Web Searching On The Vivisimo Search Engine

1y ago
26 Views
1 Downloads
418.41 KB
13 Pages
Last View : 1m ago
Last Download : 9m ago
Upload by : Elisha Lemon
Transcription

Web Searching on the Vivisimo Search EngineSherry KoshmanSchool of Information Sciences, University of Pittsburgh, 135 N. Bellefield, Pittsburgh, PA 15260.E-mail: {skoshman}@sis.pitt.eduAmanda SpinkQueensland University of Technology, Gardens Point Campus, 2 George Street, GPO Box 2434, Brisbane,QLD 4000, Australia. E-mail: ah.spink@qut.edu.auBernard J. JansenSchool of Information Sciences and Technology, The Pennsylvania State University, 329F IST Building,University Park, PA 16802. E-mail: jjansen@acm.orgThe application of clustering to Web search engine technology is a novel approach that offers structure to theinformation deluge often faced by Web searchers. Clustering methods have been well studied in research labs;however, real user searching with clustering systems inoperational Web environments is not well understood.This article reports on results from a transaction loganalysis of Vivisimo.com, which is a Web meta-searchengine that dynamically clusters users’ search results. Atransaction log analysis was conducted on 2-week’sworth of data collected from March 28 to April 4 and April25 to May 2, 2004, representing 100% of site traffic duringthese periods and 2,029,734 queries overall. The resultsshow that the highest percentage of queries containedtwo terms. The highest percentage of search sessions contained one query and was less than 1 minute in duration.Almost half of user interactions with clusters consistedof displaying a cluster’s result set, and a small percentage of interactions showed cluster tree expansion. Findings show that 11.1% of search sessions were multitasking searches, and there are a broad variety of searchtopics in multitasking search sessions. Other searchinginteractions and statistics on repeat users of the searchengine are reported. These results provide insights intosearch characteristics with a cluster-based Web searchengine and extend research into Web searching trends.IntroductionCumbersome search results lists generated by traditionalWeb search engines is a well recognized problem in Webinformation retrieval (IR). Providing the user with a meansof viewing groups of similar search results potentiallyReceived July 16, 2005; accepted November 9, 2005 2006 Wiley Periodicals, Inc. Published online 2 October 2006 in WileyInterScience (www.interscience.wiley.com). DOI: 10.1002/asi.20408enhances Web search effectiveness; however, there has beenlittle research into Web searchers’ interactions with clusteredsearch engine results. The user assumptions underlying document set clustering include the ability of the clusters toaggregate similar documents based on topic, type, chronology, or other criteria supported by the system. The clusterlabels generated are to be indicative of document content,and there is an expectation that the process should be automated and efficient. The notion is to provide an overview ofthe collection to enable users to make decisions regardingtheir selections from a document set (Hearst, 1999).Clustering search output has been incorporated in Websearch engines such as Mooter (http://mooter.com/).Mooter’s consumer searching provides clusters usingalgorithms based on psychological modeling of user information seeking. The clusters are presented in a “starburst”visualization for the user to select. The Web search engineiBoogie (http://www.iboogie.com/) is a meta-searching clustering environment that uses the Clusterizer technology(www.clusterizer.com). Each cluster is given a label basedon document content. Both the labels and the clusters arebased on term-extraction techniques that use linguistic andstatistical calculations on text summaries returned by thesearch. Similar to Vivisimo, Clusterizer technology generates clusters in real time and uses a cluster tree to structurethe cluster labels.As new clustering technologies are introduced on theWeb for IR, it is increasingly important to examine howusers interact with these systems to evaluate user–system interactions and improve system performance. This investigation is timely since its objective is to better understand thenature of user interaction with Vivisimo, a cluster-basedWeb search engine. A quantitative and qualitative transaction log analysis was conducted to examine user queriesJOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY, 57(14):1875–1887, 2006

