Tsarin Sada na Kamancecenci don Sauyin Sabis na Yanar Gizo na Ma'ana: Hanyar Tushen Sada

1. Gabatarwa

Sabis na Yanar Gizo sun zama ginshiƙai na asali na aikace-aikacen rarrabuwa na zamani. Wani babban kalubale a cikin haɗa su ta atomatik shine sarrafa gazawa ko rashin samun sabis ɗin da ke cikin su ta hanyar sauya mai tasiri. Wannan takarda ta magance wannan ta hanyar wucewa daga rarraba sauƙaƙan sabis masu sauya, ta gabatar da sabuwar hanyar tushen sada inda nodes ke wakiltar ayyukan sabis na Yanar Gizo kuma gefuna ke wakiltar kamancecenci na aiki. Wannan tsarin yana nufin samar da tsari mai wadata, mafi ƙarfi don bincike da gano sabis masu sauya, a ƙarshe inganta ƙarfi da sassaucin sabis ɗin da aka haɗa.

2. Bayan Fage & Ayyukan Da Suka Danganta

2.1. Haɗa Sabis na Yanar Gizo & Kalubale

Hangar haɗa sabis ta atomatik tana cikas da yanayin Yanar Gizo mai canzawa da sauri. Sabis na iya gaza, a sabunta su, ko zama ba su samuwa. Don haka sauya ba abin al'ajabi ba ne amma larura ne don kiyaye ci gaban sabis. Bincike na al'ada yana gano sabis don buƙata, amma sauya dole ne ya nemo maye gurbin abubuwan da aka riga aka tura yayin da ake kiyaye aikin gaba ɗaya.

2.2. Hanyoyin Sauyin Da Ake Da Su

Ayyukan da suka gabata sun fi mayar da hankali kan rarrabuwa bisa kaddarorin aiki da waɗanda ba na aiki ba (QoS). Hanyoyin gama gari sun haɗa da:

• Tushen Al'umma/Rukunin: Rarraba sabis masu kamancecenci na aiki, sau da yawa ana danganta su da ra'ayoyin ilimin ma'ana [1, 2].
• Daidaita Tsarin Mu'amala: Ayyana matakan kamancecenci (misali, daidai, maye gurbi) bisa ga ƙidaya da nau'ikan aiki/sigogi [3].

Duk da yake suna da amfani, waɗannan hanyoyin sau da yawa ba su da ƙayyadaddun bayanai da mahallin dangantaka don bincika cikakken yuwuwar sauya.

3. Tsarin Sada da Ake Shawarar

3.1. Gina Sada

Babban ƙirƙira shine ƙirƙirar sararin sauya a matsayin zane $G = (V, E)$.

• Matsakaici (V): Kowane matsayi $v_i \in V$ yana wakiltar takamaiman aiki daga tsarin mu'amalar sabis na Yanar Gizo (misali, `getWeather`, `convertCurrency`).
• Gefuna (E): Gefen da ba a jagorance ba $e_{ij} \in E$ yana haɗa matsakaici biyu $v_i$ da $v_j$ idan ayyukansu masu dacewa an ɗauke su suna da kamancecenci na aiki bisa ga ma'aunin kamancecenci $sim(v_i, v_j) > \theta$, inda $\theta$ shine bakin kamancecenci.

Wannan tsari yana canza jerin sabis marasa zurfi zuwa taswira mai wadata ta dangantaka, inda rukunoni, hanyoyi, da matsakaicin nodes ke bayyana tsarin sauya.

3.2. Ma'auni na Kamancecenci

Takardar ta ba da shawarar ma'auni huɗu na kamancecenci bisa kwatanta sigogi na shigarwa da fitarwa na ayyuka, ta amfani da bayanan ma'anarsu (misali, ra'ayoyin ilimin ma'ana). Ma'auni suna iya haɗawa da:

1. Kamancecenci na Saitin Sigogi: Kwatanta saitin ra'ayoyin shigarwa/fitarwa (misali, ma'auni na Jaccard).
2. Kamancecenci na Nau'in Sigogi: Yin la'akari da tazarar ma'ana tsakanin ra'ayoyin sigogi a cikin ilimin ma'ana.
3. Kamancecenci na Tsarin Mu'amala: Yin la'akari da tsari da ƙidaya na sigogi.
4. Ma'auni na Haɗaɗɗe: Haɗuwa mai nauyi na waɗannan da ke sama.

