathalie Perrier Patrice Leclaire February 2011 CIRRELT-2011-11 Bureaux

More information

Comment on On Discriminative vs. Generative Classifiers: A Comparison of Logistic Regression and Naive Bayes

Comment on On Discriminative vs. Generative Classifiers: A Comparison of Logistic Regression and Naive Bayes Coent on On Discriinative vs. Generative Classifiers: A Coparison of Logistic Regression and Naive Bayes Jing-Hao Xue (jinghao@stats.gla.ac.uk) and D. Michael Titterington (ike@stats.gla.ac.uk) Departent

More information

An Improved Decision-making Model of Human Resource Outsourcing Based on Internet Collaboration

An Improved Decision-making Model of Human Resource Outsourcing Based on Internet Collaboration International Journal of Hybrid Inforation Technology, pp. 339-350 http://dx.doi.org/10.14257/hit.2016.9.4.28 An Iproved Decision-aking Model of Huan Resource Outsourcing Based on Internet Collaboration

More information

International Journal of Management & Information Systems First Quarter 2012 Volume 16, Number 1

International Journal of Management & Information Systems First Quarter 2012 Volume 16, Number 1 International Journal of Manageent & Inforation Systes First Quarter 2012 Volue 16, Nuber 1 Proposal And Effectiveness Of A Highly Copelling Direct Mail Method - Establishent And Deployent Of PMOS-DM Hisatoshi

More information

ADJUSTING FOR QUALITY CHANGE

ADJUSTING FOR QUALITY CHANGE ADJUSTING FOR QUALITY CHANGE 7 Introduction 7.1 The easureent of changes in the level of consuer prices is coplicated by the appearance and disappearance of new and old goods and services, as well as changes

More information

Searching strategy for multi-target discovery in wireless networks

Searching strategy for multi-target discovery in wireless networks Searching strategy for ulti-target discovery in wireless networks Zhao Cheng, Wendi B. Heinzelan Departent of Electrical and Coputer Engineering University of Rochester Rochester, NY 467 (585) 75-{878,

More information

Equivalent Tapped Delay Line Channel Responses with Reduced Taps

Equivalent Tapped Delay Line Channel Responses with Reduced Taps Equivalent Tapped Delay Line Channel Responses with Reduced Taps Shweta Sagari, Wade Trappe, Larry Greenstein {shsagari, trappe, ljg}@winlab.rutgers.edu WINLAB, Rutgers University, North Brunswick, NJ

More information

Entity Search Engine: Towards Agile Best-Effort Information Integration over the Web

Entity Search Engine: Towards Agile Best-Effort Information Integration over the Web Entity Search Engine: Towards Agile Best-Effort Inforation Integration over the Web Tao Cheng, Kevin Chen-Chuan Chang University of Illinois at Urbana-Chapaign {tcheng3, kcchang}@cs.uiuc.edu. INTRODUCTION

More information

WORK TRAVEL MODE CHOICE MODELING USING DATA MINING: DECISION TREES AND NEURAL NETWORKS

WORK TRAVEL MODE CHOICE MODELING USING DATA MINING: DECISION TREES AND NEURAL NETWORKS TRB Paper No. 03-4348 WORK TRAVEL MODE CHOICE MODELING USING DATA MINING: DECISION TREES AND NEURAL NETWORKS Chi Xie Research Assistant Departent of Civil and Environental Engineering University of Massachusetts,

More information

Botnets Detection Based on IRC-Community

Botnets Detection Based on IRC-Community Botnets Detection Based on IRC-Counity Wei Lu and Ali A. Ghorbani Network Security Laboratory, Faculty of Coputer Science University of New Brunswick, Fredericton, NB E3B 5A3, Canada {wlu, ghorbani}@unb.ca

More information

Optimal Resource-Constraint Project Scheduling with Overlapping Modes

Optimal Resource-Constraint Project Scheduling with Overlapping Modes Optial Resource-Constraint Proect Scheduling with Overlapping Modes François Berthaut Lucas Grèze Robert Pellerin Nathalie Perrier Adnène Hai February 20 CIRRELT-20-09 Bureaux de Montréal : Bureaux de

More information

An online sulfur monitoring system can improve process balance sheets

An online sulfur monitoring system can improve process balance sheets Originally appeared in: February 2007, pgs 109-116. Used with perission. An online sulfur onitoring syste can iprove process balance sheets A Canadian gas processor used this technology to eet environental

More information

An Optimal Task Allocation Model for System Cost Analysis in Heterogeneous Distributed Computing Systems: A Heuristic Approach