presented to the system to determine characteristics of termdistribution, queries, and sessions. Sessions on the VivisimoWeb search engine were analyzed further for topic distribution and to examine the degree of multitasking search bytypical Web searchers. The overall goal of this research is tounderstand user interaction with a clustering Web searchengine and further extend the line of user interactionresearch in Web IR.Related StudiesClustering MethodsClustering methods have been well investigated in the IRenvironment (Baeza-Yates & Ribeiro-Neto, 1999; Korfhage,1997). Web-based clustering work in research labs includesadapting algorithms (Wang & Kitsregawa, 2002; Zamir &Etzioni, 1998; Zeng, He, Chen, Ma, & Ma, 2004) and usertesting (Chen & Dumais, 2000; Hearst & Pedersen, 1996).This investigation uses transaction log analysis to studyVivisimo user search characteristics. Similarly, a transactionlog analysis was used to evaluate user interaction withGrouper, a clustering interface for Web search engine results(Zamir & Etzioni, 1999). The findings from the Grouperstudy showed that users tended to examine more clustersthan hypothesized. The logs also were analyzed to compareGrouper with a traditional text-based interface, HuskySearch, to determine the number of documents clicked on bythe users. Their results showed that users followed moremultiple documents using the Grouper clustering interfaceand more single documents using HuskySearch.Transaction log analysis provides a useful method ofdetermining Web searcher trends. Xie and O’Hallaron(2002) analyzed 1-month’s worth of Vivisimo and Excitedata for repetition of query terms to determine cachingmethods. Their findings showed that query repetition iscommon among users and that caching offers the potentialto improve the efficiency of processing queries. Transactionlog analyses offer an unobtrusive method to study user interactions with traditional Web search engines (Jansen,Spink, & Saracevic, 2000; Spink & Jansen, 2004; Spink,Jansen, & Saracevic, 2001). These studies have shown thatWeb users typically enter two terms per query, one queryper session, and few use Boolean operators (Wolfram,Spink, Jansen, & Saracevic, 2001).MultitaskingA user’s single session with a Web search engine mayoften consist of seeking information on single or multipletopics. Recent studies have examined multitasking searchingon the Excite and AlltheWeb.com Web search engines.Spink, Ozmutlu, and Ozmutlu (2002) and Spink, Park,Jansen, and Pedersen (2006) showed that IR and Websearches often include multiple topics during a single searchsession, which we refer to as a multitasking search; however,limited knowledge exists on the characteristics and patterns1876of multitasking searches. Spink et al. (2006) found that 81%of two-query sessions on the AltaVista search engineincluded multiple topics, 91.3% of three or more query sessions included multiple topics, a broad variety of topics inmultitasking search sessions, and three or more query sessions sometimes contained frequent topic changes. Furtherresearch is needed to study the prevalence and characteristics of multitasking by other Web search engine users.This study extends the large-scale research of Web searchqueries to a cluster-based search engine. This analysis alsoprovides insight into user interaction with clusters, whichcurrently is unexplored and not well understood.Research QuestionsThe research questions addressed by this study include: What are the search characteristics of Vivisimo users, including the session length, query length, and use of queryoperators?What is the distribution of terms, query topics, sources, andlanguages used?What is the extent of cluster expansion by users?What is the distribution of clusters among post-initial searchrecords?What is the occurrence of repeated searching by Vivisimousers?What is the occurrence of multitasking sessions by Vivisimousers?This analysis reveals patterns of searchers using a clusterbased Web search engine.Research DesignVivisimo.comThe Vivisimo (http://vivisimo.com) interface contains adialog box for inputting queries and supports Boolean andexact phrase matching. The default search source is the Web,and a drop-down menu provides options for additionalsource selection (e.g., CBC, CNN, and Wisenut). Searchescan be limited by domain or host name, link content, Webpage, or Uniform Resource Locater (URL) information.Vivisimo offers an “Advanced” search form containing options for source and language selection, defining the numberand display of search results, deciding how links should beopened, and whether the content filter is applied. After a usersubmits a query, Vivisimo presents the clusters using a treemetaphor, which is similar to that used for viewing folders inWindows Explorer. The clusters appear on the left side of thepage, and the results pages are featured on the right of themain search page (Figure 1).Unlike typical Web search engines, which present lists ofsearch output, Vivisimo’s clustering feature creates dynamicpostsearch categories in a meta-searching environment.Users can click on cluster labels to retrieve results pages.Clusters can be expanded by clicking on the plus sign toreveal subclusters, and the cluster tree may be elongated byJOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—December 2006DOI: 10.1002/asi

