Multidirectional Knowledge Extraction Process for Creating Behavioral Personas Andrey A. Masiero1, Mayara G. Leite1, Lucia Vilela Leite Filgueiras2, Plinio Thomaz Aquino Jr.1 1
Centro Universitário da FEI - Fundação Educacional Inaciana Pe. Sabóia de Medeiros Av. Humberto A. Castelo Branco, 3972 09850-901 ± S. Bernardo Campo ± SP ± Brasil +55 11 4353 2900 {andreymasiero, mayara.gleite}@gmail.com,
[email protected] ABSTRACT
The consideration of user aspects in the design of computer systems is vital to maintaining the quality of these products. The search for user information as to their interests, needs, and behaviors, requires actions that can be costly. However, the use of such information in an inefficient, inaccurate manner, without the formation of knowledge discourages the process of user-centered development. Interface designers need to collect information and consume it efficiently. One option for documenting and efficiently consuming the information from user research is to apply the technique of personas. The extraction of knowledge in databases is the selection and processing of data with the purpose of identifying new patterns, provide greater accuracy on known patterns, and model the real world. Thus, this work presents a multidirectional process of extracting knowledge for creating behavioral personas. The processes of data mining are performed and compared with variations allowing a real knowledge extraction in the creation of personas. The process is multidirectional being able to be applied in both directions for the composition of any kind of personas. The results of the work demonstrate the extraction of knowledge originating from demographic information to the construction of personas focusing on behavioral aspects. Keywords
Behavioral Personas, User Modeling, User Profile, Knowledge Extraction, Data Mining INTRODUCTION
Digital evolution has changed the concepts of space, time and the diversity of contemporary society. Because of this companies do not need to occupy much space and can even be virtual. Cyberspace has lead companies to an era when Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. 10th Brazilian Symposyum on Human Factors in Computer Systems & 5th Latin American Conference on Human-Computer Interaction. IHC+CLIHC’2011 October 25-28, 2011, Porto de Galinhas, PE, Brazil. Copyright 2011 SBC. ISSN 2178-7697
2
Escola Politécnica da Universidade de São Paulo Av. Prof. Luciano Gualberto, trav. 3, 158 05508-900 ± São Paulo ± SP ± Brasil +55 11 3091 5200
[email protected] buying and selling is automated. All of these changes in behavior and needs alter the ability to provide value to companies, increasing competition and reducing the differentiation between them. This evolutionary scenario has been slightly slower in Brazil compared with developed countries, however, many Brazilian citizens have had access to information technology to carry out many activities. Thus, the decision to invest in creation processes based on Usability and HCI for the development of interfaces has been very helpful in the attendance of people's needs. However, in highly diverse populations there are people that are excluded by the organizations that operate in the electronic market due to the difficulty in developing systems that adapt to all sorts of human interaction difficulties. This is a very delicate topic when it comes to government services that should focus on social welfare, inclusion and accessibility, rather than profit. Companies or government departments that provide public services have many challenges in this direction. In the case of companies providing electronic government services, investments in usability are essential to ensure access for the diversity of users that characterizes a certain society to the services provided by the government. Therefore, the e-government system designers face challenges of access and accessibility whenever there is the need to design an interface. Research on user profiles applied to the electronic government motivated the development of this paper. User information should be collected and consumed efficiently. In this way, inside the field of technology there is a very large range of studies that shows effective methods to understand the behavior to target audiences increasingly heterogeneous. Creating personas is an acknowledged method for modeling user profiles, which help the designer to know the user profiles that are the target of a particular product. Often, persona-based profiles only express demographic information that characterize "who" is the 91
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Pageof Numbering, Headers and Footersthat more directly user a system and lack information Do not include headers, page consequently, numbers in your influences interfaces realfooters usageor(and its submission. will be added when the publications are design) e.g., These user behavior. assembled. Thus, this paper presents a knowledge extraction process
for creating multidirectional behavioral personas. Personas SECTIONS are to be because personas9-point were The said heading of behavioral a section should be these in Helvetica created by behavior Sections information, which are more useful in a bold in all-capitals. should be unnumbered. real interface project than only demographic information, Subsections normally used. process isshould said tobebeinmultidirectional, as The heading ofThe subsections Helvetica 9-point it gets behavioral information from demographic personas bold with only the initial letters capitalized. (Note: For suband vice-versa. Data mining procedures arethe performed sections and subsubsections, a word like or a is and not compared with variations allowing the extraction of capitalized unless it is the first word of the header knowledge in the creation of personas. Subsubsections
For this purpose, this paper presents the application context The heading for subsubsections should be in Helvetica 9that motivated the research, followed by the method used to point italic with initial letters capitalized. document user profiles. Next section presents the complete FIGURES process used to identify behavioral personas, and its results. Figures should inserted at the appropriate point in your CONTEXT ANDbeMOTIVATION text. Figures may extend over the of twotechnology columns up 17.8 In recent decades, the presence hasto been cm (7") if necessary. Black and white photographs (not common in almost all transactions. The same has occurred Polaroid prints)service may beconducted mounted by on public the camera-ready in the citizen enterprises, paper with double-sidedthe tape. A servicebetween bureau can which haveglue beenorstrengthening relationship the make a special print of your black and white photography government and people through technology. Nevertheless, for inclusion printing purposes (optional). To avoidtosmudges, the of technology in responding citizen's attach needs figures by paste or tape applied to their back surfaces only. takes into regard the high degree of diversification from the Each figure should have a figure caption in Times Roman. target audience, given that the attendance of public services is conducted the whole Color figures for should appear population, on separate without pages sodistinction that they between citizens. into a color section in the proceedings. may be collected Colordevelopment figures are ofa projects large expense the conference. The for thisforaudience requires Include them only if they are absolutely necessary – and knowledge of the user observed from field instruments, only if your submission category permits them. data collection and analysis. Filgueiras et al. [6] developed aLANGUAGE, study on theSTYLE profileAND of users from e-Poupatempo, in Sao CONTENT Paulo. E-Poupatempo is a public placeof for CHI the realization of The written and spoken language is English. services provided by Poupatempo through computers. Spelling and punctuation may consistently use any dialect According to theBritish, cited Canadian study, e-Poupatempo attendants of English (e.g., or US). Hyphenation is helped at Please that time, 30 people per day, which optional. writeonforaverage, an international audience: has justified the need to improve usability of e-gov _ Write in a straightforward style. Use simple sentence websites and thus reduce the problems encountered by the structure. Try to avoid long sentences and complex citizens. The study was based on the personas method for sentence structures. Use semicolons carefully. modeling user profiles, developing user archetypes which _ Usetocommon and basic vocabulary (e.g., use the proved be representative of a large public likeword the “unusual” rather Poupatempo users.than the word “arcane”).
