FI-20245939-A1 - Computer-implemented method for processing a request

FI20245939A1FI 20245939 A1FI20245939 A1FI 20245939A1FI-20245939-A1

Abstract

According to an embodiment, a computer-implemented method (100) for processing a request comprises: obtaining (101) a plurality of reference requests defining a request class; vectorizing (102) the plurality of reference requests, thus obtaining a plurality of reference request vectors; obtaining (103) a request; vectorizing (104) the request, thus obtaining a request vector; determining (105) whether the request belongs to the request class by comparing the plurality of reference request vectors and the request vector; and processing (106) the request based at least on whether the request belongs to the request class.

Inventors

RUUTU VILLE
RUUTU JUSSI

Assignees

ELISA OYJ

Dates

Publication Date: 20260131
Application Date: 20240730
Priority Date: 20240730

Claims (11)

1. A computer-implemented method (100) for processing a request, the method (100) comprising: obtaining (101) a plurality of reference re- quests defining a request class; vectorizing (102) the plurality of reference requests, thus obtaining a plurality of reference re- quest vectors; obtaining (103) a request; vectorizing (104) the request, thus obtaining a request vector; determining (105) whether the request belongs to the request class by comparing the plurality of ref- erence request vectors and the request vector; and processing (106) the request based at least on whether the request belongs to the request class.
2. The computer-implemented method (100) ac- cording to claim 1, wherein the comparing the plurality of reference request vectors and the request vector com- + prises calculating at least one vector space distance & between the plurality of reference request vectors and S the request vector. a z 25
3. The computer-implemented method (100) ac- > cording to claim 2, wherein the calculating the at least & one vector space distance between the plurality of ref- N erence request vectors and the request vector comprises N calculating the at least one vector space distance using cosine similarity.
4. The computer-implemented method (100) ac- cording to claim 2 or claim 3, wherein the calculating the at least one vector space distance between the plu- rality of reference request vectors and the request vec- tor comprises calculating a summary reference request vector based on the plurality of reference request vec- tors and calculating a vector space distance between the summary reference request vector and the request vector.
5. The computer-implemented method (100) ac- cording to claim 4, wherein the summary reference re- quest vector comprises a mean reference request vector and/or an average reference request vector.
6. The computer-implemented method (100) ac- cording to claim 2 or claim 3, wherein the calculating the at least one vector space distance between the plu- rality of reference request vectors and the request vec- N tor comprises calculating a plurality of vector space N distances between the request vector and the plurality = of reference reguest vectors and calculating an average ? 25 or a mean of the plurality of vector space distances a between the request vector and the plurality of refer- 8 ence request vectors. 0 S
N 7. The computer-implemented method (100) ac- cording to any of claims 2 - 6, wherein the determining whether the request belongs to the request class by comparing the plurality of reference request vectors and the request vector comprises: obtaining a threshold vector space distance; in response to the at least one vector space distance between the plurality of reference request vec- tors and the request vector being greater than the threshold vector space distance, determining that the request does not belong to the request class; and in response to the at least one vector space distance between the plurality of reference request vec- tors and the request vector being less than the thresh- old vector space distance, determining that the request belongs to the request class.
8. The computer-implemented method (100) ac- cording to any preceding claim, wherein the request class corresponds to escalation requests, the request corresponds to a request obtained from a user during a chat conversation between the user and a chatbot, and the method further comprises, in response to determining N that the request belong to the request class, performing N escalation processing based on the request. O 3 - 25
9. The computer-implemented method (100) ac- o cording to claim 8, wherein the escalation processing 8 comprises escalating the request to a human. 0 S N
10.A computing device (600), comprising at least one processor (601) and at least one memory (602) including computer program code, the at least one memory (602) and the computer program code configured to, with the at least one processor (601), cause the computing device (600) to perform the method (100) according to any preceding claim.
11. A computer program product comprising pro- gram code configured to perform the method (100) ac- cording to any of claims 1 - 9 when the computer program product is executed on a computer. <t N O N K <Q Oo O I a a o O O 0 <t N O N

Description

COMPUTER-IMPLEMENTED METHOD FOR PROCESSING A REQUEST TECHNICAL FIELD [0001] The present disclosure relates to request pro- cessing, and more particularly to a computer-implemented method for processing a request, a computing device, and a computer program product. BACKGROUND [0002] Classification of requests provided by a user can be useful in various application, such as with so- called chatbots and/or voicebots. Such classification can be performed in various ways. However, many solu- tions have technical drawbacks and challenges, such as computational expense, the need to have large amounts of training data, and/or the solution not being robust enough to reliably classify all cases. SUMMARY + 20 [0003] This summary is provided to introduce a selec- N S tion of concepts in a simplified form that are further 5 described below in the detailed description. This sum- 3 mary is not intended to identify key features or essen- = tial features of the claimed subject matter, nor is it a O 25 intended to be used to limit the scope of the claimed O 3 subject matter. <t S [0004] It is an objective to provide a computer-im- plemented method for processing a request, a computing device, and a computer program product. The foregoing and other objectives are achieved by the features of the independent claims. Further implementation forms are apparent from the dependent claims, the description and the figures. [0005] According to a first aspect, a computer-imple- mented method for processing a request comprises: ob- taining a plurality of reference requests defining a request class; vectorizing the plurality of reference requests, thus obtaining a plurality of reference re- quest vectors; obtaining a request; vectorizing the re- quest, thus obtaining a request vector; determining whether the request belongs to the request class by comparing the plurality of reference request vectors and the request vector; and processing the request based at least on whether the request belongs to the request class. [0006] In an implementation form of the first aspect, the comparing the plurality of reference request vectors and the request vector comprises calculating at least < one vector space distance between the plurality of ref- S erence reguest vectors and the reguest vector. S [0007] In another implementation form of the first S aspect, the calculating the at least one vector space I 25 distance between the plurality of reference request vec- > tors and the request vector comprises calculating the & at least one vector space distance using cosine simi- 3 larity. [0008] In another implementation form of the first aspect, the calculating the at least one vector space distance between the plurality of reference request vec- tors and the request vector comprises calculating a sum- mary reference request vector based on the plurality of reference request vectors and calculating a vector space distance between the summary reference request vector and the request vector. [0009] In another implementation form of the first aspect, the summary reference request vector comprises a mean reference request vector and/or an average ref- erence request vector. [0010] In another implementation form of the first aspect, the calculating the at least one vector space distance between the plurality of reference request vec- tors and the request vector comprises calculating a plu- rality of vector space distances between the request vector and the plurality of reference request vectors and calculating an average or a mean of the plurality of vector space distances between the request vector and + the plurality of reference reguest vectors. & [0011] In another implementation form of the first S aspect, the determining whether the request belongs to S the request class by comparing the plurality of refer- = 25 ence request vectors and the request vector comprises: > obtaining a threshold vector space distance; in re- & sponse to the at least one vector space distance between N the plurality of reference request vectors and the re- N guest vector being greater than the threshold vector space distance, determining that the request does not belong to the request class; and in response to the at least one vector space distance between the plurality of reference request vectors and the request vector be- ing less than the threshold vector space distance, de- termining that the request belongs to the request class. [0012] In another implementation form of the first aspect, the request class corresponds to escalation re- quests, the request corresponds to a request obtained from a user during a chat conversation between the user and a chatbot, and the method further comprises, in response to determining that the request belong to the request class, performing escalation processing based on the request. [0013] In another implementation form of the first aspect, the escalation processing comprises escalating the request to a human. [0014] According to a second as