FIG. 1.Vivisimo.com interface.clicking on the “More” option. Search terms can be enteredin the “Find in clusters” search box to search the clusters.The results pages are initially displayed as a result of theinitial search. Results pages are retrieved when the user clickson the clusters; additional results pages may be selected at thebottom of the window. Hyperlinks may be accessed for individual items, and Web pages may be previewed and thenopened in the results frame or in a new window. An item onthe results pages may be identified within the clusters byclicking on the “show in clusters” option next to the item.This highlights the clusters on the tree which contain theitem. The “Details” feature shows the number of results forthe sources searched.Data CollectionThe Vivisimo transaction log data used for this study represents two 1-week periods: (a) March 28 to April 4, 2004,containing 1,082,431 queries and (b) April 25 to May 2,2004, containing 927,303 queries. The transaction logrecorded 100% of the traffic on the Vivisimo Web site duringthese periods, and in total, 2,029,734 queries and 386,949unique Internet Protocol (IP) addresses were recorded.The records in the transaction log files contain a varietyof fields, including User Identification (an anonymous usercode assigned by the Vivisimo server), Date (the calendarday as recorded by the Vivisimo server), Time of Day (theclock time as recorded by the Vivisimo server), and QueryTerms (terms exactly as entered by the given user).Data AnalysisA relational database was created to import the transaction log, which is a flat ASCII file, and each record wasgiven a unique identifier. The initial query was located onthe basis of four fields (i.e., user identification, date, time ofday, and query terms). A user session was recreated by structuring a chronological series of actions on a given day. Aterm is any series of characters separated by white space orother separator. A query is the entire string of terms submitted by a searcher in a given instance of interaction. A sessionis the entire series of queries submitted by a user during oneinteraction with the Web search engine on a given day. Anidentical query is a query that is a copy of a previous querywithin the same user session. A repeat query is a query submitted more than once, irrespective of the user.Searches from both human users and agents are contained inthe transaction log. This analysis focused on queries submittedonly by humans, not by an automated process. There is not anestablished method to accurately identify human from nonhuman searchers, so most researchers utilizing transaction logsmust either overlook it (Cacheda & Viña, 2001) or assign sometemporal or interaction cutoff (Montgomery & Faloutsos,2001; Silverstein, Henzinger, Marais, & Moricz, 1999).A cutoff approach was used, and sessions with 100 orfewer queries were separated into another transaction log.This cutoff was selected because it is almost 50 times greaterthan the reported mean search session for human Websearchers and because it assured that human searches werenot excluded. The assumption was made that the analysisyielded a subset of the transaction log that contained queriessubmitted primarily by human searchers. The cutoff mayhave allowed for some agent or common user terminalsessions; however, it is broad enough in minimizing biasintroduced by too low of a cutoff threshold.When a searcher submits a query to Vivisimo, views adocument, and returns to the search engine, the Vivisimoserver logs this second visit with the identical user identification and query, but with a new time (i.e., the time of thesecond visit). Vivisimo assigns a unique code to identify auser’s multiple interactions with the system. This is beneficial information in determining how many of the retrievedresults pages the searcher visited from the search engine;however, it also introduces duplicate queries.JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—December 2006DOI: 10.1002/asi1877

The transaction log was collapsed by combining all identical queries submitted by the same user to provide uniquequeries for analyzing sessions, queries and terms, and pagesof results viewed. Uncollapsed sessions were used to obtainan accurate measure of the session duration and the numberof results pages visited. The number of identical queries bythe same user was recorded in a separate field within theremaining records when the sessions were collapsed.In addition to the fields for unique identifier and numberof identical queries, a field within each record containing thelength of the query, measured in terms, was included. Twoother tables for the collapsed dataset, one for term data andthe other for co-occurrence data, were generated. The termtable contains fields for a term and the number of times thata term occurs in the complete dataset. The co-occurrencetable contains fields for term pairs and the number of timesthat a pair occurs within the dataset, irrespective of order.The database now contains four tables (i.e., uncollapseddataset, collapsed dataset, terms, and co-occurrence). The datafrom these four tables were analyzed to investigate our researchquestions. The analysis was conducted using queries, usually aseries of layered queries, Visual Basic for Applications scripts,or a combination of the two. Key fields were extracted from thelog file for the clustering analysis of the dataset, and each querywas identified by a unique Vivisimo assigned code. A series ofTABLE 1.ResultsOverall ResultsTables 1 and 2 present the aggregate results for bothdatasets in this analysis.TermsTerm frequency. The top 30 terms were extracted, removing the terms without context (an, or, de, la, le, etc.), letters,and numbers (Table 3). The term distribution showed 73.3%of the top 30 terms were the same in both datasets. The mostfrequently used terms used were “download,” “new,”“software,” “windows,” and “sex.” Most of these terms represent a strong user focus on computing issues. Many of thetop 30 terms are strongly linked to computing, universities,students, travel, music, and downloading music. Althoughthe term sex is a very high-frequency term, sexual or pornographic queries formed less than 5%, or 1 in 20 searches.This is a similar finding to that of users of other Web searchengines (Spink, Jansen, Wolfram, & Saracevic, 2002).Aggregate analysis for Dataset 1 (3/28–4/04).Unique IP addressesTotal queriesTermsUniqueTotal termsMean terms per queryTerms per query0123 Term pairs UniqueTotal term pairsUsers’ modifying queriesSession length1 query per session2 queries per session3 queries per sessionResult pages viewedBoolean queriesQueries with other operatorsBoth Boolean and otherTerms not repeated in dataset100 Most frequently occurring terms1878UNIX text-manipulation commands were used to parse andcalculate statistics on the clustering de1SD5.3762JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—December 2006DOI: 10.1002/asi

