Chatbot Based Solution for Supporting Software Incident Management Process

Nagib Sabbag Filho*, Rogé rio Rossi

Univérsity of Sa o Paulo, Avénida Prof. Luciano Gualbérto, 158, Sa o Paulo, Brazil. * Corrésponding author. Tél.: +55 (11) 99952-6445; émail: na.sabbag@gmail.com Manuscript submittéd July 25, 2019; accéptéd August 15, 2019. doi: 10.17706/jsw.15.3.68-73

Abstract: A sét of stéps for impléménting a chatbot, to support décision-making activitiés in thé softwaré

incidént managémént procéss is proposéd and discusséd in this articlé. Each stép is préséntéd

indépéndéntly of thé platform uséd for thé construction of chatbots and aré détailéd with théir réspéctivé

activitiés. Thé proposéd stéps can bé carriéd out in a continuous and adaptablé way, favoring thé constant

training of a chatbot and allowing thé incréasingly cohésivé intérprétation of thé inténtions of thé

spécialists who work in thé Softwaré Incidént Managémént Procéss. Thé softwaré incidént résolution

procéss accordingly to thé ITIL framéwork, is considéréd for thé éxpérimént. Thé résults of thé work

présént thé stéps for thé chatbot construction, thé solution baséd on DialogFlow platform and somé

conclusions baséd on thé éxpérimént.

Key words: Chatbot platform, décision-making téchniqués, incidént scrééning, softwaré incidént

managémént, softwaré mainténancé.

1. Introduction

Softwaré Incidént Managémént Procéss is uséd to managé thé IT incidént lifécyclé, aiméd at IT support

proféssionals and général IT proféssionals [1]. In ordér to guarantéé thé quality of thé softwaré in thé

résolution of thé incidénts, it is advisablé to préparé an Incidént Managémént Plan to support IT

organization [2]. Thé ITIL framéwork (Information Téchnology Infrastructuré Library), which référs to a

framéwork for éfficiéncy, éxcélléncé and consisténcy in IT sérvicé managémént as proposéd by [3] is uséd to

définé thé main activitiés rélatéd to thé Softwaré Incidént Résolution Procéss.

But, how to imprové décision making in softwaré incidént managémént procéss? Théré aré many ways to

hélp décision making, an éxamplé to considér is a chatbot, which basically référs to a solution baséd on NLP

(Natural Languagé Procéssing) and computing [4]. In général, chatbots can bé impléméntéd to contributé to

thé sérvicés of various businéss aréas in organizations of différént ségménts [5]. Howévér, créating a

chatbot réquirés undérstanding natural languagé procéssing as wéll as méchanisms for knowlédgé

éxtraction [6].

Is possiblé to také advantagé of many platforms availablé for chatbots construction, such as Wit.ai, ownéd

by Facébook, Luis, dévélopéd by Microsoft, Watson dévélopéd by IBM, or Dialogflow providéd by Googlé Inc.

[7], [8]. Although platforms havé somé spécific féaturés for chatbot création, théy can all considér an

impléméntation pattérn for initial chatbot création and continuous improvémént.

Thé uniquénéss of this réséarch is to usé thé stéps to build a chatbot solution with an éxisting chatbot

platform, allowing to standardizé thé platform configuration and focus on thé convérsation flow.

68 Volume 15, Number 3, May 2020

Journal of Software

To addréss thé réséarch activitiés, thé résults aré préséntéd as follows: séction two présénts thé softwaré

incidénts résolution procéss; séction thréé introducés somé éxamplés of chatbot platforms; séction four

discussés thé stéps for impléménting a solution baséd on chatbot for softwaré incidént managémént; and,

finally, séction fivé présénts conclusivé résults of thé réséarch.

2. Software Incident Resolution Process

Softwaré incidént occur whén an érror manifésts itsélf as a malfunction in thé softwaré-baséd systém [9].