_ Briefly orof explain all technical terms.10 personas Through thedefine process audience segmentation, were determined to represent the users of e-Poupatempo _ Explain all acronyms the first time they are used in [6]. Each of the personas was described in your text – e.g., “World Wide Web (WWW)”.some detail relevant to the usability project, providing good knowledge about the users. The main advantage of using personas is the possibility of designing the system for more than one user profile, which makes possible the realization of a project focused on users considering fictional, although representative, portrayals of the citizen. In that study, fictional portrayals of citizens presented demographical information obtained from field research and behavioral data inferred from the observant experience.
paper,local we present the method of data mining to build _In this Explain references (e.g., not everyone knows all personas through clusters country). of behavioral information and city names in a particular consistency checked with demographic information and _ Explain “insider” comments. Ensure that your whole expectations. The objective is to document the process of audience understands any reference whose meaning you building personas based on real data. do not describe (e.g., do not assume that everyone has PERSONAS used a Macintosh or a particular application). The first formal description of the method of personas in _HCI Explain colloquial language puns. occurred in 1995, when and there wasUnderstanding a need to phrases like “red herring” requires a cultural communicate different perspectives of users on aknowledge consulting of English. Humorby andAlan ironyCooper are difficult to translate. project conducted at the technological [3, 4,unambiguous 5]. Cooper defined "personaslocalized are not _classUse forms personas: for culturally real people ... are as hypothetical archetypes of actual users ... concepts, such times, dates, currencies and numbers defined with significant rigor may and precision" [3]. or 1 May (e.g., “1-597” or “5/1/97” mean 5 January , and “seven may developed mean 7:00 am 19:00).product Personas is o'clock” a method to orassist development focused on the desires and of _ Be careful with the use of gender-specific needs pronouns customers. the methodology of User-Centered (he, she)Used andinside other gendered words (chairman, Design (UCD), personas were considered initially(e.g., as manpower, man-months). Use inclusive language stereotypes the they, result chair, of a preconceived opinion and she or he,ors/he, staff, staff-hours, persondisseminated the members of society. Nowadays, years) that isamong gender-neutral. If necessary , you may be they as archetypes that represent real beings ablearetounderstood use “he” and “she” in alternating sentences, so [10]. that the two genders occur equally often. See [5] for further advice and examples regarding and other Personas are thus compositions of gender real data and personal attributes. representative characteristics that define and detail fictitious users. Its description can be composed of ACKNOWLEDGMENTS information envisioned by demographic personality We thank CHI, PDC, CSCW volunteers, and all publications characteristics fictitious support and and staffbiographical who wrote and beings. provided helpful comments on previous versions of this document. Although fictional, personas represent a specific user and support to decision-making becomes more driven to real REFERENCES customer needsR.E. [3, 4]. 1. Anderson, Social impacts of computing: Codes of professional ethics.developers, Social Science Review Technology product such Computing as Web designers, (Winter 1992), 453-469. use10, the2concepts of audience segmentation and are not so successful because, unlike personas,Format. they do Available not take into 2. CHI Conference Publications at consideration the behavior, perception and context in which http://www.acm.org/sigchi/chipubform/. interaction occurs. Thus, it is possible to say that the 3. Conger., S., and Loch, (eds.). Ethics computer method of personas is a K.D. qualitative study and assistance in use. Commun. ACM 38, 12 (entire issue). understanding the needs and objectives of clients [10]. 4. Mackay,a personas W.E. Ethics, lies andthevideotape, in Therefore, method addresses representation Proceedings of CHI '95 (Denver CO, May 1995), ACM needs of diverse customer profiles leading into the most Press, 138-145. influential factors of human behavior.
5. and[10], Taskthe Force onfictional Bias-Free Language. ForSchwartz, Pruitt andM., Adlin use of characters in Guidelines for Bias-Free Writing. Indiana University the development of audience and segmentation can serve a Press, in Bloomington IN,ways: 1995.it increases the usability of company the following product and its usefulness; it improves the streamlining processes of team work; it improves a company's ability to attend customer needs while serving the company; and it increases profits. DATA CLUSTERING
Clustering is a procedure for the division or merging The columns on the last pageData of equal ofshould data collectionsbe that follow recognized length. standards such as clusters that are mutually exclusive. Thus, data clustering identifies data with the same pattern into a cluster. The partitions show distinct data patterns [7]. Basically, in the clustering process, groups are formed in order to minimize the distance between elements of a group and maximizing the distance between clusters. 92
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Page Clustering Numbering, and Footers Data is Headers applied in this paper to group relevant Do not include headers, footers or page numbers in your features characterizing the user profile based on: submission. These be added when the before publications are demographic data,will reported expectation use and assembled. satisfaction after completion of interactive subjective activities in e-government. SECTIONS
when these classes (e.g., are sparsely in space (project _algorithm Explain local references not everyone knows all data). city names in a particular country).
_
Explain “insider” comments. Ensure that your whole audience understands any reference whose meaning you do not describe (e.g., do not assume that everyone has used a Macintosh or a particular application).
_
Explain colloquial language and puns. Understanding phrases like “red herring” requires a cultural knowledge of English. Humor and irony are difficult to translate.
_
Subsubsections K-Means algorithm performs the following steps to a The heading forofsubsubsections be in Helvetica 9particular group non-clustered should data [7],[11]: point italic with initial letters capitalized. 1. Select n objects to be the centers of clusters;
Use unambiguous forms for culturally localized concepts, such as times, dates, currencies and numbers (e.g., “1-5- 97” or “5/1/97” may mean 5 January or 1 May , and “seven o'clock” may mean 7:00 am or 19:00).
_
2. Repeat stepsbe 3 and 4 while there are changes in cluster; Figures should inserted at the appropriate point in your text.Reorganize Figures may theFOXVWHUWKDWLW¶VFORVHst two columns up to 17.8 3. eachextend objectover to the to. cm (7") if necessary. Black and white photographs (not 4. Recalculate the centers of clusters. Polaroid prints) may be mounted on the camera-ready paper with glue or double-sided tape. A service bureau can make a special print your black white photography Equation 1 shows theofformula usedand by K-Means to define for printing avoid smudges, attach the distance purposes between (optional). objects andTothe center of a cluster. figureswe byhave pastea or tape applied theirformula back surfaces When smaller result oftothis it meansonly. that Eachobject figure isshould have figure caption Times Roman.of the closer to athe center. Thein minimization Equation 1 means thatappear K-Means ends it execution. Color figures should on separate pages so [7]. that they may be collected into a color section in the proceedings. Color figures are a large expense for the conference. Include them only if they are absolutely necessary – and only if your submission category permits them.
Be careful with the use of gender-specific pronouns (he, she) and other gendered words (chairman, manpower, man-months). Use inclusive language (e.g., she or he, s/he, they, chair, staff, staff-hours, personyears) that is 1gender-neutral. If necessary you may be Figure ± Data clustering with close, data able to use “he” and “she” in alternating sentences, so that the two genders occur equally often. See [5] for further advice and examples regarding gender and other personal attributes.
