Computer-implemented method for providing a response to a user

Abstract

According to an embodiment, a computer-implemented method (100) for providing a response to a user comprises: obtaining (101) speech data from a user; performing (102) sentiment analysis at least on the speech data, thus obtaining a sentiment of the speech data; performing (103) context analysis, thus obtaining a context of the speech data; obtaining (104) a selection matrix, wherein each element of the selection matrix corresponds to a tone of a response and a content of a response; selecting (105) an element of the selection matrix based at least on the sentiment of the speech data and the context of the speech data; and providing (106) a response to the user based on the selected element of the selection matrix.

Description

COMPUTER- IMPLEMENTED METHOD FOR PROVIDING A RESPONSE TO A USERTECHNICAL FIELD

[0001] The present disclosure relates to speech processing, and more particularly to a computer-implemented method for providing a response to a user, a computing device , and a computer program product .BACKGROUND

[0002] When an automated system, such as a so-called voicebot , interacts with a user via speech, the system may need to be able to handle different types of requests and provide appropriate responses . Further, the user may be speaking, such as asking questions or making orders , with various tones of voice .SUMMARY

[0003] This summary is provided to introduce a selection of concepts in a s implif ied form that are further described below in the detailed description . This summary is not intended to identify key features or essential features of the claimed subj ect matter, nor is it intended to be used to limit the scope of the claimed subj ect matter .

[0004] It is an obj ective to provide a computer-implemented method for providing a response to a user, a computing device , and a computer program product . Theforegoing and other obj ectives 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-implemented method for processing a request comprises : A computer-implemented method for providing a response to a user, the method comprising : obtaining speech data from a user ; performing sentiment analysis at least on the speech data, thus obtaining a sentiment of the speech data ; performing context analysis , thus obtaining a context of the speech data ; obtaining a selection matrix, wherein each element of the selection matrix corresponds to a tone of a response and a content of a response ; selecting an element of the selection matrix based at least on the sentiment of the speech data and the context of the speech data ; and providing a response to the user based on the selected element of the selection matrix .

[0006] In an implementation form of the first aspect , the performing sentiment analysis at least on the speech data comprises : performing a speech-to-text conversion on the speech data thus obtaining text data and performing sentiment analysis on the text data ; and / or performing sentiment analysis directly on the speech data .

[0007] In another implementation form of the first aspect , the performing sentiment analysis at least on the speech data further comprises : performing sentiment analysis at least on one speech data previously obtained from the user ; and obtaining a total sentiment based onthe sentiment of the speech data and the sentiment of the at least one speech data previously obtained from the user ; and the selecting the element of the selection matrix based at least on the sentiment of the speech data and the context of the speech data comprises selecting the element of the selection matrix based at least on the total sentiment and the context of the speech data .

[0008] In another implementation form of the first aspect , the obtaining the speech data from the user comprises continuously obtaining speech data from the user and the performing sentiment analysis at least on the speech data further comprises continuously performing sentiment analysis on the continuously obtained speech data .

[0009] In another implementation form of the first aspect , the performing the context analysis comprises : obtaining context information related to the speech data and performing the context analysis based at least on the context information ; performing context analysis at least on the speech data ; and / or obtaining history information related to the speech data and performing the context analysis based at least on the history information .

[0010] In another implementation form of the first aspect , the performing context analysis at least on the speech data comprises : performing context analysis based on at least one keyword in the speech data; and / or performing context analysis using a machine learning model .

[0011] In another implementation form of the first aspect , the content of the response comprises a prerecorded audio response or data for generating an audio response .

[0012] In another implementation form of the first aspect , the providing the response to the user based on the selected element of the selection matrix comprises generating the response using a generative artificial intelligence model .

[0013] In another implementation form of the first aspect , the sentiment of the speech data is represented by a floating point number .

[0014] In another implementation form of the first aspect , the obtaining speech data from the user comprises obtaining the speech data via a voice call and the providing the response to the user comprises providing the response to the user via the voice call .

[0015] According to a second aspect , a computing device compri ses at least one processor and at least one memory including computer program code , the at least one memory and the computer program code being configured to , with the at least one proces sor, cause the computing device to perform the method according to the first aspect .

[0016] According to a third aspect , a computer program product comprises program code configured to perform the method according to the first aspect when the computer program product is executed on a computer .