4. Cikakkun Bayanai na Fasaha & Hanyar Aiki

4.1. Tsarin Lissafi

Ma'auni na asali zai iya zama aikin kamancecenci mai nauyi. Bari $I_x, O_x$ su zama saitin ra'ayoyin ma'ana don shigarwa da fitarwa na aikin $x$. Ana iya ayyana maki kamancecenci tsakanin ayyuka $a$ da $b$ kamar haka:

$sim(a, b) = \alpha \cdot \text{sim}_{input}(I_a, I_b) + \beta \cdot \text{sim}_{output}(O_a, O_b)$

inda $\alpha + \beta = 1$ nauyi ne, kuma $\text{sim}_{input/output}$ zai iya zama ma'aunin kamancecenci na saiti kamar:

$\text{Jaccard}(X, Y) = \frac{|X \cap Y|}{|X \cup Y|}$

Don kamancecenci na ma'ana tsakanin ra'ayoyi ɗaya ɗaya $(c_i, c_j)$, ana iya haɗa ma'auni na tushen ilimin ma'ana kamar Wu & Palmer ko Lin kamancecenci, ana ɗauka daga ayyukan da aka kafa a cikin ilimin harshe na kwamfuta da wakilcin ilimi kamar yadda ake gani a albarkatun kamar bayanan WordNet.

4.2. Misalin Tsarin Bincike

Yanayi: Sabis ɗin haɗa tikitin tafiya ya gaza lokacin da aikinsa na "FlightSearch" ya zama ba ya samuwa.

1. Gano Matsakaici: Nemo matsayin aikin `FlightSearch` da ya gaza a cikin sada na kamancecenci.
2. Binciken Maƙwabta: Bincika maƙwabtansa kai tsaye (ayyuka masu kamancecenci sosai). Waɗannan su ne zaɓuɓɓukan sauya na farko (misali, `SearchFlights`, `FindAirfare`).
3. Gano Hanya: Idan babu maƙwabci kai tsaye da ya samu, bincika hanyoyin tsalle-2. Wani aikin `SearchTravel` zai iya haɗa `FlightSearch` zuwa `BusSearch`. Duk da yake ba maye gurbi kai tsaye ba ne, `BusSearch` na iya zama madadin da za a iya amfani da shi a cikin sake tsara haɗawa.
4. Binciken Rukuni: Gano rukunin da ke ɗauke da matsayin da ya gaza. Duk ayyuka a cikin wannan rukunin suna raba ainihin kamancecenci na aiki, suna ba da tafki na yuwuwar maye gurbi.
5. Binciken Matsakaici: Matsakaicin da ke da babban matsayi na tsakiya yana wakiltar ayyuka "na gama gari" ko "na gama gari", mai yuwuwar zama maye gurbi mafi ƙarfi.

Wannan tsarin yana wucewa daga yanke shawara na "mai sauya/ba mai sauya ba" zuwa bincike mai mataki, mai mahallin madadin.

5. Kimantawa ta Gwaji & Sakamako

5.1. Bayanan Gwaji & Saiti

An yi kimantawa akan ma'auni na sabis na Yanar Gizo masu bayanin ma'ana (misali, bayanin OWL-S ko SAWSDL). An gina sada ta amfani da ma'auni daban-daban na kamancecenci da bakuna.

5.2. Binciken Tsarin Sada & Binciken

Takardar ta yi kwatancen kimantawa na tsarin sada da aka samar. Ma'auni masu mahimmanci da aka yi bincike suka haɗa da:

• Rarraba Mataki: Don gano idan sada ba shi da ma'auni (ƴan cibiyoyi) ko bazuwar.
• Ma'aunin Haɗawa: Yana auna yadda maƙwabta suke daure, yana nuna al'ummomin aiki.
• Abubuwan Haɗin Haɗin: Yana bayyana rukunin sabis ɗin da ke ware.
• Tsawon Hanya: Matsakaicin mafi gajeren hanya tsakanin matsakaici, yana nuna yadda dangantakar sauya ke "nisa".