ACKNOWLEDGMENTS
K-Means The heading of a section should be in Helvetica 9-point
There a lot of research overshould grouping data with similar or bold inisall-capitals. Sections be unnumbered. related information, i.e. clustering [2], [8]. Among Subsections algorithms used with the aim of clustering, K-Means is The heading of subsections should be in Helvetica 9-point currently one of the most used due to its performance and bold with only the initial letters capitalized. (Note: For subease of use [12]. K-means algorithm is used to create sections and subsubsections, a word like the or a is not clusters with different data types like images, texts, and capitalized unless it is the first word of the header other kinds of data [12].
FIGURES
LANGUAGE, STYLE AND CONTENT
The written and spoken language of CHI is English. Equation 1 - Minimum Square K-Means Spelling and punctuation may consistently use any dialect of English (e.g., British, Canadian or US). Hyphenation is optional. for an international audience: There arePlease somewrite inconveniences to this algorithm. To minimize those errors in an optimal manner, eithersentence to 2 or _ Write in a straightforward style. Use simple Nstructure. clusters, itTry takes a long time, because this algorithm is to avoid long sentences and complex NP-hard, because you can not verify it or perform it in sentence structures. Use semicolons carefully. polynomial time, and also this is a greedy algorithm that _ Use common and the word only converges to basic local vocabulary minimums,(e.g., anduseonly this “unusual” rather than the word “arcane”). convergence to a global minimum if clusters are well _ Briefly define or explain all technical terms. separated [7]. Due to this problem K-Means directly on the _ Explain all acronyms the isfirst time dependent they are used in startup of the centers of clusters [13].(WWW)”. This case implies the your text – e.g., “World Wide Web standard deviation is small and the initialization of the algorithm is sparse, finding a cluster in this data can become a very difficult task.
We thank CHI, PDC, CSCW volunteers, all publications support and staff who wrote and provided helpful comments on previous versions of this document. REFERENCES
1. Anderson, R.E. Social impacts of computing: Codes of professional ethics. Social Science Computing Review 10, 2 (Winter 1992), 453-469. 2. CHI Conference Publications Format. Available at http://www.acm.org/sigchi/chipubform/. 3. Conger., S., and Loch, K.D. (eds.). Ethics and computer use. Commun. 12 (entire issue). Figure 2 ±ACM Data38, clustering with sparse data 4. Mackay, W.E. Ethics, lies and videotape, in Proceedings of CHI '95 (Denver CO, May 1995), ACM Even with138-145. these disadvantages K-Means is widely used in Press, scientific research and commercial applications in data 5. Schwartz, M., and Task Force on algorithm Bias-Free to Language. mining, even with the difficulty of the find the Guidelines for Bias-Free Writing. Indiana University best solution, it can find a solution that satisfies the Press,inBloomington 1995. problem considerableIN, time and it's easily applicable [11]. PROCEDURE FOR CREATING PERSONAS
To accomplish this paper, it was necessary a data analysis from user profiles obtained through questionnaires before and after services to e-Government. Figure 3 shows the steps necessary for the implementation of this proccess.
section presents preparationlength. for the process, The columns on the last pageNext should be theofdataequal where data will be filtered and normalized. Following the
Figure 1 shows the formation of clusters with close data (project data), the left identifies the classes in a twodimensional space and beside the result of the partitions created by K-Means, where it realizes that the algorithm does not perform the partitions isolating each class in its own cluster. In Figure 2 we can observe the behavior of the
clustering process will be detailed, from the implementation of K-Means to the formation of the personas, which represent their clusters. Finally, the analysis of the personas established is discussed showing the multidirectional part of this process.
93
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Page Numbering, Headers and Footers
Do not include headers, footers or page numbers in your submission. These will be added when the publications are assembled.
_
Explain local references (e.g., not everyone knows all city names in a particular country).
_
Explain “insider” comments. Ensure that your whole audience understands any reference whose meaning you do not describe (e.g., do not assume that everyone has used a Macintosh or a particular application).
_
Explain colloquial language and puns. Understanding phrases like “red herring” requires a cultural knowledge of English. Humor and irony are difficult to translate.
_
Use unambiguous forms for culturally localized concepts, such as times, dates, currencies and numbers (e.g., “1-5- 97” or “5/1/97” may mean 5 January or 1 May , and “seven o'clock” may mean 7:00 am or 19:00).
_
Be careful with the use of gender-specific pronouns (he, she) and other gendered words (chairman, manpower, man-months). Use inclusive language (e.g., she or he, s/he, they, chair, staff, staff-hours, personyears) that is gender-neutral. If necessary , you may be able to use “he” and “she” in alternating sentences, so that the two genders occur equally often. See [5] for further advice and examples regarding gender and other personal attributes.
SECTIONS
The heading of a section should be in Helvetica 9-point bold in all-capitals. Sections should be unnumbered. Subsections
The heading of subsections should be in Helvetica 9-point bold with only the initial letters capitalized. (Note: For subsections and subsubsections, a word like the or a is not capitalized unless it is the first word of the header Subsubsections
The heading for subsubsections should be in Helvetica 9point italic with initial letters capitalized. FIGURES
Figures should be inserted at the appropriate point in your text. Figures may extend over the two columns up to 17.8 cm (7") if necessary. Black and white photographs (not Polaroid prints) may be mounted on the camera-ready paper with glue or double-sided tape. A service bureau can make a special print of your black and white photography for printing purposes (optional). To avoid smudges, attach figures by paste or tape applied to their back surfaces only. Each figure should have a figure caption in Times Roman. Color figures should appear on separate pages so that they may be collected into a color section in the proceedings. Color figures are a large expense for the conference. Include them only if they are absolutely necessary – and only if your submission category permits them. LANGUAGE, STYLE AND CONTENT
The written and spoken language of CHI is English. Spelling and punctuation may consistently use any dialect of English (e.g., British, Canadian or US). Hyphenation is optional. Please write for an international audience:
ACKNOWLEDGMENTS
We thank CHI, PDC, CSCW volunteers, all publications support and staff who wrote and provided helpful comments on previous versions of this document. REFERENCES
1. Anderson, R.E. Social impacts of computing: Codes of professional ethics. Social Science Computing Review 10, 2 (Winter 1992), 453-469. 2. CHI Conference Publications Format. Available at http://www.acm.org/sigchi/chipubform/. 3. Conger., S., and Loch, K.D. (eds.). Ethics and computer use. Commun. ACM 38, 12 (entire issue).
_
Write in a straightforward style. Use simple sentence structure. Try to avoid long sentences and complex sentence structures. Use semicolons carefully.
4. Mackay, W.E. Ethics, lies and videotape, in Proceedings of CHI '95 (Denver CO, May 1995), ACM Press, 138-145.
_
Use common and basic vocabulary (e.g., use the word “unusual” rather than the word “arcane”).