[0017] Many of the attendant features wil l be more readily appreciated as they become better understood by reference to the following detailed description considered in connection with the accompanying drawings .DESCRIPTION OF THE DRAWINGS

[0018] In the following, example embodiments are described in more detail with reference to the attached figures and drawings , in which :

[0019] Fig . 1 illustrates a flow chart representation of a method according to an embodiment ;

[0020] Fig . 2 illustrates a flow chart representation of a method according to another embodiment ;

[0021] Fig . 3 illustrates a schematic representation of a selection matrix according to an embodiment ;

[0022] Fig . 4 illustrates a schematic representation of a selection matrix according to another embodiment ; and

[0023] Fig . 5 illustrates a schematic representation of a computing device according to an embodiment .

[0024] In the following, like reference numerals are used to des ignate li ke parts in the accompanying drawings .DETAILED DESCRIPTION

[0025] In the following description, reference is made to the accompanying drawings , which form part of thedisclosure , and in which are shown, by way of illustration, specific aspects in which the present disclosure may be placed . It is understood that other aspects may be utilised, and structural or logical changes may be made without departing from the scope of the present disclosure . The following detailed description, therefore , is not to be taken in a limiting sense , as the scope of the present disclosure is defined by the appended claims .

[0026] For instance , it is understood that a disclosure in connection with a described method may also hold true for a corresponding device or system configured to perform the method and vice versa . For example , if a specific method step is described, a corresponding device may include a unit to perform the described method step, even if such unit is not explicitly described or il lustrated in the f igures . On the other hand, for example , if a specific apparatus is described based on functional units , a corresponding method may include a step performing the described functionality, even if such step is not explicitly described or illustrated in the figures . Further, it is understood that the features of the various example aspects described herein may be combined with each other, unless specifically noted otherwise .

[0027] Fig . 1 illustrates a flow chart representation of a method according to an embodiment .

[0028] According to an embodiment , a computer-implemented method 100 for providing a response to a user comprises obtaining 101 speech data from a user .

[0029] Herein, the speech data may comprise audio data comprising speech, such as speech of the user . For example , the speech data may compri se a question uttered by the user .

[0030] The obtaining 101 speech data from the user may comprise , for example , obtaining the speech data via a telecommunications network, such as over the internet , or via a telephone network . For example, the method 100 may be performed as a part of a so-called voicebot running on a server . The voicebot can communicate with the user via speech over a voice call , such as a call in the telephone network or a voice over internet protocol (VoI P) .

[0031] The method 100 may further comprise performing102 sentiment analysis at least on the speech data, thus obtaining a sentiment of the speech data .

[0032] The sentiment analysis may be performed based at least on the speech data . In other embodiments , other data may be used in addition to the speech data and / or the speech data may be processed, such as pre-processed, before performing the sentiment analysis .

[0033] The method 100 may further comprise performing103 context analysis , thus obtaining a context of the speech data .

[0034] The context analysis may be performed based at least on the speech data and / or other data . In otherembodiments , other data may be used in addition or al ternatively to the speech data and / or the speech data may be processed, such as pre-proces sed, before performing the context analysis .

[0035] The method 100 may further comprise obtaining104 a selection matrix, wherein each element of the selection matrix corresponds to a tone of a response and a content of a response .

[0036] The obtaining 104 the selection matrix may comprise , for example , obtaining the selection matrix from a database .

[0037] The tone of the response may define the tone voice for providing the response to the user .

[0038] The content of the response may define the content of the response to be provided to the user .

[0039] The method 100 may further comprise selecting105 an element of the selection matrix based at least on the sentiment of the speech data and the context of the speech data .

[0040] Each element of the selection matrix may correspond to a specific sentiment and / or context of the speech data . For example , each row of the selection matrix may correspond to a specific context or a range / set of contexts of the speech data and each column of the selection matrix may correspond to a specific sentiment or a range / set of sentiments of the speech data or vice versa .

[0041] The method 100 may further comprise providing 106 a response to the user based on the selected element of the selection matrix .

[0042] In some embodiments , the response can be generated based on the selected element of the selection matrix . In other embodiments , the selected element of the selection matrix may comprise the response as , for example , audio data .