TABLE 2.Aggregate analysis for Dataset 2 (4/25–5/02).Unique IPQueriesTermsUniqueTotalMean terms per queryTerms per query0123 Users’ modifying queriesSession size1 query2 queries3 queriesResult pagesBoolean queriesQueries withother operatorsBoth Boolean and otherTerms not repeatedin dataset100 Mostfrequentlyoccurring x85%0.418.830.050.8100%58Min0Term 861No.55.9%547,80918.9Term Co-OccurrenceAlthough a term analysis is useful, it is sometimes difficult to determine the specific usage of a term intended by asearcher outside the framework of a particular query. Inthese cases, a term co-occurrence (Leydesdorff, 1989) ismore helpful. Tables 4 and 5 show the co-occurrences for theVivisimo datasets, including the percentage calculated of thetop 10 co-occurring term pairs.QueriesQueries per day. Both datasets showed very similar querydistribution patterns during the week (Figure 2). The meanqueries per day were 135,304 for Dataset 1 and 115,913 forDataset 2. Approximately 80% of queries were enteredduring weekdays, with about 5% fewer queries submittedper day on the weekends.Query length. Table 6 shows the range of query lengths asdefined by the number of terms per query. Zero indicates anull search. In both datasets, the highest percentage (29.4 and30%) of queries contained two terms. Approximately 72% ofthe queries in both datasets contained one to three terms. Further analysis of the queries showed that a small percentage(2.6%) in both datasets contained Boolean operators. Approximately 20% of the queries contained other operators.Most frequent queries. The most frequent queries werecalculated for both datasets, and 80% of the top 10 repeatedqueries are alike for both datasets (Tables 7 and 8). The common terms found in both datasets are shaded in Table 8,which shows that 61.5% of terms were present in bothdatasets.SessionsSessions per day. Figure 3 shows the percentage ofsessions distributed across both datasets. The mean sessionsper day for Dataset 1 were 34,971 and 34,432 for Dataset 2.Session length by query. Table 9 shows that 1 in 3Vivisimo users entered only one term during their session.Some 1 in 5 Vivisimo users entered only two terms duringJOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—December 2006DOI: 10.1002/asi1879

TABLE 3.Top 30 term frequency.Dataset 1 (3/28–4/04)TABLE 4.Dataset 2 0.000787,5842.58092,8750.0319Dataset 1 top 10 term co-occurrences.TABLE 5.Dataset 2 top 10 term 14,096100.010,274100.0their session. In addition, 1 in 10 Vivisimo users enteredonly three terms during their session, and 3 in 10 Vivisimousers entered more than three terms in their session.Session duration and interactions. Table 10 shows the aggregate statistics for session duration. Session duration wasmeasured from the time the first query was submitted untilthe user departed the search engine for the last time (i.e.,does not return) on a given day. This definition allows for themeasurement of the total user time on the search engine and1880the time spent viewing the first and all subsequent Web documents, except the final document. The final viewing time isnot available since the Web search engine server records thetime stamp. A limitation is that the time between visits fromthe Web document to the search engine may not have beenentirely spent viewing the Web document.The average session duration is 11 2 min, and the mostfrequently occurring duration value in the dataset is less than1 min for both datasets. Table 11 shows one in two Vivisimosessions were less than 1 min in duration, 1 in 10 VivisimoJOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—December 2006DOI: 10.1002/asi