Bayanin Ginshiƙi (An fahimta): Taswirar ginshiƙi tana kwatanta Matsakaicin Ma'aunin Haɗawa a cikin sada da aka gina tare da ma'auni huɗu daban-daban na kamancecenci. Ma'auni na 3 (Tsarin Mu'amala) yana iya haifar da ma'auni mafi girma, yana nuna yana samar da tsari mai ƙarfi, kamar al'umma, wanda ake so don gano ƙungiyoyin sauya bayyananne. Taswirar layi tana nuna yadda Adadin Abubuwan Haɗin Haɗin ke canzawa tare da bakin kamancecenci $\theta$: babban $\theta$ yana haifar da ƙananan abubuwa da yawa (sauya mai tsauri), yayin da ƙaramin $\theta$ ya haɗa su zuwa ƙananan, manyan abubuwa (sauya mai faɗi).

Babban Sakamako: Hanyar sada ta yi nasara wajen bayyana ƙarin cikakkun bayanai da tsari na sabis masu sauya idan aka kwatanta da rarrabuwa mara zurfi. Ya ba da damar gano ba kawai maye gurbi kai tsaye ba har ma da madadin kai tsaye da al'ummomin aiki, yana tabbatar da ainihin hasashe.

6. Fahimta ta Asali & Bincike Mai Zurfi

Fahimta ta Asali: Aikin Cherifi juyi ne mai hikima daga ɗaukar sauya sabis a matsayin matsalar kasida zuwa ɗaukar shi a matsayin matsalar kewayawa sada. Ainihin darajar ba kawai a lissafa maye gurbi masu yuwuwa ba ne, amma a fahimci yanayin kusancin aiki. Wannan yana kama da sauyi a cikin tsarin ba da shawara daga tacewa haɗin gwiwa mai sauƙi zuwa hanyoyin tushen zane waɗanda ke ɗaukar ƙwaƙƙwaran yanayin dangantaka, wani yanayi da aka rubuta da kyau a cikin wallafe-wallafe daga cibiyoyi kamar Aikin Bincike na Sada na Stanford.

Kwararar Hankali: Hankali yana da ban sha'awa: 1) Aikin sabis ana ayyana shi ta hanyar ayyuka. 2) Kamancecenci na aiki ana iya ƙididdige shi ta hanyar daidaita I/O na ma'ana. 3) Don haka, sada na waɗannan dangantakar kamancecenci a zahiri yana taswira ƙasar sauya. Wannan yana motsa abin da ke haifar da sauya daga bincike mai amsawa zuwa binciken tsari mai himma. Amfani da bayanan ma'ana yana da mahimmanci a nan—shi ne abin da ya ɗaga hanyar daga daidaita suna na tsari zuwa kwatanta aiki mai ma'ana, darasi da aka koya daga ƙoƙarin Yanar Gizo na Ma'ana.

Ƙarfi & Kurakurai: Ƙarfinsa shine gaskiyar wakilci. Sada a zahiri yana ɗaukar "matakan rabuwa" tsakanin sabis, yana ba da ba kawai 'yan takara ba har ma da madadin da aka jera da zaɓuɓɓukan fallback. Yana ƙetare tsaurin rarrabuwa cikin kyakkyawa. Duk da haka, kurakuri na takardar, gama gari a cikin samfuran sada na farkon zamani, shine dogaro mai yawa akan inganci da kasancewar bayanan ma'ana. A duniyar gaske, sabis da yawa ba su da cikakkun bayanin OWL-S. Ma'aunin kamancecenci da aka ba da shawara, duk da yake suna da hankali, suma sun ɗan ɓace; Aiwowinsu na gaske a duniya da hayaniyar, rashin cikakken, ko metadata iri-iri shine gwaji na gaske. Bugu da ƙari, binciken da alama ya fi mayar da hankali kan tabbatar da tsarin sada maimakon ainihin ƙimar nasarar sauya a cikin injin haɗawa mai rai—ainihin KPI.