[0043] It should be appreciated that at least some of the operations of the method 100 may be performed in various orders . For example, in some embodiments , operation 104 may be performed before operations 101 - 103 and / or operations 102 and 103 may be performed in any order . In some embodiments , operation 104 may be performed well in advance of other operations of the method 100 . For example , the selection matrix may be configured in advance and may be saved to a database or similar . Then, when the speech data is obtained 101 , the selection matrix may be obtained from the database and utili zed for performing the operations of the method 100 .

[0044] The method 100 can, for example, enable voice- bots and similar systems to provide more appropriate responses to users .

[0045] The method 100 can, for example , improve the user experience with voicebots and similar systems .

[0046] Fig . 2 illustrates a flow chart representation of a method according to another embodiment .

[0047] In the embodiment of Fig. 2, the obtaining 104 the selection matrix comprises forming a selection matrix 201, forming 202 the content of each element of the selection matrix, and forming 203 the tone of each element of the selection matrix.

[0048] The selection matrix can be formed based on the use case, since different use cases can have different contexts, tones for the response, and / or contents for the response. For example, in a customer service use case, the user may be presenting a complaint, placing an order, reporting a fault / f allure, or making an inquiry. In other use cases, the contexts may be different .

[0049] The forming 202 the content of the selection matrix may comprise, for example, generating pre-recorded audio responses or data for generating audio responses. The data for generating audio responses may comprise, for example, a part of a prompt to be provided to a generative artificial intelligence model.

[0050] The forming 203 the tone of the selection matrix may comprise, for example, forming data that can be used to configure the tone of voice of the response. For example, a speech synthesis markup language (SSML) script can be formed.

[0051] Since the operations 201 - 203 can form the selection matrix and the content and sentiment of the elements of the selection matrix, after operations 201 - 203 have been performed, the selection matrix can be, for example, saved in memory, such as in a database. Theselection matrix may then be reused later without needing to perform operations 201 - 203 . Thus , later the obtaining 104 the selection matrix may comprise obtaining the selection matrix from, for example , memory .

[0052] According to an embodiment , the performing sentiment analysis at least on the speech data compri ses : performing a speech-to-text conversion on the speech data thus obtaining text data and performing sentiment analysis on the text data ; and / or performing sentiment analysis directly on the speech data .

[0053] For example , in the embodiment of Fig . 2 , operation 204 comprises the obtaining 101 speech data from a user and converting the speech data to text data using a speech-to-text conversion .

[0054] In the embodiment of Fig . 2 , in operation 205 it can be checked whether the conversion continues . I f the conversion continues , new speech data can be obtained from the user in operation 204 . I f the conversation does not continue , the procedure can end at operation 206 .

[0055] According to an embodiment , the performing 102 sentiment analysis at least on the speech data further comprises : performing sentiment analysis at least on one speech data previously obtained from the user ; and obtaining a total sentiment based on the sentiment of the speech data and the sentiment of the at least one speech data previously obtained from the user ; and the selecting the element of the selection matrix based at least on the sentiment of the speech data and the context ofthe speech data comprises selecting the element of the selection matrix based at least on the total sentiment and the context of the speech data .

[0056] The speech data previously obtained from the user may comprise , for example , speech data previously obtained from the user during the voice call , such as what the user has said previously during the voice call . Thus , by obtaining the total sentiment based on the sentiment of the speech data and the sentiment of the at least one speech data previously obtained from the user and selecting the element of the selection matrix based at least on the total sentiment and the context of the speech data, both what the user has said previ ously during the voice call and what the user is saying currently can be used to assess the sentiment of the user and provide an appropriate response .

[0057] According to an embodiment , the obtaining 101 the speech data from the user comprises continuously obtaining speech data from the user and the performing sentiment analysis at least on the speech data further comprises continuously performing sentiment analysis on the continuously obtained speech data .

[0058] The sentiment of continuously obtained speech data can be quantified by, for example , calculating a moving average over the sentiment obtained via the continuously performed sentiment analysis . The moving average can be calculated over any number of obtained speech data . Thus , the sentiment of the user may beassessed in a continuous fashion and an appropriate response can be provided . Alternatively, the sentiment of continuously obtained speech data can be quantified using some other technique , such as a Kalman filter .

[0059] According to an embodiment , the performing 103 the context analysis comprises : obtaining context information related to the speech data and performing the context analysis based at least on the context information ; performing context analysis at least on the speech data ; and / or obtaining history information related to the speech data and performing the context analysis based at least on the history information .

