Design of the system of automated generation of poetry works

2021;
: 01-14
https://doi.org/10.23939/ujit2021.02.001
Received: October 15, 2021
Accepted: November 23, 2021
1
Lviv Polytechnic National University, Lviv, Ukraine
2
Lviv Polytechnic National University, Lviv, Ukraine

 Features of designing a system of automated generation of poetic works, which opens up new opportunities for artistic speech and show business, especially the preparation of poems and songs have been considered. Quite often lyrics without special content become successful due to the lack of complex plots, as well as due to the unobtrusiveness and ease of perception by listeners. Well-known literature sources and available software products that can generate poetic works by combining different methods and algorithms are analyzed. It has been established that none of them is able to ensure the content and uniqueness of the poetic work at the same time, especially in the Ukrainian language. The existing approaches to the generation of poetic works are analysed, among which the relevant is a method based on templates, generation and testing, evolutionary algorithms and the method based on specific cases. Peculiarities of generating poetical works, first of all rhyming rules, types of strophes, poetic rhythms and sizes have been investigated. An approach to automated generation of poetic works using evolutionary algorithms and a method based on specific cases have been developed. Their combination resembles a sequence of actions for creative personalities when creating poems or writing lyrics.

Peculiarities of neural network organization for automated generation of poetic works have been considered. It is proposed to perform neural network training using the method of inverse propagation and using a genetic algorithm. The principle of operation of algorithms for finding optimal solutions which contain such consecutive stages as initialization, evaluation of solutions, population selection, evolution of solutions, is analysed. Their interaction and various opportunities for neural network learning have been investigated in detail. An algorithm has been developed according to which the software application will analyse the poetic works offered by the user and generate new variants of it on the basis received from the neural network of logically connected words or lines of the verse in the poem. The user can edit both the components of the poem and the generated poetic works, and thus can train the neural network. The specification of requirements to the software application has been developed, the basic requirements to the user interface are defined, and also potential classes of users who will use it are established.

  1. Be­lek­ho­va, L. I. (2009). Syntac­tic or­ga­ni­za­ti­on of texts of mo­dern Ame­ri­can po­etry: cog­ni­ti­ve-se­mi­otic and lin­guo-syner­ge­tic as­pects. Bul­le­tin of VN Ka­ra­zin Khar­kiv Na­ti­onal Uni­ver­sity, 57(838), 20–28. Khar­kiv: KhNU Pub­lis­hing Hou­se. [In Uk­ra­ini­an].
  2. Bod­yanskiy, Ya., Po­pov, S., & Rybalchen­ko, T. (2008). Mul­ti­la­yer neu­ro-fuzzy net­work for short term electric lo­ad fo­re­cas­ting. Lec­tu­re No­tes in Com­pu­ter Sci­en­ce – Ber­lin, He­idel­berg: Sprin­ger-Ver­lag, 5010, 339–348. Ret­ri­eved from: https:// link.sprin­ger.com/chap­ter/10.1007/978-3-540-79709-8_34
  3. Brick, O. M. (1927). Rhythm and syntax. New Lef, 3–6, 15–37. Mos­cow: Go­siz­dat. [In Rus­si­an].
  4. Ca­za. (2021). Synap­tic.js. The ja­vascript archi­tec­tu­re-free neu­ral net­work lib­rary for no­de.js and the brow­ser. Ret­ri­eved from: https://ca­za.la/synap­tic/#/.
  5. Cic­hoc­ki, A., & Un­be­ha­uen, R. (1993). Neu­ral Net­works for Op­ti­mi­za­ti­on and Sig­nal Pro­ces­sing. Stuttgart: Te­ub­ner, 526 p. Ret­ri­eved from: https://www.ama­zon.com/Neu­ral-Net­works-Op­ti­mi­za­ti­on-Sig­nal-Pro­ces­sing/dp/0471930105
  6. Di­ag­rams.net. (2021). Draw.io. Ret­ri­eved from: https://abo­ut.draw.io/abo­ut-us/
  7. Du, K.-L., & Swamy, M. N. S. (2014). Mul­ti­la­yer Per­ceptrons: Archi­tec­tu­re and Er­ror Backpro­pa­ga­ti­on. Neu­ral Net­works and Sta­tis­ti­cal Le­ar­ning, 83–126. https://doi.org/10.1007/978-1-4471-5571-3_4
  8. Eic­hen­ba­um, B.M. (1987). The the­ory of the "For­mal Met­hod". Abo­ut li­te­ra­tu­re. Mos­cow: Sov. Wri­ter, 375–408. [In Rus­si­an].