5. Schwartz, M., and Task Force on Bias-Free Language. Guidelines for Bias-Free Writing. Indiana University Press, Bloomington IN, 1995.
_ _
Briefly define or explain all technical terms. Explain all acronyms the first time they are used in your text – e.g., “World Wide Web (WWW)”.
The columns on the last page should be of equal length. Figure 3 ± Procedure for creating personas
94
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Preparing DatabaseHeaders and Footers Page Numbering,
For thisinclude research, information waspage obtained from two Do not headers, footers or numbers in your questionnaires: a questionnaire evaluate the expectation submission. These will be addedtowhen the publications are of citizens about the computer use on the web to reach their assembled. goal of electronic services ³SUH-VHUYLFH TXHVWLRQQDLUH´ SECTIONS and a questionnaire applied after use to collect user The heading of a section should be in Helvetica 9-point satisfaction according to experience ³Sost-service bold in all-capitals. Sections should be unnumbered. TXHVWLRQQDLUH´ Subsections
A large database of exactly 6580 records was collected, The heading of subsections should be in Helvetica 9-point being 3812 pre-service questionnaires and 2768 postbold with only the initial letters capitalized. (Note: For subservice questionnaires. Pre-service questionnaires enquire sections and subsubsections, a word like the or a is not users about their expectations to perform the services, while capitalized unless it is the first word of the header post-service questionnaires enquire about the ease of use of Subsubsections these services and user´s impression regarding the past The heading for subsubsections should be in Helvetica 9experience. pointfirst italicstep withisinitial lettersofcapitalized. The a quality data process, in which user input data is verifyed against invalid information, for FIGURES example, age equal to zeroat(very common when user Figures should be inserted the appropriate pointthe in your does not wish to extend reveal over his/her In these up cases the text. Figures may theage). two columns to 17.8 invalid incompleteBlack recordandwas removed from (not the cm (7") or if necessary. white photographs database. Polaroid prints) may be mounted on the camera-ready paper with glue double-sidedwas tape.separated A serviceinto bureau can In sequence, theor information answer make a special print of your black and white photography records to post-service questionnaire and pre-service for printing purposes (optional). To avoid attacha questionnaire, because people who smudges, answered figures by paste or tape applied to their back surfaces only. questionnaire did not answer the other one and the number Each figure should have a figure caption in Times Roman. of questions between them is also different.
Color questionnaire figures should models appear on separate pagesdemographic so that they Both have identical may be collected into a color section in the proceedings. questions with the aim to allow the junction with the other Color figuresBecause are a of large expense the conference. information. impact on thefor service completion Include them only if they are absolutely necessary and time, only one questionnaire model was applied to– each only if your submission category permits them. citizen. Another purpose of the division was to reduce the time of participation research, increasing the receptivity LANGUAGE, STYLE in AND CONTENT of citizen. and For the pre-service questionnaire were Thethewritten spoken language of CHI isthere English. 14 questions, while for may the post-service thereanywere 28 Spelling and punctuation consistently use dialect questions, below. of Englishpresented (e.g., British, Canadian or US). Hyphenation is optional. Please an international audience: Questions made write in thefor pre-service questionnaire: Write in a straightforward style. Use simple sentence x_ Age structure. Try to avoid long sentences and complex x sentence Educational level Use semicolons carefully. structures. x_ Profession Use common and basic vocabulary (e.g., use the word than the word “arcane”). x “unusual” Monthlyrather income x_
Briefly define explain Occupation fororthose whoall dotechnical not haveterms. an income
x_
Explaintechnology all acronyms the experienced first time they are used in Digital already your text – e.g., “World Wide Web (WWW)”. x Experience using computers x x x
Frequency of Internet use Place where Internet access happens
_ Explain local references (e.g., not everyone knows all city names in ainparticular country).questionnaire: Questions made the post-service _x Explain “insider” comments. Ensure that your whole Age audience understands any reference whose meaning you x do Educational not describeLevel (e.g., do not assume that everyone has a Macintosh or a particular application). x used Profession / Occupation _x
Explain Monthly colloquial Income language and puns. Understanding phrases like “red herring” requires a cultural knowledge x Justification for who does not have an income of English. Humor and irony are difficult to translate. x Which equipment did you use? _ Use unambiguous forms for culturally localized x concepts, Experience computers suchusing as times, dates, currencies and numbers (e.g., “1-597” or “5/1/97” mean 5 January or 1 May x How often do you use themay Internet? , and “seven o'clock” may mean 7:00 am or 19:00). x Where do you usually access the Internet? _ Be careful with the use of gender-specific pronouns x (he,Howshe) muchand help other did the caregiver gendered provide? words (chairman, manpower, man-months). Use inclusive language (e.g., x Will you use this service again? she or he, s/he, they, chair, staff, staff-hours, personx Was it easy to perform the service? years) that is gender-neutral. If necessary , you may be x able Didtoituse take“he” long and to perform thealternating service? sentences, so “she” in that the two genders occur equally often. [5] for x Were you satisfied when you performed theSee service? further advice and examples regarding gender and other x personal Were you nervous using the service? attributes. x Did you feel safe when using the Internet? ACKNOWLEDGMENTS We thankyou CHI, CSCW volunteers, all publications x Did findPDC, the service page? support and staff who wrote and provided helpful x Did you the scroll bar? of this document. comments onuse previous versions x Did you read text on the screen? REFERENCES
x Anderson, Did you understand theimpacts text? of computing: Codes of 1. R.E. Social professional ethics. Social x Did you click on buttons orScience icons? Computing Review 10, 2 (Winter 1992), 453-469. x Did you print a document? 2. CHI Conference Publications Format. Available at x http://www.acm.org/sigchi/chipubform/. Did you use the mouse? x Conger., Did you S., useand the Loch, keyboard? 3. K.D. (eds.). Ethics and computer use. Commun. ACM 38, 12 (entire issue). x Did you fill in any boxes?
4. W.E. Ethics,links? lies and videotape, in x Mackay, Did you use navigation Proceedings of CHI '95 (Denver CO, May 1995), ACM x Press, Did you use a search engine? 138-145. x Schwartz, Did you complete the room? Language. 5. M., and your Task purpose Force onin Bias-Free Guidelines for Bias-Free Writing. Indiana University In this step a de-normalization of the original database was Press, Bloomington IN, 1995. performed, i.e., the answers of each questionnaire were grouped in a database table for easy browsing and the manipulation of information. Based on the two tables created, variables were grouped to the formation of eight data groups. Table 1 gives a description of each group and the motivation that justified the need for this type of division.
The columns ononthe Time the user expects to spend service last page should be of equal length.
x
Do you expect use to the computer? (5-point scale)
x
Will it be easy to realize the service?
x
Will it be tricky to use the Internet?
x
Do you think it will be safe to use the Internet?
#Group
Description
Motivation
A
All variables from preservice questionnaire
To find personas composed by demographic and expectation data.