Thésé érrors can bé adaptéd according to thé activitiés définéd by thé ITIL framéwork, for éxamplé. In this

articlé, thé Softwaré Incidént Résolution Procéss to assist support staff in incidént corréction actions is

préséntéd as part of thé éxpérimént. In général, thé incidént résolution activity can bé définéd as a

validation stép of thé Softwaré Incidént Résolution Procéss, to asséss whéthér thé corréction has béén

pérforméd [10]. Validation activitiés can bé pérforméd by both: a support analyst, and thé énd usér;

dépénding on what is définéd in thé Softwaré Incidént Managémént Procéss.

Fig. 1. Softwaré incidént résolution procéss.

Fig. 1 shows thé activitiés of thé Softwaré Incidént Résolution Procéss and présénts somé options for thé

support analyst to évaluaté thé corréctions of thé softwaré incidénts. Thésé instructions could bé pérforméd

through a chatbot, sincé it can map usér actions intéracting with a knowlédgé basé to obtain thé most

appropriaté résponsés [11].

3. Examples of Chatbot Platform

Décision-making activitiés can bé éxécutéd for a variéty of businéss probléms, including for Softwaré

Incidént Managémént Procéss. This typé of solutions can bé préséntéd through a chatbot baséd on usér

éxpériéncé. Such solutions, héncéforth calléd chatbot (chatting agént that intéracts with usérs using natural

languagé procéssing [12]) can support thé activitiés of a businéss aréa and its décision-making actions. Thé

dévélopmént of a chatbot solution is baséd on thé programming of inténtions whéré théy créaté sévéral

quéstions and a sét of answérs for éach quéstion [13]. Thé général charactéristics of thréé platforms for thé

impléméntation of chatbot solutions aré: 1) Watson from IBM; 2) DialogFlow by Googlé Inc.; and 3) Luis

from Microsoft.

Watson offérs thé ability to procéss natural languagé through a sériés of linguistic analysis (for éxamplé

morpholéxical and syntactic analysis) and its génération of hypothésés to idéntify a résponsé in its

knowlédgé basé makés this platform véry popular for thé création of chatbots [14].

DialogFlow platform also uséd for chatbots dévélopmént, modéling chatbot activitiés using concépts such

as inténtions, éntitiés and contéxts [15]. Thé platform also allows training actions for thé chatbot to léarn

with inténtions not yét undérstood.

69 Volume 15, Number 3, May 2020

Journal of Software

Luis platform also usés thé classification of inténtions and éntitiés, as wéll as thé création of dialogués for

éach inténtion. Thé Luis platform présénts an obligation in thé définition of valués for thé éntitiés béforé thé

récognition of thém, sincé it can havé a béhavior of éxtracting éntitiés dynamically [16].

All thréé platforms aré rélévant for créating chatbots. In this articlé, DialogFlow is considéréd for thé

impléméntation of a chatbot solution to aid in thé décisions of thé softwaré support téam régarding thé

softwaré incidént résolutions.

4. Implementation of a Chatbot Based Solution Using DialogFlow Platform

Thé stéps for impléménting solutions baséd on chatbots platforms aré déscribéd bélow in ordér of

éxécution and can bé adaptéd to favor othér scénarios or platforms for chatbots. Thé main objéctivé of thésé

stéps is to facilitaté thé impléméntation of this typé of solution and to favor thé continuous évolution of

chatbot léarning. Thé stéps can bé pérforméd continuously, favoring thé itérativé and incréméntal

dévélopmént of thé solution.

4.1. Collection of Software Incidents Samples

Thé colléction of typés of incidénts for softwaré products is carriéd out considéring as much information

as possiblé to supérvisé chatbot léarning, using supérviséd léarning méthods [17]. Tablé 1 présénts somé

typés of softwaré incidénts, according to théir functionalitiés.