[0060] The context information related to the speech data may comprise , for example , data indicating which phone number the user has called to initiate the voice call . For example , i f the user has cal led a number for placing orders the context of the voice call may be different than if the user has called a number for making failure reports .

[0061] In some embodiments , alternatively or additionally, the context information may be obtained also from other resources , such as external resources . For example , the context information may comprise information about the user . The information about the user may be obtained from, for example , a customer relationship management (CRM) system .

[0062] Alternatively or additionally, context information may be obtained from a ticketing system . I f the user is asking about an error / f al lure with a system, theticketing system may indicate, for example, if the error is already in the ticketing system and what is the status of the ticket.

[0063] The performing context analysis at least on the speech data may comprise, for example, identifying the matter the user is calling about based on the speech data. For example, the speech data may be converted to text data using a speech-to-text conversion and the matter the user is calling about may be identified based on the text data using, for example, the techniques disclosed herein.

[0064] In some cases, the user may be calling about more than one matter. For example, the user may have a question about a service and may also want to place an order. In some embodiments, the performing context analysis at least on the speech data may comprise, for example, identifying a plurality of matters the user is calling about based on the speech data.

[0065] The history information related to the speech data may comprise, for example, data indicating the history of the voice call. For example, the history information may indicate if the user has talked with a human agent during the voice call, how long the user has had to wait, and / or to what services the voice call has been connected to.

[0066] According to an embodiment, the performing context analysis at least on the speech data comprises:performing context analysis based on at least one keyword in the speech data; and / or performing context analysis using a machine learning model.

[0067] The performing context analysis using a machine learning model may comprise, for example, performing the context analysis using machine learning classification or a generative artificial intelligence model.

[0068] The performing context analysis based on at least one keyword in the speech data may comprise, for example, identifying at least one preconfigured keyword in the speech data. Each keyword may correspond to a context. For example, if the speech data comprises words that are typically associated with placing an order, the context analysis can determine that the context of the speech data is placing an order.

[0069] According to an embodiment, the sentiment of the speech data is represented by a floating point number .

[0070] For example, a value of 0 - 0.4 may indicate a negative sentiment, a value of 0.4 - 0.7 may indicate a neutral sentiment, and a value of 0.7 - 1.0 may indicate a negative sentiment. The scale of the sentiment may not be linear and / or the sentiment may be scaled in other fashions. For example, in other embodiments, the sentiment may be represented by an integer value between 0 and 100, an integer value between 0 and 255, or by a floating-point value or an integer value in any other interval .

[0071] The sentiment analysis may be performed, for example , using a third-party service .

[0072] According to an embodiment , the obtaining 101 speech data from the user comprises obtaining the speech data via a voice call and the providing 106 the response to the user comprises providing the response to the user via the voice call .

[0073] Fig . 3 illustrates a schematic representation of a selection matrix according to an embodiment .

[0074] In the embodiment of Fig . 3 , an example of how an element of the selection matrix can be selected based on the context of the speech data and the sentiment of the speech data is illustrated . In the embodiment of Fig . 3 , each row of the selection matrix 300 corresponds to a different context and each column corresponds to a different sentiment .

[0075] Each element in the selection matrix 300 can correspond to a tone of the response and a content of the response .

[0076] According to an embodiment , the content of the response comprises a pre-recorded audio response or data for generating an audio response .

[0077] The data for generating an audio response may be used to generate the response . For example , the data for generating the audio response may comprise text and the response can be generated using a text-to-speech conversion .

[0078] In some embodiments , the data for generating the audio response may comprise a prompt or a part of aprompt that can be provided to a machine learning model , such as a large language model (LLM) , to generate the response . For example , the prompt may comprise a system section and the system section may be used to instruct the LLM to provide a response to the audio data provided by the user or a response to text data converted from the audio data . Alternatively or additionally, the content obtained from the selection matrix may be used to instruct the LLM to provide the response in specific form . For example , the selected element of the selection matrix may indicate that the response should be formulated in a specific way, such as in an apologi zing manner, informative manner, or enthusiastic manner .

[0079] According to an embodiment , the providing 106 the response to the user based on the selected element of the selection matrix comprises generating the response using a generative artificial intelligence model .

[0080] The tone of the response may be used to configure the tone of voice of the response . For example , in some embodiments , the response may be generated using a generative artificial intelligence model and the tone and the content may be provided to the model in a prompt . In other embodiments , the tone of voice of the text-to- speech conversion may be configurable and can be configured according to the tone indicated by the selection matrix .