TABLE 8.RankFIG. 2.TABLE 6.Queries per day.Query length.Dataset 1 (3/28–4/04)Length012345678910 10TABLE 7Dataset 2 1314151617181920212223242526Dataset 2 top repeat queries and common terms.QueryOccurrences%“Mark Twain”Looney TunesGoogleCloningYahooEbaySexparis HiltonDictionaryyahoo.comhotmailspybot“Ralph Nader”pornmapquestfnordgameslyricssearch enginesKentucky set 1 top repeat queries.QueryOccurrences%Google“Mark Twain”CloningLooney TunesYAHOOSexEbayScience Lesson Plans“Ralph Nader”yahoo.comwww.lolitas.comTestChocolate Chipssite de blog fotos de casaisde sexoDictionaryGamesparis HiltonSexoBackgammonbritney spears picturesHotmailHomenspeladosPornbritney spears clipsLyricsCarolina 380.05420.03030.0300FIG. 3.TABLE 9.Sessions per day.Session length.Dataset 1 (3/28–4/04)QueryLength12345678910 10Dataset 2 1.91.51.15.2279,766100275,456100.0

TABLE 10.Aggregate statistics for vivisimo sessions.Dataset 1 (3/28–4/04)Additionally, one in seven queries were related to people,places, or things. These queries include personal names orthe names of locations.Dataset 2 ons1:36:3523:59:06 1s 1 min3:48:393.87214115.41:34:0123:59:44 1s 1 min3:43:493.841,035115.96Language and SourcesMeanMaxMinModeSDTABLE 11.Session duration distribution.Dataset 1 (3/28–4/04)Session duration 1 min1–5 min5–10 min10–15 min15–30 min30–60 min1–2 hr2–3 hr3–4 hr 4 hrDataset 2 00.0sessions were less than 5 min in duration, and 1 in 20Vivisimo sessions were between 5 and 10 min in duration.Overall, 45% of sessions were less than 1 min in duration forboth datasets.Table 12 shows the results from a random sample of3,600 sessions containing 4,883 queries that were classifiedinto 11 nonmutually exclusive topic categories developed bySpink, Jansen, et al. (2002). One in five queries submitted toVivisimo related to commerce, travel, employment, or theeconomy. This finding relates to the “business” and “management” terms that occur in the top 30 most frequent terms(Table 1). One in five queries were indiscernible or nonEnglish, which represents a sizable proportion of all queries.TABLE 12.Session topic distribution.TopicsNo. oftopics%Commerce, Travel, Employment,or EconomyIndiscernible or Non-EnglishPeople, Places, or ThingsComputers or InternetSocial, Culture, Ethnic, or ReligionHealth or SciencesEducation or HumanitiesSex or PornographyPerforming or Fine ArtsGovernmentEntertainment or 1396543324,8831001882Table 13 shows that the majority of queries in bothdatasets did not specify a language (90%); however, thehighest foreign-language selection was German (3%). Themajority of source requests selected the Web for Vivisimosearches (Table 14). Nine in 10 Vivisimo users did not request a language. The ranking of the German Web source isconsistent with the previous language distribution, and maybe facilitated by the availability of Vivisimo’s interface inGerman via a “Deutsch” option on the initial search page.ClustersQuery DistributionClustering data were available for analysis in the seconddataset (April 25–May 2, 2004). Figure 4 shows the query distribution with a cutoff of 20 queries. The distribution of queriesper IP address shows that the largest percentage of IP addressesexhibited one query occurrence. Small percentages of IPaddresses ( .05) contained 21 or more queries. IP addressesthat showed above 100 or more query occurrences constitutedless than .0015. This query limit was used as a cutoff point toaccount for the prevalence of network address translationbased firewalls that hide multiple users behind one IP addressand the use of dynamic addressing for dial-up connections. AVivisimo search returns three frame records. The form framedefines the frame set, including tree and list frames. The treeframe represents clusters, and the list frame corresponds to theresults pages. The total number of records was 4,219,925. Thelist records constitute 44% of the total number of records, andthe tree records represent 29% of the total.The higher percentage of list records suggests that moreresults pages were viewed than were cluster expansions.This finding is consistent with the pattern of cluster usagediscussed in the next section. Postquery records were examined to look solely at user interaction with Vivisimo. Listrecords were used approximately three and a half times morethan were the tree records. This means that the users clickedon clusters to retrieve results pages. Almost half (48.26%) ofthe postquery records involved displaying the results pagesthat come from clicking on a cluster.Cluster ExpansionCluster expansion indicates user activity in clicking on the“ ” sign next to a cluster label and expanding the tree to revealsubclusters.Figure 5 shows the distribution and extent to which clusters were expanded. Clusters are most frequently expandedonce, representing about 2.3% of the total number of allrecords (Koshman, Spink, & Jansen, 2005). The maximumnumber of clusters expanded in a record is 26 (.00014%).JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE AND TECHNOLOGY—December 2006DOI: 10.1002/asi