  9. Gas­pa­rov, M. L. (1985). Op­po­si­ti­on "ver­se-pro­se" and the for­ma­ti­on of Rus­si­an li­te­rary ver­se. Rus­si­an ver­si­fi­ca­ti­on: Tra­di­ti­ons and de­ve­lop­ment prob­lems, 4, 264–277. Mos­cow. [In Rus­si­an].
  10. Gas­pa­rov, M. L. (1994). Lin­gu­is­tics of ver­se. Iz­ves­tia RAN. Li­te­ra­tu­re and Lan­gua­ge Se­ri­es, 53(6), 28–35. Mos­cow. [In Rus­si­an].
  11. Gas­pa­rov, M. L. (2001). Verb rhyme and the syntax of a po­etic li­ne. Rus­si­an lan­gua­ge in sci­en­ti­fic co­ve­ra­ge, 1, 148–160. Mos­cow. [In Rus­si­an].
  12. Ge­qay, R., & Liu, T. (1997). Non­li­ne­ar mo­de­ling and pre­dic­ti­on with fe­ed for­ward and re­cur­rent net­works. Physi­ca D, 108, 119–134. https://doi.org/10.1016/S0167-2789(97)82009-X
  13. Ger­vas, P. (2002). Explo­ring Qu­an­ti­ta­ti­ve Eval­ua­ti­ons of the Cre­ati­vity of Au­to­ma­tic Po­ets. Pab­lo Ger­vas. 15th Eu­ro­pe­an Con­fe­ren­ce on Ar­ti­fi­ci­al In­tel­li­gen­ce. Ret­ri­eved from: http:// nil.fdi.ucm.es/si­tes/de­fa­ult/fi­les/Ger­va­sE­CA­Iws2002.pdf.
  14. Girshick, R., Do­nah­ue, J., Dar­rell, T., & Ma­lik, J. (2016). Re­gi­on-Ba­sed Con­vo­lu­ti­on Net­works for Ac­cu­ra­te Ob­ject De­tec­ti­on and Seg­men­ta­ti­on. In IEEE Tran­sac­ti­ons on Pat­tern Analysis and Mac­hi­ne In­tel­li­gen­ce, 38(1), 142–158. https://doi.org/10.1109/TPA­MI.2015.2437384
  15. Hu­go Gonçalo Oli­ve­ira, Raq­uel Hervás, Al­ber­to Díaz & Pab­lo Gervás. (2017). Mul­ti­lan­gua­ge Ex­ten­si­on and Eval­ua­ti­on of a Po­etry Ge­ne­ra­tor. In Jo­ur­nal of Na­tu­ral Lan­gua­ge En­gi­ne­ering, 23(6), 929–967. https://doi.org/10.1017/S1351324917000171
  16. Hu­go Gonçalo Oli­ve­ira, Ti­ago Men­des, Ana Bo­avi­da, Ai Na­ka­mu­ra & Mar­ga­re­ta Ac­ker­manc. (2019). Co-Po­eTryMe: In­te­rac­ti­ve po­etry ge­ne­ra­ti­on. In Cog­ni­ti­ve Systems Re­se­arch, 54, 199–216. https://doi.org/10.1016/j.cogsys.2018.11.012
  17. Hu­go Gonçalo Oli­ve­ira. (2017). O Po­eta Ar­ti­fi­ci­al 2.0: Incre­asing me­aningful­ness in a po­etry ge­ne­ra­ti­on Twit­ter bot. In Pro­ce­edings of the Workshop on Com­pu­ta­ti­onal Cre­ati­vity in Na­tu­ral Lan­gua­ge Ge­ne­ra­ti­on (CC-NLG 2017), 11–20, San­ti­ago de Com­pos­te­la, Spa­in. ACL Press. https://doi.org/10.18653/v1/W17-3902
  18. Hu­go Gonçalo Oli­ve­ira. (De­cem­ber, 2015). Tra-la-lyrics 2.0: Au­to­ma­tic ge­ne­ra­ti­on of song lyrics on a se­man­tic do­ma­in In Jo­ur­nal of Ar­ti­fi­ci­al Ge­ne­ral In­tel­li­gen­ce, 6(1), 87–110. https://doi.org/10.1515/jagi-2015-0005
  19. IEEE 730 Stan­dard for Softwa­re Qua­lity As­su­ran­ce Plans. (2014). The Insti­tu­te of Electri­cal and Electro­nics En­gi­ne­ers, Inc. Ret­ri­eved from: https://stan­dards.ieee.org/stan­dard/730-2014.html.