95
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Page Numbering, Headers and Footers To find personas All headers, variables from Do not include footers or pagecomposed numbersbyin your B post-service demographic, service are submission. These will be added when the publications questionnaire satisfaction and ease-ofassembled. use data SECTIONS Demographic variables demographic The Cheading from of apre section should beToinfind Helvetica 9-point and postpersonas bold in all-capitals. Sections should be unnumbered. service questionnaire Subsections Demographic variables To find demographic TheDheading of subsections from pre-serviceshould be in Helvetica 9-point personas questionnaire bold with only the initial letters capitalized. (Note: For sub-
sections and subsubsections, a word like the or a is not Demographic variables To find demographic capitalized unless it is the first E from post-service word of the header personas Subsubsectionsquestionnaire of service should To be findinexpectation The FheadingVariables for subsubsections Helvetica 9point italic with expectation initial letters capitalized. personas FIGURES G
Variables of service satisfaction
To find satisfaction personas
Figures should be inserted at the appropriate point in your of declared Tocolumns find ease-ofuse17.8 text.HFiguresVariables may extend over the two up to easy use personas cm (7") if necessary. Black and white photographs (not Polaroid be mounted on the camera-ready Table 1 ±prints) Groupmay variables to create typification personas paper with glue or double-sided tape. A service bureau can make a special print of your black and white photography Groups A and B were created To in an attempt to achieve for printing purposes (optional). avoid smudges, attach personas relatively largetogroup of variables figures byfrom pastea or tape applied their back surfaceswhich only. represent behavior andRoman. / or Each figure different should havekinds a figureofcaption in Times characteristics. In groups C, D and E only variables related Color figures should appear on separate pages so that they to the demographic data from users were separated, and as may be collected into a color section in the proceedings. this data was common to both questionnaires, a Color figures are a large expense for the conference. combination of the two sources was performed in an Include them only if they are absolutely necessary – and attempt to achieve better detail for generated personas. only if your submission category permits them. The others groups - F, G and H - were separated by specific LANGUAGE, STYLE AND CONTENT data from each of the questionnaires, representing the The written and spoken language of CHI is English. service expectation information, satisfaction with the Spelling and punctuation may consistently use any dialect service and ease of use offered by the service, respectively. of English (e.g., British, Canadian or US). Hyphenation is Except groups A for andanBinternational which used audience: all the variables, optional.for Please write two issues were not considered for this study, because they _ Write in a straightforward Use of simple sentence were multivalued and during thestyle. execution the algorithm structure. Try to avoid long sentences and complex could generate a range from little interest to the results of sentence structures. Use semicolons carefully. the research as discussed in the next topics. The questions excluded were: and basic vocabulary (e.g., use the word _ Use common “unusual” rather thandid theyou word “arcane”). x Which equipment use? Briefly define or explain technical terms. x_ How often do you use theall Internet? _ Explain acronyms the infirst they aretoused in These issues all were not used thistime procedure create your textbut – e.g., “World WideinWeb (WWW)”. clusters, were useful other models of data
transformation, also used to characterize the persona after identifying the groups.
applying clustering algorithm chosenknows for this _Before Explain localthe references (e.g., not everyone all research normalization datacountry). procedure was performed. The city names in a particular purpose of this procedure is to establish a standard between _ Explain “insider” comments. Ensure that your whole these values and stay inside the same range. This action audience understands any reference whose meaning you supports the algorithm to obtain a good result when the data do not describe (e.g., do not assume that everyone has has many variables. used a Macintosh or a particular application). This normalization study was done between 0 and 1, as it is _the most Explain colloquial and puns. Understanding common way tolanguage obtain standardization [9]. phrases like “red herring” requires a cultural knowledge Two steps were necessary to perform the normalization of English. Humor and irony are difficult to translate. of data: the first one was code the answers to numbers, Table _2 shows Use an unambiguous forperformed; culturally thelocalized example of forms encoding second concepts, such as times, dates, currencies numbers step was the application of Equation 2, where and the new data (e.g., “1-5- by 97”dividing or “5/1/97” may mean 5 January or 1 May was reached the current given by the maximum andvariable “seven o'clock” mean 7:00 am or 19:00). of, the which it may represents. _ Be careful with the use of gender-specific pronouns (he, she) and other gendered words (chairman, Before After(e.g., manpower, man-months). UseAfter inclusive language coding coding normalization she or he, s/he, they, chair, staff, staff-hours, personNDA years) that is gender-neutral. If 0necessary , you0.0 may be ableUp_to_R$_509.00 to use “he” and “she” in alternating sentences, so 1 0.2 that the two genders occur equally often. See [5] for From_R$_510.00_to_R$_ 2 0.4 further advice 1529.00and examples regarding gender and other personal attributes. From_R$_1530.00_to_R$ ACKNOWLEDGMENTS _2549.00
3
0.6
We Above_R$_2550.00 thank CHI, PDC, CSCW volunteers, all publications 4 0.8 support and staff who wrote and provided helpful No income 5 comments on previous versions of this document. 1.0 Table 2 - Data transformation for normalization REFERENCES 1. Anderson, R.E. Social impacts of computing: Codes of professional ethics. Social Science Computing Review 10, 2 (Winter 1992), 453-469. 2. CHI Conference Publications Format. Available at Equation 2 - Data normalization calculus http://www.acm.org/sigchi/chipubform/. 3. Conger., S., and Loch, K.D. (eds.). Ethics and computer Commun. 12 (entire issue). Theuse. entire processACM was 38, implemented using MATLAB due to its ease of W.E. use, trustworthiness reuse. Thus 4. Mackay, Ethics, lies and andcode videotape, in those eight groups uploaded the1995), MATLAB Proceedings of CHIwere '95 (Denver CO,toMay ACM workspace in matrix shapes to better manipulate them. Press, 138-145. Following the M., implementation of the normalization to 5. Schwartz, and Task Force on data Bias-Free Language. eachGuidelines group, thefor identification phase was performed to the Bias-Free Writing. Indiana University clusters which each IN, sample Press,toBloomington 1995.belonged. To this end, we applied the algorithm K-Means native in the MATLAB library which was defined as parameters, groups A to H to perform an create clusters and the number of cluster that is expected to generate from each group, in this case 10. The number of 10 clusters follows the pattern of work [1] in the same context of the electronic government.
From this point the eight groups are ready to be implemented into the remaining procedure.
identifying be the clusters, the relevance of each cluster The columns on the last pageAfter should of equal length. is calculated in respect to group used. This calculation is
Clustering procedure
given based on Equation 3 below:
In the procedure to create clusters, we attempted to identify natural patterns among data, aiming to insert them into the same group to identify similar characteristics among the information belonging to the group.