An Optimal Task Allocation Model for System Cost Analysis in Heterogeneous Distributed Computing Systems: A Heuristic Approach An Optial Tas Allocation Model for Syste Cost Analysis in Heterogeneous Distributed Coputing Systes: A Heuristic Approach P. K. Yadav Central Building Research Institute, Rooree- 247667, Uttarahand (INDIA)

More information

Online Community Detection for Large Complex Networks

Online Community Detection for Large Complex Networks Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence Online Counity Detection for Large Coplex Networks Wangsheng Zhang, Gang Pan, Zhaohui Wu, Shijian Li Departent

More information

Study on the development of statistical data on the European security technological and industrial base

Study on the development of statistical data on the European security technological and industrial base Study on the developent of statistical data on the European security technological and industrial base Security Sector Survey Analysis: France Client: European Coission DG Migration and Hoe Affairs Brussels,

More information

Real Time Target Tracking with Binary Sensor Networks and Parallel Computing

Real Time Target Tracking with Binary Sensor Networks and Parallel Computing Real Tie Target Tracking with Binary Sensor Networks and Parallel Coputing Hong Lin, John Rushing, Sara J. Graves, Steve Tanner, and Evans Criswell Abstract A parallel real tie data fusion and target tracking

More information

Method of supply chain optimization in E-commerce

Method of supply chain optimization in E-commerce MPRA Munich Personal RePEc Archive Method of supply chain optiization in E-coerce Petr Suchánek and Robert Bucki Silesian University - School of Business Adinistration, The College of Inforatics and Manageent

More information

Efficient Key Management for Secure Group Communications with Bursty Behavior

Efficient Key Management for Secure Group Communications with Bursty Behavior Efficient Key Manageent for Secure Group Counications with Bursty Behavior Xukai Zou, Byrav Raaurthy Departent of Coputer Science and Engineering University of Nebraska-Lincoln Lincoln, NE68588, USA Eail:

More information

RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION. Henrik Kure

RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION. Henrik Kure RECURSIVE DYNAMIC PROGRAMMING: HEURISTIC RULES, BOUNDING AND STATE SPACE REDUCTION Henrik Kure Dina, Danish Inforatics Network In the Agricultural Sciences Royal Veterinary and Agricultural University

More information

AutoHelp. An 'Intelligent' Case-Based Help Desk Providing. Web-Based Support for EOSDIS Customers. A Concept and Proof-of-Concept Implementation

AutoHelp. An 'Intelligent' Case-Based Help Desk Providing. Web-Based Support for EOSDIS Customers. A Concept and Proof-of-Concept Implementation //j yd xd/_ ' Year One Report ":,/_i',:?,2... i" _.,.j- _,._".;-/._. ","/ AutoHelp An 'Intelligent' Case-Based Help Desk Providing Web-Based Support for EOSDIS Custoers A Concept and Proof-of-Concept Ipleentation

More information

Audio Engineering Society. Convention Paper. Presented at the 119th Convention 2005 October 7 10 New York, New York USA

Audio Engineering Society. Convention Paper. Presented at the 119th Convention 2005 October 7 10 New York, New York USA Audio Engineering Society Convention Paper Presented at the 119th Convention 2005 October 7 10 New York, New York USA This convention paper has been reproduced fro the authors advance anuscript, without

More information

Evaluating Software Quality of Vendors using Fuzzy Analytic Hierarchy Process

Evaluating Software Quality of Vendors using Fuzzy Analytic Hierarchy Process IMECS 2008 9-2 March 2008 Hong Kong Evaluating Software Quality of Vendors using Fuzzy Analytic Hierarchy Process Kevin K.F. Yuen* Henry C.W. au Abstract This paper proposes a fuzzy Analytic Hierarchy

More information

( C) CLASS 10. TEMPERATURE AND ATOMS

( C) CLASS 10. TEMPERATURE AND ATOMS CLASS 10. EMPERAURE AND AOMS 10.1. INRODUCION Boyle s understanding of the pressure-volue relationship for gases occurred in the late 1600 s. he relationships between volue and teperature, and between

More information

6. Time (or Space) Series Analysis

6. Time (or Space) Series Analysis ATM 55 otes: Tie Series Analysis - Section 6a Page 8 6. Tie (or Space) Series Analysis In this chapter we will consider soe coon aspects of tie series analysis including autocorrelation, statistical prediction,