  20. Inspi­red. (No­vem­ber 27, 2020). Go­og­le's ar­ti­fi­ci­al in­tel­li­gen­ce al­go­rithm has le­ar­ned to wri­te po­etry, imi­ta­ting fa­mo­us po­ets. Ret­ri­eved from: https://inspi­red.com.ua/stre­am/al­gorytm-shtuchno­go-in­te­lek­tu-vid-go­og­le-navchyvsya-pysaty-virshi-nas­li­du­yuchy-vi­domyh-po­etiv/
  21. Is­tan­bul. (2021). Ja­vaScript test co­ve­ra­ge ma­de simple. Ret­ri­eved from: https://is­tan­bul.js.org/.
  22. Ja­cob­son, R. O. (1975). Lin­gu­is­tics and po­etics. Struc­tu­ra­lism: pros and cons: Sat. transla­ted ver­ses. Mos­cow Prog­ress, 193–230. [In Rus­si­an].
  23. Jie Wang, Chengzhi Zhang, Mengying Zhang & San­hong Deng. (22 Jun 2018). Ci­ta­ti­on AS: A To­ol of Au­to­ma­tic Sur­vey Ge­ne­ra­ti­on Ba­sed on Ci­ta­ti­on Con­tent. Jo­ur­nal of Da­ta and In­for­ma­ti­on Sci­en­ce, 20–37. https://doi.org/10.2478/jdis-2018-0007
  24. Khro­len­ko, A. T. (2012). Au­to­ma­ted con­cor­dan­ce: ex­pe­ri­en­ce of cre­ati­on and prac­ti­ce of use. Ret­ri­eved from: https://cyber­le­nin­ka.ru/ar­tic­le/n/av­to­ma­ti­zi­ro­vannyy-kon­kor­dans-opyt-soz­da­ni­ya-i-prak­ti­ka-is­pol­zo­va­ni­ya. [In Rus­si­an].
  25. Kol­mo­go­rov, A. N. (2002). Li­ne, stan­za and ver­se as a rhythmic system. Grin­ba­um O. N. Ma­te­ri­als of the XXXI All-Rus­si­an Sci­en­ti­fic and Met­ho­do­lo­gi­cal Con­fe­ren­ce of Te­ac­hers and Postgrad­ua­tes of the Phi­lo­lo­gi­cal Fa­culty of St. Pe­tersburg Sta­te Uni­ver­sity, 4(2), 12–28. [In Rus­si­an].
  26. Lot­man, M. Yu. (1999). Analysis of the po­etic text. The struc­tu­re of the ver­se. Lot­man Yu. M. On Po­ets and Po­etry, 4, 18–253. SPb.: Pub­lis­hing hou­se of SPb. Ret­ri­eved from: https://www.rut­he­nia.ru/lot­man/pa­pers/apt/. [In Rus­si­an].