96
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Page Numbering, Headers and Footers
Do not include headers, footers or page numbers in your submission. These will be added when the publications are assembled. Equation 3 - Cluster relevance calculus SECTIONS
The heading of a section should be in Helvetica 9-point Where: bold in all-capitals. Sections should be unnumbered. xSubsections R is relevance The heading of subsections x E is number of elements should be in Helvetica 9-point bold with only the initial letters capitalized. (Note: For subxsections k is number of the cluster a word like the or a is not and subsubsections, capitalized unless it isthe theclusters first word the header From this moment, areofdefined, and we also know the relevance of each to their origin group, but can Subsubsections not identify which persona that characterizes each cluster. The heading for subsubsections should be in Helvetica 9-
pointidentify italic with initialpersonas letters capitalized. To these we used three arithmetic methods: mean given by Equation 4, median represented by FIGURES Equation 5 and mode that presents datapoint of inhigher Figures should be inserted at the appropriate your occurrence within the variable. text. Figures may extend over the two columns up to 17.8 cm (7") if necessary. Black and white photographs (not Polaroid prints) may be mounted on the camera-ready paper with glue or double-sided tape. A service bureau can make a special print of your black and white photography for printing purposes (optional). To avoid smudges, attach figures by paste or tape applied to their back surfaces only. Equation ± Mean calculus Each figure should have a4figure caption in Times Roman. Color figures should appear on separate pages so that they may be collected into a color section in the proceedings. Color figures are a large expense for the conference. Include them only if they are absolutely necessary – and only if your submission permits them. Equationcategory 5 ± Median calculus LANGUAGE, STYLE AND CONTENT
The written and spoken language of CHI is English. Through thesepunctuation methods 30may personas were obtained, 10 for Spelling and consistently use any dialect each method above, using the information of the 8 groups of English (e.g., British, Canadian or US). Hyphenation is of data composed in the data preparation, totalizing 240 optional. Please write for an international audience: personas generated by four different types (demographic, _ Write inofause, straightforward style. Use simple sentence expectation ease of use and satisfaction of use). structure. Try to avoid long sentences and complex Analyzing Personas sentence structures. Use carefully. After the identification of semicolons personas, personas were selected from datacommon group C,and which demographic both _ Use basichas vocabulary (e.g.,data usefrom the word questionnaires to analyze a higher level of detail. “unusual” rather than theonword “arcane”).
During the detailed analysis personas created by each _ Briefly define or explainofallthe technical terms. method it was noted that some of the personas generated _ Explain all acronyms the first time they are used in actually belonged to another cluster, due to their similarity your text – e.g., “World Wide Web (WWW)”. or difference of tiny data. Thus, we eliminated duplication and increased the reference cluster values for every one which owned such similarity.
_x
For mode, instead of a (e.g., total of personas ratified Explain local references not10 everyone knows all 9 personas. citywith names in a particular country).
point “insider” a search comments. in the database wasthat carried for _At this Explain Ensure your out whole each of theunderstands personas formed by checking themeaning existenceyou of audience any reference whose those, anddescribe at the same some that of the personashas of do not (e.g.,time do finding not assume everyone user ease of use, which usedsatisfaction, a Macintosh expectation or a particularand application). corresponded. _ Explain colloquial language and puns. Understanding Table 3 presents how many instances each persona were phrases like “red herring” requires aofcultural knowledge found in the databases. of English. Humor and irony are difficult to translate.
_
Use unambiguous forms for culturally localized concepts, such as times, dates, currencies and numbers Median (e.g., “1-5- 97” or “5/1/97” may mean 5 January or 1 May Persona may Pre Post , and “seven o'clock” mean 7:00 Total am or 19:00). 1 50 39 89 _ Be careful with the use of gender-specific pronouns 0 0 (he, she) and 2 other 0gendered words (chairman, manpower, man-months). Use inclusive language (e.g., 3 0 0 0 she or he, s/he, they, chair, staff, staff-hours, person4 0 1 , you may be years) that is gender-neutral. If1necessary able to use “he”5 and “she” 6 in 4alternating 10 sentences, so that the two genders occur equally often. See [5] for 6 0 5 5 further advice and examples regarding gender and other 7 21 25 46 personal attributes. 8 50 58 108 ACKNOWLEDGMENTS
We thank CHI, PDC, CSCW volunteers, all publications support and staff Meanwho wrote and provided helpful comments on previous versions of this document. Persona Pre Post Total REFERENCES
1 0 of computing: 1 1. Anderson, R.E.1Social impacts Codes of 2 0 Science 0 0 professional ethics. Social Computing Review 10, 2 (Winter 1992), 453-469. 3 18 21 39 2. CHI Conference Publications Format. Available at 4 0 0 0 http://www.acm.org/sigchi/chipubform/. 5 0 0 0 3. Conger., S., and Loch, K.D. (eds.). Ethics and computer 6 0 0 0 use. Commun. ACM 38, 12 (entire issue). 10 lies 4 and 14 videotape, in 4. Mackay, W.E.7 Ethics, 1 CO, 1May 1995), ACM Proceedings of 8CHI '95 0(Denver Press, 138-145.
5. Schwartz, M., and Task Force on Bias-Free Language. Mode Guidelines for Bias-Free Writing. Indiana University Persona IN,Pre Press, Bloomington 1995.Post Total 1 12 4 16 2
6
2
8
3
1
0
1
4
17
10
27
5
4
7
11
7
34
40
74
8
5
6
11
The columns on the last page should be of50equal 6 40 90length.
In the end of the analysis of each one of the three methods (mean, median and mode) the following results were obtained: x
For median, instead of a total of 10 personas ended with a score of 8 personas.
x
For mean, instead of a total of 10 personas 8 personas were identified.
9 60 30 90 Table 3 - Count the number of occurrences per persona
97
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Page Numbering, Headers and Footers
Do not include headers, numbers in your The information in Table 3footers showsorthepage existence of personas submission. These will be added when the publications are with great representation in the database for both the assembled. questionnaires. SECTIONS for the cases of personas generated by the Furthermore, The heading of aand section should in Helvetica 9-point methods, mean median, are be personas composed of bold in all-capitals. Sections should be unnumbered. information from the database, however there is no similar combination Subsectionsof those stored in the database. The identify heading the of subsections should be in Helvetica 9-point To cases of complex personas just identify bold with only the initial letters capitalized. (Note: For those personas which have no occurrence, for cases subthat sections and subsubsections, a word like the than or azero, is not have obtained a result in the selection greater it capitalized unless it isa the first word ofpersona, the header was possible to find corresponding created from the data of the groups F, G and H, because these presented Subsubsections the combination in some records database. 9Thesame heading for subsubsections shouldfrom be the in Helvetica point italic with initial letters capitalized. RESULTS When clustering was performed with data of the groups A FIGURES and B were at notthesatisfactory. Figures those shouldresults be inserted appropriateThe pointK-means in your algorithm did not achieve a good behavior while text. Figures may extend over the two columns upworking to 17.8 with too ifmany variables - 41and from the database of postcm (7") necessary. Black white photographs (not questionnaires andbe27mounted from theondatabase of before Polaroid prints)usemay the camera-ready questionnaires - and personas generated in 80% of can the paper with glueuse or double-sided tape. A service bureau cases are equal to each other. This occurred because the make a special print of your black and white photography greater variation of (optional). data was To between answers of for printing purposes avoid the smudges, attach multivalued issues, which affected growth of the other figures by paste or tape applied to their back surfaces only. values. Each figure should have a figure caption in Times Roman.