[0081] Fig . 4 illustrates a schematic representation of a selection matrix according to another embodiment .

[0082] In the embodiment of Fig. 4, an example of how an element of the selection matrix can be selected based on the context of the speech data and the sentiment of the speech data is illustrated. In the embodiment of Fig. 4, each row of the selection matrix 300 corresponds to a different context in the set of making a complaint, reporting a failure, making an inquiry, and placing an order and each column corresponds to a different sentiment in the set of negative, neutral, and positive.

[0083] Each element in the selection matrix 300 can correspond to a tone of the response and a content of the response. For example, in the embodiment of Fig. 4, tone 1 may be empathetic, tone 3 can be friendly and helpful, and tone 5 can be cheerful. Tone 2 and tone 4 may correspond to other tones of voice.

[0084] For example, if the sentiment of the speech data corresponds to a value of 0.2, i.e. negative, and the context of the speech data correspond to a complaint, the tone of the response can be empathetic (tone 1) and the content of the response can be apologizing (content 1 ) .

[0085] For example, if the sentiment of the speech data corresponds to a value of 0.5, i.e. neutral, and the context of the speech data correspond to an inquiry, the tone of the response can be friendly and helpful (tone 3) and the content of the response can be informative (content 3) .

[0086] For example, if the sentiment of the speech data corresponds to a value of 0.8, i.e. positive, andthe context of the speech data correspond to placing an order, the tone of the response can be friendly and cheerful ( tone 5 ) and the content of the response can be enthusiastic ( content 5 ) .

[0087] For example , the audio data may comprise audio of the user making a complaint that a specific service does not work . An LLM may be instructed to provide a brief response asking for further information in an apologi zing manner .

[0088] For example , the audio data may comprise audio of the user requesting to place and order . An LLM may be instructed to provide a brief response asking for further information in a neutral manner or in an enthusiastic manner depending on the sentiment of the audio data .

[0089] The examples disclosed herein are only limited examples of the functionality of the selection matrix . In other embodiments , the selection matrix may be of any si ze and each element of the selection matrix may comprise any combination of a tone of a response and a content of a response .

[0090] The response can be formulated by, for example , providing the tone and the context to a large language model (LLM) . For example the prompt provided to the LLM can instruct the LLM to formulate the response according to the tone of the response and the context of the response obtained from the selection matrix .

[0091] Alternatively or additionally, the tone of the response can be used in a text-to-speech conversion . Forexample, the selection matrix may comprise each content in a text form . The content can be converted to speech using a text-to-speech conversion and the tone may be provided to the text-to-speech conversion to define the tone of voice desired for the response .

[0092] In some embodiments , the content of the response indicated by the selection matrix may indicate how the response should be formulated without defining the exact content of the response . For example , the content of the response indicated by the selection matrix may comprise a prompt or a part of a prompt defining how an LLM should formulate the response . Other content of the response , such as an answer to question presented by the user, can be provided by the LLM. In some embodiments , the LLM can utili zed, for example , retrieval- augmented generation (RAG) in order to answer the question presented by the user . For example , if the user presents a question about how to use a specific ser- vice / device , the LLM can answer the question using data about the service / device , such as a user manual , via RAG and the content of the response indicated by the selection matrix can define how the response should be formulated by the LLM .

[0093] Fig . 5 illustrates a schematic representation of a computing device according to an embodiment .

[0094] According to an embodiment , a computing device 500 comprises at least one processor 501 and at least one memory 502 including computer program code , the atleast one memory 502 and the computer program code configured to, with the at least one processor 501, cause the computing device 500 to perform the method 100.

[0095] The computing device 500 may comprise at least one processor 501. The at least one processor 501 may comprise, for example, one or more of various processing devices, such as a co-processor, a microprocessor, a digital signal processor (DSP) , a processing circuitry with or without an accompanying DSP, or various other processing devices including integrated circuits such as, for example, an application specific integrated circuit (ASIC) , a field programmable gate array (FPGA) , a microprocessor unit (MCU) , a hardware accelerator, a special-purpose computer chip, or the like.