  27. Man­dic, D. P., & Cham­bers, J. A. (2001). Re­cur­rent Neu­ral Net­works for Pre­dic­ti­on: Le­ar­ning Al­go­rithms, Archi­tec­tu­res and Sta­bi­lity. Chic­hes­ter: John Wi­ley&Sons, 285 p. https://doi.org/10.1002/047084535X
  28. Ma­nu­rung, H. (2004). An evo­lu­ti­onary al­go­rithm appro­ach to po­etry ge­ne­ra­ti­on. Doc­to­ral Dis­ser­ta­ti­on for Techni­cal Sci­en­ces. Uni­ver­sity of Edin­burgh. Col­le­ge of Sci­en­ce and En, 367 p.
  29. Ma­zur, M. (2015). A Step by Step Backpro­pa­ga­ti­on Example. Matt Ma­zur. Ret­ri­eved from: https://mattma­zur.com/2015/03/ 17/a-step-by-step-backpro­pa­ga­ti­on-example/.
  30. Mic­ha­le­wicz, Z. (1992). Ge­ne­tic alo­rithms + da­ta struc­tu­res = evo­lu­ti­onary prog­rams. Mic­ha­le­wicz. Char­lot­te, USA: Uni­ver­sity of North Ca­ro­li­na, 388 p.
  31. Mil­ler, S. (2015). Mind: How to Bu­ild a Neu­ral Net­work. Ste­ven Mil­ler. Ret­ri­eved from: http://ste­ven­mil­ler888.git­hub.io/mind-how-to-bu­ild-a-neu­ral-net­work
  32. Moc­ha: simple, fle­xib­le, fun. (2021). Ret­ri­eved from: https://moc­hajs.org/.
  33. Mon­go, D. B. (2021). Bu­ild fas­ter. Bu­ild smar­ter. Ret­ri­eved from: https://www.mon­godb.com/.
  34. Ne­bor­si­na, N. P. (1997). The syntax of po­etic spe­ech as a sub­ject of lin­guo­po­etic re­se­arch (ba­sed on the ma­te­ri­al of English and Ame­ri­can po­etry of the 16th-20th cen­tu­ri­es). Doc­to­ral Dis­ser­ta­ti­on for Phi­lo­lo­gi­cal Sci­en­ces (10.02.04 – Ger­ma­nic Lan­gua­ges). Mos­cow sta­te un-t them. M. V. Lo­mo­no­sov, 356 p. [In Rus­si­an].
  35. Oli­vi­era H. G. (2016). Po­eTryMe: a ver­sa­ti­le plat­form for po­etry ge­ne­ra­ti­on. Hu­go Gon­ca­lo Oli­vi­era // CI­SUC, Uni­ver­sity of Co­imbra, Por­tu­gal. Ret­ri­eved from: https://eden.dei.uc.pt/ ~hro­liv/pubs/Gon­ca­lo­Oli­ve­ira2012_c3gi_CRC.pdf.
  36. Panchen­ko, T. V. (2007). Ge­ne­tic al­go­rithms. Astrak­han: Astrak­han Uni­ver­sity, 86 p. [In Rus­si­an].
  37. Po­em Ge­ne­ra­tor. (2021). Mas­ter­pi­ece Ge­ne­ra­tor. Ret­ri­eved from: https://www.po­em-ge­ne­ra­tor.org.uk/.
  38.  Po­em. (2019). Ne­og­ran­ka.ru. Ret­ri­eved from: http://ne­og­ran­ka.ru/ge­ne­ra­tor_sti­hov.html.
  39. Po­eTryMe. (2014). Uni­ver­sity of Co­imbra. Ret­ri­eved from: http://po­etryme.dei.uc.pt/.
  40. Ren, S., He, K., Girshick, R., & Sun, J. (2017). Fas­ter R-CNN: To­wards Re­al-Ti­me Ob­ject De­tec­ti­on with Re­gi­on Pro­po­sal Net­works. In IEEE Tran­sac­ti­ons on Pat­tern Analysis and Mac­hi­ne In­tel­li­gen­ce, 39(8), 1137–1149. https://doi.org/10.1109/TPA­MI.2016.2577031
  41. She­vel­yo­va-Gar­kus­ha, N. V. (2010). Se­man­tic and functi­onal fe­atu­res of rhythmic-syntac­tic or­ga­ni­za­ti­on of texts of mo­dern Ame­ri­can po­etry. Sci­en­ti­fic Bul­le­tin of VN Ka­ra­zin Khar­kiv Na­ti­onal Uni­ver­sity. Ser : Ro­ma­no-Ger­ma­nic phi­lo­logy, 896(61), 137–143. Ret­ri­eved from: http://eKhSU­IR.kspu.edu/handle/123456789/473. [In Uk­ra­ini­an].