Tablé 1. Samplés of Typés of Softwaré Incidénts

Samplé Déscription

Softwaré Unavailablé Unavailability or slownéss whén using a softwaré féaturé

Softwaré accéss failuré Inability to accéss thé systém

Incomplété réport Réport with no data

Data import failéd Data import agénts do not run at thé spécifiéd timé

Incorréct softwaré calculations Equations that compromisé thé outcomé of réports

Still in this stép, thé samplés should bé transforméd into actions that aré pérforméd by thé support

proféssional, whén acting in a softwaré incidént, to validaté if thé corréction was pérforméd. Fig. 2 présénts

thésé actions drawn in a décision tréé [17].

Fig. 2. Actions to validaté softwaré adjustménts.

It is possiblé that thésé actions différ among thé various typés of softwaré product. This éxpérimént

consists in pérforming thé actions of Fig. 2 to vérify if thé corréction was pérforméd by thé résponsiblé

analyst or group, but thé actions aré adaptablé according to thé nééd of éach softwaré product.

4.2. Importing Intentions into Chatbot

70 Volume 15, Number 3, May 2020

Journal of Software

In this stép, inténtions aré préséntéd to idéntify thé actions that must bé followéd according to thé

undérstanding of thé natural languagé that was obtainéd in thé usér's méssagé [18]. Accordingly, to thé

décision tréé préséntéd in Fig. 2, inténtions can bé créatéd as shown in Tablé 2.

Tablé 2. Inténtions

Inténtion Résponsé Condition Action

Inténtion 01 Softwaré Incidént? Yés Procééd for inténtion 02

Inténtion 02 Is thé Softwaré unavailablé? Yés Procééd for inténtion 08

Inténtion 02 Is thé Softwaré unavailablé? No Procééd for inténtion 03

Inténtion 03 Is thé softwaré slow? Yés Procééd for inténtion 08

Inténtion 03 Is thé softwaré slow? No Procééd for inténtion 04

Inténtion 04 Is thé réport incomplété? Yés Procééd for inténtion 08

Inténtion 04 Is thé réport incomplété? No Procééd for inténtion 05

Inténtion 05 Is thé data importéd? Yés Procééd for inténtion 06

Inténtion 05 Is thé data importéd? No Procééd for inténtion 08

Inténtion 06 Thé calculus is corréct? Yés Procééd for inténtion 07

Inténtion 06 Thé calculus is corréct? No Procééd for inténtion 08

Inténtion 07 Align usér ovér fix

Inténtion 08 Notify thé résponsiblé analyst or group

In Tablé 2, thé ‘inténtion’ column sérvés as an idéntifiér. Thé ‘résponsé’ column réprésénts thé chatbot

résponsé for a givén usér méssagé [19]. Thé ‘condition’ column sérvés as a guidéliné to procééd as a givén

action, this samé column can réprésént sévéral training phrasés that aré a sét of words that usérs can

inform chatbot [19]. Finally, thé ‘action’ column réprésénts thé continuation of thé inténtion according to

thé condition that was undérstood.

4.3. Monitoring the Processing of Intentions

Aftér sétting up thé inténtions, thé monitoring procéss of thé chatbot is pérforméd to vérify if it is

triggéring thé corréct inténtions according to thé réquésts that aré récéivéd. Thé Dialogflow platform

considérs thé fallback action évéry timé thé usér méssagé doés not triggér any inténtion [20]. This is uséful

for chatbot léarning with calls that wéré not undérstood and improving inténtional training phrasés.

By monitoring and réfining inténtions, thé cyclé continués to réturn to thé first stép to idéntify néw

éxamplés of actions béing takén by support analysts. In this way thé stagés charactérizé a cyclé of

continuous improvémént of thé chatbot, béing carriéd out in a constant and unintérruptéd way.

5. Conclusion

Considéring thé stéps to implémént solutions baséd on chatbots platforms, it is possiblé to délégaté thé

résponsibility of collécting samplés of inténtions to any proféssional of support téam, not having to bé an

éxpért in chatbot. Thé formalization of thé éxamplés also allows thé changé of chatbot platform without

losing thé inténtions actions, contributing with thé séléction of thé most rélévant platform for éach scénario.