TABLE 13.Distribution of language requests.Dataset 1 (3/28–4/04)123456789101112TABLE 14.Dataset 2 (4/25–5/02)LanguageOccurrences%Occurrences%None FrenchDutchJapaneseAll .11,082,431100.00927,303100.0Distribution of source requests.Dataset 1 (3/28–4/04)12345678910111213WebGermanWebMSN, Netscape, Lycos,Looksmart, OvertureEnglish WebeBayPubMedGoogle, Lycos, Netscape,Looksmart, MSN, Britannica,DogpileMSN, Netscape, Lycos,Looksmart, FindWhatAltaVista, MSN, Netscape,Lycos, Looksmart, FindWhatAltaVistaMSNGoogle, Fast, MSN, BBCAll Other CombinationsFIG. 4.Dataset 2 ,9320.30.30.2

Web search engines is a well recognized problem in Web information retrieval (IR). Providing the user with a means of viewing groups of similar search results potentially enhances Web search effectiveness; however, there has been little research into Web searchers’interactions with clustered search engine results. The user assumptions .File Size: 418KB

Related Documents:

May 02, 2018 · D. Program Evaluation ͟The organization has provided a description of the framework for how each program will be evaluated. The framework should include all the elements below: ͟The evaluation methods are cost-effective for the organization ͟Quantitative and qualitative data is being collected (at Basics tier, data collection must have begun)

On an exceptional basis, Member States may request UNESCO to provide thé candidates with access to thé platform so they can complète thé form by themselves. Thèse requests must be addressed to esd rize unesco. or by 15 A ril 2021 UNESCO will provide thé nomineewith accessto thé platform via their émail address.

̶The leading indicator of employee engagement is based on the quality of the relationship between employee and supervisor Empower your managers! ̶Help them understand the impact on the organization ̶Share important changes, plan options, tasks, and deadlines ̶Provide key messages and talking points ̶Prepare them to answer employee questions

Chính Văn.- Còn đức Thế tôn thì tuệ giác cực kỳ trong sạch 8: hiện hành bất nhị 9, đạt đến vô tướng 10, đứng vào chỗ đứng của các đức Thế tôn 11, thể hiện tính bình đẳng của các Ngài, đến chỗ không còn chướng ngại 12, giáo pháp không thể khuynh đảo, tâm thức không bị cản trở, cái được

Le genou de Lucy. Odile Jacob. 1999. Coppens Y. Pré-textes. L’homme préhistorique en morceaux. Eds Odile Jacob. 2011. Costentin J., Delaveau P. Café, thé, chocolat, les bons effets sur le cerveau et pour le corps. Editions Odile Jacob. 2010. Crawford M., Marsh D. The driving force : food in human evolution and the future.

Le genou de Lucy. Odile Jacob. 1999. Coppens Y. Pré-textes. L’homme préhistorique en morceaux. Eds Odile Jacob. 2011. Costentin J., Delaveau P. Café, thé, chocolat, les bons effets sur le cerveau et pour le corps. Editions Odile Jacob. 2010. 3 Crawford M., Marsh D. The driving force : food in human evolution and the future.

MARCH 1973/FIFTY CENTS o 1 u ar CC,, tonics INCLUDING Electronics World UNDERSTANDING NEW FM TUNER SPECS CRYSTALS FOR CB BUILD: 1;: .Á Low Cóst Digital Clock ','Thé Light.Probé *Stage Lighting for thé Amateur s. Po ROCK\ MUSIC AND NOISE POLLUTION HOW WE HEAR THE WAY WE DO TEST REPORTS: - Dynacó FM -51 . ti Whárfedale W60E Speaker System' .