More information

Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and migration algorithms

Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and migration algorithms Energy Efficient VM Scheduling for Cloud Data Centers: Exact allocation and igration algoriths Chaia Ghribi, Makhlouf Hadji and Djaal Zeghlache Institut Mines-Téléco, Téléco SudParis UMR CNRS 5157 9, Rue

More information

Fuzzy Evaluation on Network Security Based on the New Algorithm of Membership Degree Transformation M(1,2,3)

Fuzzy Evaluation on Network Security Based on the New Algorithm of Membership Degree Transformation M(1,2,3) 324 JOURNAL OF NETWORKS, VOL. 4, NO. 5, JULY 29 Fuzzy Evaluation on Networ Security Based on the New Algorith of Mebership Degree Transforation M(,2,3) Hua Jiang School of Econoics and Manageent, Hebei

More information

Towards Change Management Capability Assessment Model for Contractors in Building Project

Towards Change Management Capability Assessment Model for Contractors in Building Project Middle-East Journal of Scientific Research 23 (7): 1327-1333, 2015 ISSN 1990-9233 IDOSI Publications, 2015 DOI: 10.5829/idosi.ejsr.2015.23.07.120 Towards Change Manageent Capability Assessent Model for

More information

Partitioned Elias-Fano Indexes

Partitioned Elias-Fano Indexes Partitioned Elias-ano Indexes Giuseppe Ottaviano ISTI-CNR, Pisa giuseppe.ottaviano@isti.cnr.it Rossano Venturini Dept. of Coputer Science, University of Pisa rossano@di.unipi.it ABSTRACT The Elias-ano

More information

Information Processing Letters

Information Processing Letters Inforation Processing Letters 111 2011) 178 183 Contents lists available at ScienceDirect Inforation Processing Letters www.elsevier.co/locate/ipl Offline file assignents for online load balancing Paul

More information

The AGA Evaluating Model of Customer Loyalty Based on E-commerce Environment

The AGA Evaluating Model of Customer Loyalty Based on E-commerce Environment 6 JOURNAL OF SOFTWARE, VOL. 4, NO. 3, MAY 009 The AGA Evaluating Model of Custoer Loyalty Based on E-coerce Environent Shaoei Yang Econoics and Manageent Departent, North China Electric Power University,

More information

PRODUCT REVIEW RANKING SUMMARIZATION

PRODUCT REVIEW RANKING SUMMARIZATION PRODUCT REVIEW RANKING SUMMARIZATION N.P.Vadivukkarasi, Research Scholar, Department of Computer Science, Kongu Arts and Science College, Erode. Dr. B. Jayanthi M.C.A., M.Phil., Ph.D., Associate Professor,

More information

AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES

AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES Int. J. Appl. Math. Coput. Sci., 2014, Vol. 24, No. 1, 133 149 DOI: 10.2478/acs-2014-0011 AN ALGORITHM FOR REDUCING THE DIMENSION AND SIZE OF A SAMPLE FOR DATA EXPLORATION PROCEDURES PIOTR KULCZYCKI,,

More information

Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2

Exploiting Hardware Heterogeneity within the Same Instance Type of Amazon EC2 Exploiting Hardware Heterogeneity within the Sae Instance Type of Aazon EC2 Zhonghong Ou, Hao Zhuang, Jukka K. Nurinen, Antti Ylä-Jääski, Pan Hui Aalto University, Finland; Deutsch Teleko Laboratories,

More information

A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials

A Study on the Chain Restaurants Dynamic Negotiation Games of the Optimization of Joint Procurement of Food Materials International Journal of Coputer Science & Inforation Technology (IJCSIT) Vol 6, No 1, February 2014 A Study on the Chain estaurants Dynaic Negotiation aes of the Optiization of Joint Procureent of Food

More information

Load balancing over redundant wireless sensor networks based on diffluent

Load balancing over redundant wireless sensor networks based on diffluent Load balancing over redundant wireless sensor networks based on diffluent Abstract Xikui Gao Yan ai Yun Ju School of Control and Coputer Engineering North China Electric ower University 02206 China Received

More information

This paper studies a rental firm that offers reusable products to price- and quality-of-service sensitive

This paper studies a rental firm that offers reusable products to price- and quality-of-service sensitive MANUFACTURING & SERVICE OPERATIONS MANAGEMENT Vol., No. 3, Suer 28, pp. 429 447 issn 523-464 eissn 526-5498 8 3 429 infors doi.287/so.7.8 28 INFORMS INFORMS holds copyright to this article and distributed