  42. Sku­lac­he­va, T. V. (1989). On the qu­es­ti­on of the in­te­rac­ti­on of rhythm and syntax in a li­ne of po­etry (English and Rus­si­an iam­bic tet­ra­me­ter). Iz­ves­tia RAN. Li­te­ra­tu­re and Lan­gua­ge Se­ri­es, 48(2), 156–165. Mos­cow. [In Rus­si­an].
  43. Slov­nik.ua. (2005). Uk­ra­ini­an lan­gua­ge and cul­tu­re por­tal. Insti­tu­te of Lin­gu­is­tics. O. O. Po­tebny. Ret­ri­eved from: https://slovnyk.ua/. [In Uk­ra­ini­an].
  44. Stay­ko­va, Ka­men­ka. (15 Jul 2014). Na­tu­ral Lan­gua­ge Ge­ne­ra­ti­on and Se­man­tic Techno­lo­gi­es. Cyber­ne­tics and In­for­ma­ti­on Techno­lo­gi­es, 3–23. https://doi.org/10.2478/cait-2014-0015
  45. To­mas­hevsky, B. V. (2002). Li­te­ra­tu­re the­ory. Po­etics: Textbo­ok. al­lo­wan­ce. Mos­cow: As­pect Press, 334 p. [In Rus­si­an].
  46. Tynya­nov, Yu. N. (1993). Li­te­rary fact. Com­pi­led by O. I. No­vi­ko­va. Mos­cow: Hig­her. shk., 23–121. [In Rus­si­an].
  47. Tynya­nov, Yu. N. (1993). The prob­lem of po­etic lan­gua­ge. Babylon: Bul­le­tin of Yo­ung Li­te­ra­tu­re, 2(18), 86–90. Mos­cow: AR­GO-RISK. [In Rus­si­an].
  48. Vi­nog­ra­dov, V. V. (1975). From the his­tory of the study of po­etics (20s). Iz­ves­tia of the Aca­demy of Sci­en­ces of the USSR. Li­te­ra­tu­re and Lan­gua­ge Se­ri­es, 3, 259–272. Mos­cow. [In Rus­si­an].
  49. Wang, J., & Hu, S. (2002). Glo­bal asympto­tic sta­bi­lity and glo­bal ex­po­nen­ti­al sta­bi­lity of con­tin­uo­us-ti­me re­cur­rent neu­ral net­works. IEEE Trans. Au­to­ma­tic Control, 47, 802–807. https://doi.org/10.1109/TAC.2002.1000277
  50. Wil­li­ams, R. J., & Zip­ser, D. (1989). A Le­ar­ning Al­go­rithm for Con­tin­ually Run­ning Fully Re­cur­rent Neu­ral Net­works. Neu­ral Com­pu­ta­ti­on, 1, 270–280. https://doi.org/10.1162/ne­co.1989.1.2.270
  51. Yu­anzhi, Ke, & Ma­sa­fu­mi, Ha­gi­wa­ra. (03 May 2017). An English Neu­ral Net­work that Le­arns Texts, Finds Hid­den Know­led­ge, and Answers Qu­es­ti­ons. Jo­ur­nal of Ar­ti­fi­ci­al In­tel­li­gen­ce and Soft Com­pu­ting Re­se­arch, 229–242. https://doi.org/10.1515/jaiscr-2017-0016
  52. Zak­ha­rov, Vic­tor. (21 Dec 2019). Ways of Au­to­ma­tic Iden­ti­fi­ca­ti­on of Words Be­lon­ging to Se­man­tic Fi­eld. Jo­ur­nal of Lin­gu­is­tics, 234–243. https://doi.org/10.2478/jazcas-2019-0054