Fahimta Mai Aiki: Ga masu aiki, wannan bincike ya ba da umarni ga ayyuka biyu: Na farko, saka hannun jari a cikin bayanin ma'ana na tsarin mu'amalar sabis; shine man fetur don wannan inji mai ƙarfi. Na biyu, haɗa kayan aikin binciken sada (kamar Gephi ko NetworkX) cikin sarrafa rajistar sabis. Kar a adana sabis kawai; taswira su. Ga masu bincike, mataki na gaba a bayyane yake: haɗa wannan samfurin. Haɗa halayen QoS a matsayin nauyin gefe (ƙirƙirar sada mai yawa). Haɗa yanayin lokaci don ƙirƙira canjin sabis. Bincika koyon inji, watakila ta amfani da Cibiyoyin Jijiyoyin Zane (GNNs), don hasashen hanyoyin haɗin sauya daga bayanan ɓangare, kama da yadda samfura kamar GraphSAGE ke aiki. Nan gaba na ingantaccen haɗa sabis yana cikin waɗannan zane-zane masu wadata, masu koyo.

7. Hangar Aikace-aikace & Hanyoyin Gaba

Samfurin sauya na tushen sada yana da aikace-aikace masu ban sha'awa fiye da dawo da gazawa na asali:

• Kasuwannin Sabis Masu Sauƙi: Nuna yanayin sabis a matsayin zane-zane masu mu'amala ga masu bayarwa da masu amfani.
• Haɓaka Haɗawa: Yin amfani da hanyoyin sada don gano sabbin sarƙoƙin sabis waɗanda ke cimma irin wannan buri tare da abubuwa daban-daban, mai yuwuwar inganta farashi ko aiki.
• Haɗa Tsarin Tsoho: Taswirar APIs na ƙananan sabis na zamani da ayyukan tsarin tsoho don nemo dabarun nadewa ko maye gurbi.
• Ƙarfin Hankali Mai Himma: Sa ido kan "lafiya" na mahimman matsakaicin cibiyoyin a cikin sada na sauya da kuma kafin haka samun madadin.

Hanyoyin Bincike na Gaba:

1. Haɗawa tare da QoS: Ƙirƙirar sada masu yawa inda wani Layer shine kamancecenci na aiki kuma wani shine alaƙar QoS, ta amfani da dabarun bincike sada masu yawa.
2. Kamancecenci na Tushen Koyo: Yin amfani da NLP da koyo mai zurfi (misali, masu fassara jumla kamar BERT) don ƙididdige kamancecenci na aiki daga bayanin sabis maras tsari, rage dogaro akan ma'ana mai tsari.
3. Ci gaban Sada Mai Sauƙi: Haɓaka samfuran inda sada na sauya ke sabuntawa cikin ainihin lokaci yayin da ake buga sabis, sabuntawa, ko ƙi.
4. Sauya Mai Bayyanawa: Yin amfani da tsarin sada don samar da bayanin da mutum zai iya karantawa don dalilin da ya sa aka zaɓi wani takamaiman sabis a matsayin maye gurbi (misali, "An zaɓe shi saboda yana raba kashi 80% na abubuwan shigar da kuke buƙata kuma an haɗa shi ta hanyar cibiyar sabis mai aminci").

8. Nassoshi

1. Klusch, M., & Gerber, A. (2006). Semantic Web Service Composition Planning with OWLS-XPlan. Proceedings of the AAAI Fall Symposium on Semantic Web for Collaborative Knowledge Acquisition.
2. Dong, X., et al. (2004). Similarity Search for Web Services. Proceedings of the 30th VLDB Conference.
3. Mokhtar, S. B., et al. (2006). Efficient Semantic Service Discovery in Pervasive Computing Environments. Proceedings of the 4th ACM International Middleware Conference.
4. Stanford Network Analysis Project (SNAP). http://snap.stanford.edu. (Don ra'ayoyin binciken sada da kayan aiki).
5. Wu, Z., & Palmer, M. (1994). Verbs Semantics and Lexical Selection. Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics. (Don ma'auni na kamancecenci na ma'ana).
6. Hamilton, W., Ying, R., & Leskovec, J. (2017). Inductive Representation Learning on Large Graphs. Advances in Neural Information Processing Systems 30 (NIPS 2017). (Don Cibiyoyin Jijiyoyin Zane kamar GraphSAGE).