As for figures those groups G andonH,separate creatingpages clusters Color shouldF,appear so achieved that they amay satisfactory value because there was good diversity be collected into a color section in the proceedings. among those generated personas. However, clusters Color figures are a large expense for thethese conference. when theifalgorithm were very intermittent, Includegenerating them only they are absolutely necessary – due and to theif standard deviation between this data only your submission category permits them.lies very low making those clusters often stay empty providing an error LANGUAGE, STYLE AND CONTENT in the algorithm K-Means. The written and spoken language of CHI is English. Between groups C, D andmay E, clustering wasusesuccessful as Spelling and punctuation consistently any dialect expected and very similar among the 3 groups, but it of English (e.g., British, Canadian or US). Hyphenation is is clear thatPlease when write usingfora an larger amount of data the result optional. international audience: presented is better. _ Write in a straightforward style. Use simple sentence With this fact just personas generated from the group C structure. Try to avoid long sentences and complex have been chosen to perform analysis in more detail, as it sentence structures. Use semicolons carefully. contained data from both questionnaires. _ Use common and basic vocabulary (e.g., use the word Verifying those numbers of occurrences of each persona in “unusual” rather than the word “arcane”). the original database and identifying that the mode was the _ Briefly explainoccurrences all technicalfrom terms.each of the only methoddefine that or showed personas directly in the database. _ Explain all acronyms the first time they are used in
your text or – e.g., “World Webchange (WWW)”. The mean median mayWide slightly the values and end up in values which originally there is no occurrence in the database, unlike the mode which only returns the highest frequency value in the data. Therefore execution of the method for mode can achieve the most return queries in the database.
methods support the ratification procedure of _otherExplain local references (e.g., not aiming everyone knows all a more representative persona. city names in a particular country). interesting result is that Ensure in mostthat cases forwhole each _Another Explain “insider” comments. your persona of understands group C reached in the database, it was also audience any reference whose meaning you found as most values, that personas of user do notsuch describe (e.g.,frequent do not assume everyone has satisfaction, expectation and ease application). of use - belong to group used a Macintosh or a particular G, F and H respectively - match, see Table 4. _ Explain colloquial language and puns. Understanding Asphrases a last step of procedure, it was noted that there was a like “red herring” requires a cultural knowledge correspondence between personas through the opposite of English. Humor and irony are difficult to translate. path, starting from the personas of user satisfaction, _expectation Use unambiguous forms for culturally localized and ease of use to demographic and the result concepts, such as times, dates, currencies and was positive for personas to occurrences in the numbers database (e.g., “1-597”3.or “5/1/97” may mean 5 January or 1 May presented Table , and “seven o'clock” may mean 7:00 am or 19:00). In general, the highest combination of occurrences found, _existed Be careful the useconsidered of gender-specific pronouns betweenwith personas appropriate after (he, she) and other words (chairman, clustering and analysis in anygendered of the paths opted to combine manpower, man-months). Use inclusive language (e.g., them. See example below in Table 4: she or he, s/he, they, chair, staff, staff-hours, personyears) that is gender-neutral. If necessary , you may be able to use “he”Demographic and “she” inPersona alternating sentences, so that the two genders occur equally often. See [5] for Variable Value further advice and examples regarding gender and other Age 30 personal attributes. Educational Level high school ACKNOWLEDGMENTS We thank CHI, PDC, CSCW volunteers, all publications from R$ 510,00 to Monthly Income support and staff who wrote and provided helpful R$ 1.529,00 comments on previous document. Without Income versions of thisnot answered Adjutant REFERENCES Occupation Assistant Codes of 1. Anderson, R.E. Social impacts of computing: professional ethics. Social Science Attendant Computing Review 10, 2 (Winter 1992), 453-469. I use the computer and have Computing Experience made courses 2. CHI Conference Publications Format. Available at Frequency of Internet use Frequently http://www.acm.org/sigchi/chipubform/.
3. Conger., S., and Loch, K.D. (eds.). Ethics and computer use. Commun. ACM 38, 12 (entire issue). Expectation Persona 4. Mackay, W.E. Ethics, lies andValue videotape, in Variable Proceedings of CHI '95 (Denver CO, May 1995), ACM Where do you use the Alone Press,computer? 138-145. 5. Complicated Schwartz, M., and the Task Force on Bias-Free Language. to use Guidelines for Bias-Free Writing. Disagree Indiana University internet? Press, IN, 1995. Do youBloomington feel safe on the Agree Internet? How long did it take to Up to 5 minutes use the service? Was it easy to use the Agree service?