Glossary of Social Security Terms (Vietnamese) Term. Thuật ngữ. Giải thích. Application for a Social Security Card. Đơn xin cấp Thẻ Social Security. Mẫu đơn quý vị cần điền để xin số Social Security hoặc thẻ thay thế. Baptismal Certificate. Giấy chứng nhận rửa tội

More than words-extreme You send me flying -amy winehouse Weather with you -crowded house Moving on and getting over- john mayer Something got me started . Uptown funk-bruno mars Here comes thé sun-the beatles The long And winding road .

Phần II: Văn học phục hưng- Văn học Tây Âu thế kỷ 14- 15-16 Chương I: Khái quát Thời đại phục hưng và phong trào văn hoá phục hưng Trong hai thế kỉ XV và XVI, châu Âu dấy lên cuộc vận động tư tưởng và văn hoá mới rấ

Food outlets which focused on food quality, Service quality, environment and price factors, are thè valuable factors for food outlets to increase thè satisfaction level of customers and it will create a positive impact through word ofmouth. Keyword : Customer satisfaction, food quality, Service quality, physical environment off ood outlets .

query log) for analysis.A search engine transaction log is an electronic record of the interactions that have occurred dur-ing a searching episode between a Web search engine and users searching for content on that Web search engine. Just as transaction logs have yielded comprehensive documenta-

Lời Nói Đầu K inh Bát-Nhã (Prajna) đƣợc lƣu hành rất sớm tại Ấn độ. Khoảng 700 năm sau khi Phật diệt độ (cuối thế kỷ II đầu thế kỷ III Tây lịch), lúc Bồ-tát Long Thọ

UNESCO in consultation with thé National Commission for UNESCO as well as b non- overnmental or anizations NGOs in officiai artnershi with UNESCO. Nominations must focus on a s ecific ESD ro'ect or ro ramme. Each Member State or NGO can make u to three nominations for an édition of thé Pri

1.2. Chương Trình 0% Lãi Suất Ưu Đãi Mua Sắm không áp dụng cho Chủ thẻ Tín Dụng Thương Mại. The Installment Plan With 0% Interest is not applicable for HSBC Business Credit Card. 1.3. Loại tiền tệ được sử dụng trong Chương Trình 0% L

For centuries, Baccarat has been privileged to create masterpieces for royal households throughout the world. Honoring that legacy we have imagined a tea service as it might have been enacted in palaces from St. Petersburg to Bangalore. Pairing our menus with world-renowned Mariage Frères teas to evoke distant lands we have

HƯỚNG DẪN LỰA CHỌN DÂY & CÁP HẠ THẾ DÂY & CÁP HẠ THẾ A/ LỰA CHỌN DÂY & CÁP : Khi chọn cáp, khách hàng cần xem xét những yếu tố sau: - Dòng điện định mức - Độ sụt áp - Dòng điện ngắn mạch - Cách lắp đặt - Nhiệt độ môi trường hoặc nhiệt độ đất

Niagara University-Toronto niagara.instructure.com Canada Simon Fraser University Canvas.sfu.ca Canada University of British Columbia (UBC) Canvas.ubc.ca Canada University of Toronto learn.utoronto.ca Canada . 5 Nắm bắt xu thế phát triển công nghệ của thế giới, Trường Đại học Công nghệ .

deficiencies of full-text searching in academic library databases. Because full-text searching relies on matching words in a search query with words in online resources, it is an inefficient method of finding information in a dat

GHAMI Asia HARRIS GHAVIMI HARTLEYClarita GIL Maria GIRMA Turufat GOMES Marcio GOMEZ Luis GOMEZ Jessica GOMEZ Marie GOTTARDI Giannino GORDON Natasha GREAVES Cynthia GREENWOOD Peter GRIFFIN Daniel HABIB Assema Kedir HABIB Fatuma Kedir HABIB Jemal Kedir HABIB Merema Kedir HABIB Mehammed Kedir HABIB Mojda HABIB Shemsu Kedir HADDADI Rkia HADGAY Ismal HAKIM Hamid HAKIM Mohamed HAMDAN Rkia HAMDAN .