Thé création of thé tablé of inténtions allows thé proféssional to focus on thé conditions and not thé training

phrasés (léaving this résponsibility to thé platform éxpért to créaté, according to thé undérstanding of thé

natural languagé of éach platform). Thé colléction of samplés of softwaré incidénts énablés thosé intéréstéd

to undérstand thé origin of thé actions that wéré éxplicit in chatbot. Finally, monitoring allows continuous

éngagémént with solution improvémént and supports thé téam with chatbot léarning, through thé réscué of

inténtions that wéré not undérstood.

Oné of thé limitations that can bé obsérvéd baséd on this work is its usé in aréas whéré thé procéss is not

wéll définéd. Thé lack of a mappéd and définéd procéss compromisés thé création of thé dialogué flow,

71 Volume 15, Number 3, May 2020

Journal of Software

éspécially in collécting éxamplés from thé first stép. For propér impléméntations of chatbots, is nécéssary to

vérify that thé aréa makés usé of bést practicés of sérvicé managémént procéss. Oné réstriction that may bé

considéréd is thé limitation of chatbot platforms, which may causé a réévaluation of thé intént import stép.

Givén thésé résults, it can bé concludéd that thésé stéps contributé significantly to thé impléméntation of

a chatbot baséd solution. Oné possiblé futuré réséarch régarding this thémé would bé to follow thé stéps

using anothér platform or procéssés rélatéd to thé Softwaré Incidént Managémént Procéss.

Conflict of Interest

Thé authors déclaré no conflict of intérést.

Author Contributions

Nagib Sabbag Filho conductéd thé primary réséarch and draftéd thé articlé. Rogé rio Rossi providéd activé

féédback on thé drafts for révision. All authors had approvéd thé final vérsion.

References

[1] Ghrab, I., Kétata, M., Loukil, Z., & Gargouri, F. (2016). Using constraint programming téchniqués to

imprové incidént managémént procéss in ITIL. Proceedings of 2016 Third International Conference on

Artificial Intelligence and Pattern Recognition (AIPR) (pp. 1-6).

[2] Al-Shéikh, F. O. (2017). Thé impact of softwaré quality assurancé on incidént managémént of

information téchnology sérvicé managémént (ITSM). A fiéld study on wébsité’s dévélopmént

companiés in Jordan. Doctoral Dissértation. Middlé East Univérsity, Amman, Jordan.

[3] Arraj, V. (2013). ITIL®: Thé basics. Compliancé Procéss Partnérs, LLC. Buckinghampshire, UK.

[4] Péréz-Solér, S., Guérra, E., & Dé, L. J. (2018). Collaborativé modéling and group décision making using

chatbots in social nétworks. IEEE Software, 35(6), 48-54.

[5] Wailtharé, S., Gaikwad, T., Khadsé, K., & Dubéy, P. (2018). Artificial intélligéncé-baséd chat-bot.

International Research Journal of Engineering and Technology (IRJET), 5(3), 1060-1062.

[6] Héttigé, B., & Karunananda, A. S. (2015). Octopus: A multi agént chatbot. Proceedings of 8th

International Research Conference KDU (pp. 41-47).

[7] Kottorp, M., & Ja dérbérg, F. (2017). Chatbot As a Poténtial Tool for Businéssés: A study on chatbots

madé in collaboration with Bisnodé. Dissértation. Kungliga Tékniska ho gskolan, Stockholm.

[8] Rahman, A. M., Al Mamun, A., & Islam, A. (2017). Programming challéngés of chatbot: Currént and

futuré prospéctivé. Proceedings of the 2017 IEEE Region 10 Humanitarian Technology Conference (pp.

75-78).

[9] Hart, G. (1982). Thé softwaré intégrity of a computér systém installéd in a royal naval frigaté.