More information

Modeling Parallel Applications Performance on Heterogeneous Systems

Modeling Parallel Applications Performance on Heterogeneous Systems Modeling Parallel Applications Perforance on Heterogeneous Systes Jaeela Al-Jaroodi, Nader Mohaed, Hong Jiang and David Swanson Departent of Coputer Science and Engineering University of Nebraska Lincoln

More information

Modeling Nurse Scheduling Problem Using 0-1 Goal Programming: A Case Study Of Tafo Government Hospital, Kumasi-Ghana

Modeling Nurse Scheduling Problem Using 0-1 Goal Programming: A Case Study Of Tafo Government Hospital, Kumasi-Ghana Modeling Nurse Scheduling Proble Using 0-1 Goal Prograing: A Case Study Of Tafo Governent Hospital, Kuasi-Ghana Wallace Agyei, Willia Obeng-Denteh, Eanuel A. Andaa Abstract: The proble of scheduling nurses

More information

Identification and Analysis of hard disk drive in digital forensic

Identification and Analysis of hard disk drive in digital forensic Identification and Analysis of hard disk drive in digital forensic Kailash Kuar Dr. Sanjeev Sofat Dr. Naveen Aggarwal Phd(CSE) Student Prof. and Head CSE Deptt. Asst. Prof. CSE Deptt. PEC University of

More information

Research on News Video Multi-topic Extraction and Summarization

Research on News Video Multi-topic Extraction and Summarization International Journal of New Technology and Research (IJNTR) ISSN:2454-4116, Volume-2, Issue-3, March 2016 Pages 37-39 Research on News Video Multi-topic Extraction and Summarization Di Li, Hua Huo Abstract

More information

Markovian inventory policy with application to the paper industry

Markovian inventory policy with application to the paper industry Coputers and Cheical Engineering 26 (2002) 1399 1413 www.elsevier.co/locate/copcheeng Markovian inventory policy with application to the paper industry K. Karen Yin a, *, Hu Liu a,1, Neil E. Johnson b,2

More information

A Multi-Core Pipelined Architecture for Parallel Computing

A Multi-Core Pipelined Architecture for Parallel Computing Parallel & Cloud Coputing PCC Vol, Iss A Multi-Core Pipelined Architecture for Parallel Coputing Duoduo Liao *1, Sion Y Berkovich Coputing for Geospatial Research Institute Departent of Coputer Science,

More information

Using Bloom Filters to Refine Web Search Results

Using Bloom Filters to Refine Web Search Results Using Bloo Filters to Refine Web Search Results Navendu Jain Departent of Coputer Sciences University of Texas at Austin Austin, TX, 78712 nav@cs.utexas.edu Mike Dahlin Departent of Coputer Sciences University

More information

The Design and Implementation of an Enculturated Web-Based Intelligent Tutoring System

The Design and Implementation of an Enculturated Web-Based Intelligent Tutoring System The Design and Ipleentation of an Enculturated Web-Based Intelligent Tutoring Syste Phaedra Mohaed Departent of Coputing and Inforation Technology The University of the West Indies phaedra.ohaed@gail.co

More information

ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET

ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET ESTIMATING LIQUIDITY PREMIA IN THE SPANISH GOVERNMENT SECURITIES MARKET Francisco Alonso, Roberto Blanco, Ana del Río and Alicia Sanchis Banco de España Banco de España Servicio de Estudios Docuento de

More information

SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS

SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS WHITE PAPER SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS Quanti Solutions. Advancing HIM through Innovation HEALTHCARE SUPPORTING YOUR HIPAA COMPLIANCE EFFORTS Quanti Solutions. Advancing HIM through Innovation

More information

Quality evaluation of the model-based forecasts of implied volatility index

Quality evaluation of the model-based forecasts of implied volatility index Quality evaluation of the odel-based forecasts of iplied volatility index Katarzyna Łęczycka 1 Abstract Influence of volatility on financial arket forecasts is very high. It appears as a specific factor

More information

Generating Certification Authority Authenticated Public Keys in Ad Hoc Networks

Generating Certification Authority Authenticated Public Keys in Ad Hoc Networks SECURITY AND COMMUNICATION NETWORKS Published online in Wiley InterScience (www.interscience.wiley.co). Generating Certification Authority Authenticated Public Keys in Ad Hoc Networks G. Kounga 1, C. J.