The columns on the last page should be of equal length. Number of Demographic Occurrences from
Though we can observe which of those personas was generated from mean and median, which have high relevance, we also found through mode, thereby serving as a way of affirmation for these personas. In this case those
Expectation - 7 of 50 Number of Expectation Occurrences from Demographic - 13 of 212 Table 4 ± Comparing personas
98
Proceedings of IHC+CLIHC’2011 – Technical Session on HCI and the Workplace Pageresult Numbering, Headers and4Footers The presented in Table shows the possibility to Do notpersonas include headers, page numbers in your create typified footers from a ordatabase and make the submission. These will betoadded the publications are combination of personas attendwhen the needs of the designer. assembled. In the last two rows of Table 4 the number of times the
expectation SECTIONS persona appears is located when the demographic persona was should searched the database, for The heading of a section be in Helvetica 9-point example, searchingSections exactlyshould demographic persona data bold in all-capitals. be unnumbered. displayed in the table identifies 50 occurrences of the Subsections expectation bringing together these results the The heading personas, of subsections should be in Helvetica 9-point highest number of occurrence identified 7 is exactly the bold with only the initial letters capitalized. (Note: For subexpectation persona displayed in the table. sections and subsubsections, a word like the or a is not As the process performed is multidirectional searching the capitalized unless it is the first word of the header expectation persona, the greatest number of data occurrence Subsubsections found in the demographic personas composes exactly the The heading for subsubsections should be in Helvetica 9same as Table 4. point italic with initial letters capitalized. CONCLUSION FIGURES Upon execution of this study, it is concluded that despite
Figures should bethe inserted at the appropriate point some limitations k-means algorithm showed wasina your tool text. Figures may extend over the two columns up to even 17.8 to a positive result in the application of this procedure, cm (7") if necessary. and whitegroups photographs (not with the difficulty of theBlack data clustering A, B , F, G Polaroid and H. prints) may be mounted on the camera-ready paper with glue or double-sided tape. A service bureau can We presented procedure to build personas from the makehave a special printa of your black and white photography analysis of clusters identified by K-Means. Furthermore it for printing purposes (optional). To avoid smudges, attach is verified that personas can be accurately identified by the figures by paste or tape applied to their back surfaces only. process of derivation of the profile to datainfrom database Each figure should have a figure caption Times Roman.or derivation to identity between the typification of personas. Color figures should appear on separate pages so that they The of personas can occur by simplified methods. may creation be collected into a color section in the proceedings. The simple when expense using limited dataconference. set or the Colormethod figuresis are a large for the characters are only created by the Include them if they are imaginary absolutely scenario necessaryprocess. – and The method presented in this paper is effective for large only if your submission category permits them. databases. Especially when the data are group data: LANGUAGE, STYLE AND CONTENT behavioral data in this work. It is a method of creating The written and spoken is English. personas is not simplified; language it applies of dataCHI mining process Spelling and punctuation may consistently use any dialect and proof of training data. The personas created with this of English (e.g., British, Canadian or US). Hyphenation is method have been recognized as characters that identify optional. Please write for an international audience: individuals representative of the environment that the data were originally _ Write in a collected. straightforward style. Use simple sentence structure. Try to avoid sentences and complex With the objective of thislong study accomplished to the sentence structures. semicolons carefully. efficiency definition of the Use process established, measurement research of vocabulary the application process has _ Use common and basic (e.g., use the word continued. It is an important step to establish parameters to “unusual” rather than the word “arcane”). complete the automation of the process with the aspects of _ Briefly define or explain all technical terms. artificial intelligence. _ Explain all acronyms the first time they are used in ACKNOWLEDGMENTS your text –would e.g., “World Web (WWW)”. The authors like to Wide acknowledge the contributions of the team from e-Poupatempo represented by Iara Barbarian and Carlos Torres of the Operations Superintendent (SCO) of the São Paulo State. Special acknowledge to "Fundação de Amparo à Pesquisa do Estado de São Paulo" (FAPESP) for financial support for this publication.
de São (e.g., Paulo. Departamento de _ Universidade Explain local references not everyone knows all Engenharia Computação e Sistemas Digitais (2008). city names in de a particular country). Davidson, I.; And Wagstaff, L. _2. Basu, ExplainS.;“insider” comments. Ensure that your K. whole Constrained Clustering - Advances in meaning Algorithms, audience understands any reference whose you & Applications. Press, 2009. doTheory not describe (e.g., do CRC not assume that everyone has used a Macintosh or a particular application). 3. Cooper, A., The inmates are running the asylum: Why products drive us crazy and how to restore _ high-tech Explain colloquial language and puns. Understanding the sanity. Indianapolis, Ind.: Sams, 1999. phrases like “red herring” requires a cultural knowledge English. A., Humor and ironyR.M., are difficult translate. 4.ofCooper, Reimann, About toFace 2.0 The Essentials of Interaction Design. John Wiley & Sons; _ Use unambiguous forms for culturally localized 2nd edition, 2003 concepts, such as times, dates, currencies and numbers “1-5-A., 97”Reimann, or “5/1/97”R.,may mean D., 5 January 1 May 5.(e.g., Cooper, Cronin, About orFace 3: , and o'clock” may mean 7:00 am orJohn 19:00). The“seven Essentials of Interaction Design. Wiley & 2007 with the use of gender-specific pronouns _ Sons, Be careful 6.(he, Filgueiras, Luciaother et al.gendered Personas words como Modelo de she) and (chairman, manpower, inclusive Eletrônico. language (e.g., Usuários man-months). de Serviços Use de Governo In: Conferência Latinoamericana De Interação Humano she or he, s/he, they, chair, staff, staff-hours, person-Computador, 2005, México. Anais. , Cuernavaca: years) that is gender-neutral. If necessary you may be able to use2005. “he”p.and “she” in alternating sentences, so CLIHC, 319-324. that the two genders occur equally SeeK-means. [5] for 7. Jain, A. K. Data clustering: 50 yearsoften. beyond further advice and examples regarding gender and other Award winning papers from the 19th International personal attributes. Conference on Pattern Recognition (ICPR), 19th International Conference in Pattern Recognition (ICPR), ACKNOWLEDGMENTS 1, 2010), 651-666. doi: We 31, thank 8CHI, (June PDC, CSCW volunteers, all publications 10.1016/j.patrec.2009.09.011. support and staff who wrote and provided helpful comments this document. 8. Larose,onD.previous T. Dataversions MiningofMethods and Models. A JOHN WILEY & SONS, INC PUBLICATION, 2007. REFERENCES 1. R.E. Social impacts of computing: Codes de of 9. Anderson, Lattin, J.; Carroll, J. D. And Green, P. E. Análise professional ethics. Social Science Computing Review Dados Multivariados. CENGAGE Learning, 2011. 10, 2 (Winter 1992), 453-469. 10. Pruitt, J.; Adlin, T. The Persona Lifecycle : Keeping 2. CHI Format. Design. AvailableThe at PeopleConference in Mind Publications Throughout Product http://www.acm.org/sigchi/chipubform/. Morgan Kaufmann Series in Interactive Technologies.Interactive Technologies, 2006. 3. Conger., S., and Loch, K.D. (eds.). Ethics and computer use. Commun. ACM 38, 12 (entire issue). 11. Vattani, A. k-means requires exponentially many iterations even plane. lies In Proceedings of the 25th 4. Mackay, W.E.in the Ethics, and videotape, in annual symposium geometry (SCG Proceedings of CHI on '95 Computational (Denver CO, May 1995), ACM '09) (2009), 324-332. DOI=10.1145/1542362.1542419 Press, 138-145. http://doi.acm.org/10.1145/1542362.1542419. 5. Schwartz, M., and Task Force on Bias-Free Language. 12. Guidelines Xinwu, L. for Research on Writing. Text Clustering Algorithm Bias-Free Indiana University Based on K_means and SOM. In Proceedings of the Press, Bloomington IN, 1995. 2008 International Symposium on Intelligent Information Technology Application Workshops (IITAW '08) (2008), 341-344. DOI=10.1109/IITA.Workshops.2008.13 http://dx.doi.org/10.1109/IITA.Workshops.2008.13. 13. Wang, H.; Yang, H.; Xu, Z. And Yuan, Z. A Clustering
Algorithm Use K-Means in Intrusion The columns on the last page should beSOM of and equal length.
REFERENCES
1. Aquino JR., P.T. PICaP : padrões e personas para expressão da diversidade de usuários no projeto de interação. Tese (Doutorado) - Escola Politécnica da
Detection. In Proceedings of the 2010 International Conference on E-Business and E-Government (ICEE '10) (2010), 1281-1284. DOI=10.1109/ICEE.2010.327 http://dx.doi.org/10.1109/ICEE.2010.327.
99