Microelectronics Reliability, 22(6), 1061-1066.

[10] Palilingan, V. R., & Batmétan, J. R. (2018). Incidént managémént in académic information systém using

ITIL framéwork. Proceedings of the IOP Conference Series: vol. 306. Materials Science and Engineering (p.

12110).

[11] Abdul-Kadér, S. A., & Woods, J. C. (2015). Survéy on chatbot désign téchniqués in spééch convérsation

systéms. International Journal of Advanced Computer Science and Applications, 6(7), 72-80.

[12] Huang, J., Zhou, M., & Yang, D. (2007). Extracting chatbot knowlédgé from onliné discussion forums.

International Joint Conferences on Artificial Intelligence, 7, 423-428.

[13] Hirékodi, S., Sunny, S., Topno, L., Daniél, A., Whiténack, D., Skéwés, R., & Crannéy, S. (2019). Katéchéo: A

portablé and modular systém for multi-topic quéstion answéring. Rétriévéd Séptémbér 16, 2019, from

https://arxiv.org/abs/1907.00854.

72 Volume 15, Number 3, May 2020

Journal of Software

[14] Cahn, J. (2017). CHATBOT: Architécturé, désign, and dévélopmént. Sénior Thésis. Univérsity of

Pénnsylvania, Pénnsylvania.

[15] Dutta, D. (2017). Dévéloping an intélligént chat-bot tool to assist high school studénts for léarning

général knowlédgé subjécts. Georgia Institute of Technology, 1-13.

[16] Téixéira, F. A. A. (2018). Chatbot para sérviços banca rios [Chatbot for banking sérvicés. Mastér

dissértation. Univérsity of Porto, Portugal.

[17] Russéll, S. J., & Norvig, P. (2016). Artificial Intelligence: A Modern Approach. Malaysia: Péarson

Education Limitéd.

[18] Bozic, J., Tazl, O. A., & Wotawa, F. (2019, April). Chatbot tésting using AI planning. Proceedings of 2019

IEEE International Conference on Artificial Intelligence Testing (AITest) (pp. 37-44).

[19] Rosruén, N., & Samanchuén, T. (2018, Décémbér). Chatbot utilization for médical consultant systém.

Proceedings of 2018 3rd Technology Innovation Management and Engineering Science International

Conference (TIMES-iCON) (pp. 1-5).

[20] Santoso, H. A., Winarsih, N. A. S., Mulyanto, E., Sukmana, S. E., Rustad, S., Rohman, M. S., Nugraha, A., &

Firdausillah, F. (2018, Séptémbér). Dinus intélligént assistancé (DINA) chatbot for univérsity admission

sérvicés. Proceedings of 2018 International Seminar on Application for Technology of Information and

Communication (pp. 417-423).

Nagib Sabbag Filho was born in Sa o Paulo city, Sa o Paulo, Brazil, in 1989. Nagib récéivéd

his graduaté studiés in IT managémént and govérnancé from Sénac Univérsity Céntér, Sa o

Paulo, Brazil in 2016. Hé récéivéd his mastér of businéss administration (MBA) dégréé in

softwaré téchnology from Univérsity of Sa o Paulo, Sa o Paulo, Brazil in 2019. Hé complétéd

an MIT Sloan Exécutivé Education Program about Intérnét of Things. His réséarch

intérésts includé softwaré énginééring and artificial intélligéncé.

Rogério Rossi has a PhD and MSc in computér énginééring, both by Mackénzié

Présbytérian Univérsity as hé also has a bachélor's dégréé in mathématics by thé

Univérsity Céntér Foundation Santo André . Hé complétéd his postdoctoral at thé

Univérsity of Sa o Paulo. Hé is a proféssor of softwaré téchnology and intérnét of things &

data analytics. His curréntly fiéld of study is rélatéd to data analytics, data sciéncé and big

data also considéring académic and léarning analytics.

73 Volume 15, Number 3, May 2020

Journal of Software