More information

AUC Optimization vs. Error Rate Minimization

AUC Optimization vs. Error Rate Minimization AUC Optiization vs. Error Rate Miniization Corinna Cortes and Mehryar Mohri AT&T Labs Research 180 Park Avenue, Florha Park, NJ 0793, USA {corinna, ohri}@research.att.co Abstract The area under an ROC

More information

Automatic Name-Face Alignment to Enable Cross-Media News Retrieval

Automatic Name-Face Alignment to Enable Cross-Media News Retrieval Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence Autoatic ae-face Alignent to Enable Cross-Media ews Retrieval Yueie Zhang*, Wei Wu*, Yang Li*, Cheng Jin*, Xiangyang

More information

Adaptive Modulation and Coding for Unmanned Aerial Vehicle (UAV) Radio Channel

Adaptive Modulation and Coding for Unmanned Aerial Vehicle (UAV) Radio Channel Recent Advances in Counications Adaptive odulation and Coding for Unanned Aerial Vehicle (UAV) Radio Channel Airhossein Fereidountabar,Gian Carlo Cardarilli, Rocco Fazzolari,Luca Di Nunzio Abstract In

More information

CPU Animation. Introduction. CPU skinning. CPUSkin Scalar:

CPU Animation. Introduction. CPU skinning. CPUSkin Scalar: CPU Aniation Introduction The iportance of real-tie character aniation has greatly increased in odern gaes. Aniating eshes ia 'skinning' can be perfored on both a general purpose CPU and a ore specialized

More information

Option B: Credit Card Processing

Option B: Credit Card Processing Attachent B Option B: Credit Card Processing Request for Proposal Nuber 4404 Z1 Bidders are required coplete all fors provided in this attachent if bidding on Option B: Credit Card Processing. Note: If

More information

Resource Allocation in Wireless Networks with Multiple Relays

Resource Allocation in Wireless Networks with Multiple Relays Resource Allocation in Wireless Networks with Multiple Relays Kağan Bakanoğlu, Stefano Toasin, Elza Erkip Departent of Electrical and Coputer Engineering, Polytechnic Institute of NYU, Brooklyn, NY, 0

More information

Workflow Management in Cloud Computing

Workflow Management in Cloud Computing Workflow Manageent in Cloud Coputing Monika Bharti M.E. student Coputer Science and Engineering Departent Thapar University, Patiala Anju Bala Assistant Professor Coputer Science and Engineering Departent

More information

Design of Model Reference Self Tuning Mechanism for PID like Fuzzy Controller

Design of Model Reference Self Tuning Mechanism for PID like Fuzzy Controller Research Article International Journal of Current Engineering and Technology EISSN 77 46, PISSN 347 56 4 INPRESSCO, All Rights Reserved Available at http://inpressco.co/category/ijcet Design of Model Reference

More information

Endogenous Credit-Card Acceptance in a Model of Precautionary Demand for Money

Endogenous Credit-Card Acceptance in a Model of Precautionary Demand for Money Endogenous Credit-Card Acceptance in a Model of Precautionary Deand for Money Adrian Masters University of Essex and SUNY Albany Luis Raúl Rodríguez-Reyes University of Essex March 24 Abstract A credit-card

More information

Example: Suppose that we deposit $1000 in a bank account offering 3% interest, compounded monthly. How will our money grow?

Example: Suppose that we deposit $1000 in a bank account offering 3% interest, compounded monthly. How will our money grow? Finance 111 Finance We have to work with oney every day. While balancing your checkbook or calculating your onthly expenditures on espresso requires only arithetic, when we start saving, planning for retireent,

More information

A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS

A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS A CHAOS MODEL OF SUBHARMONIC OSCILLATIONS IN CURRENT MODE PWM BOOST CONVERTERS Isaac Zafrany and Sa BenYaakov Departent of Electrical and Coputer Engineering BenGurion University of the Negev P. O. Box

More information

Hybrid spam filtering for mobile communication

Hybrid spam filtering for mobile communication coputers & security 29 (2) 446 459 available at www.sciencedirect.co journal hoepage: www.elsevier.co/locate/cose Hybrid spa filtering for obile counication Ji Won Yoon a, Hyoungshick Ki b, *, Jun Ho Huh

More information

Algorithmica 2001 Springer-Verlag New York Inc.

Algorithmica 2001 Springer-Verlag New York Inc. Algorithica 2001) 30: 101 139 DOI: 101007/s00453-001-0003-0 Algorithica 2001 Springer-Verlag New York Inc Optial Search and One-Way Trading Online Algoriths R El-Yaniv, 1 A Fiat, 2 R M Karp, 3 and G Turpin

More information

Leak detection in open water channels

Leak detection in open water channels Proceedings of the 17th World Congress The International Federation of Autoatic Control Seoul, Korea, July 6-11, 28 Leak detection in open water channels Erik Weyer Georges Bastin Departent of Electrical

More information