[0096] The computing device 500 may further comprise a memory 502. The memory 502 may be configured to store, for example, computer programs and the like. The memory 502 may comprise one or more volatile memory devices, one or more non-volatile memory devices, and / or a combination of one or more volatile memory devices and nonvolatile memory devices. For example, the memory 502 may be embodied as magnetic storage devices (such as hard disk drives, magnetic tapes, etc.) , optical magnetic storage devices, and semiconductor memories (such as mask ROM, PROM (programmable ROM) , EPROM (erasable PROM) , flash ROM, RAM (random access memory) , etc.) .

[0097] The computing device 500 may further comprise other components not illustrated in the embodiment ofFig . 5 . The computing device 500 may comprise , for example , an input / output bus for connecting the computing device 500 to other devices . Further, a user may control the computing device 500 via the input / output bus .

[0098] When the computing device 500 is configured to implement some functionality, some component and / or components of the computing device 500 , such as the at least one processor 501 and / or the memory 502 , may be configured to implement this functionality . Furthermore , when the at least one processor 501 is configured to implement some functionality, this functionality may be implemented using program code comprised, for example , in the memory .

[0099] The computing device 500 may be implemented at least partially using, for example , a computer, some other computing device , or similar .

[0100] The method 100 and / or the computing device 500 may be utilised in, for example, voicebot applications . A voicebot may be configured to obtain information from users by, for example , phone and convert the voice information into text information using automatic speech recognition (ASR) . The voicebot may be , for example , running on a server of a company and be provided as a service to the clients of the company . The method 100 can be used to provide appropriate responses to the users .

[0101] Any range or device value given herein may be extended or altered without losing the effect sought .Also any embodiment may be combined with another embodiment unless explicitly disallowed .

[0102] Although the subj ect matter has been described in language specific to structural features and / or acts , it is to be understood that the subj ect matter defined in the appended claims is not necessarily limited to the specific features or acts described above . Rather, the specific features and acts described above are disclosed as examples of implementing the claims and other equivalent features and acts are intended to be within the scope of the claims .

[0103] It will be understood that the benefits and advantages described above may relate to one embodiment or may relate to several embodiments . The embodiments are not limited to those that solve any or all of the stated problems or those that have any or all of the stated benefits and advantages . It wil l further be understood that reference to ' an ' item may refer to one or more of those items .

[0104] The steps of the methods described herein may be carried out in any suitable order, or simultaneously where appropriate . Additionally, individual blocks may be deleted from any of the methods without departing from the spirit and scope of the subj ect matter described herein . Aspects of any of the embodiments described above may be combined with aspects of any of the other embodiments described to form further embodiments without losing the effect sought .

[0105] The term ' comprising ' is used herein to mean including the method, blocks or elements identified, but that such blocks or elements do not comprise an exclusive list and a method or apparatus may contain addi- tional blocks or elements .

[0106] It will be understood that the above description is given by way of example only and that various modif ications may be made by those ski lled in the art . The above specification, examples and data provide a complete description of the structure and use of exemplary embodiments . Although various embodiments have been described above with a certain degree of particularity, or with reference to one or more individual embodiments , those skilled in the art could make numer- ous alterations to the disclosed embodiments without departing from the spirit or scope of this specification .

Claims

CLAIMS :

1. A computer-implemented method (100) for providing a response to a user, the method (100) comprising : obtaining (101) speech data from a user; performing (102) sentiment analysis at least on the speech data, thus obtaining a sentiment of the speech data; performing (103) context analysis, thus obtaining a context of the speech data; obtaining (104) a selection matrix, wherein each element of the selection matrix corresponds to a tone of a response and a content of a response; selecting (105) an element of the selection matrix based at least on the sentiment of the speech data and the context of the speech data; and providing (106) a response to the user based on the selected element of the selection matrix.

2. The computer-implemented method (100) according to claim 1, wherein the performing (102) sentiment analysis at least on the speech data comprises: performing a speech-to-text conversion on the speech data thus obtaining text data and performing sentiment analysis on the text data; and / or performing sentiment analysis directly on the speech data.3 . The computer-implemented method ( 100 ) according to claim 1 or claim 2 , wherein the performing( 102 ) sentiment analysis at least on the speech data further comprises : performing sentiment analysis at least on one speech data previously obtained from the user ; and obtaining a total sentiment based on the sentiment of the speech data and the sentiment of the at least one speech data previously obtained from the user ; and the selecting the element of the selection matrix based at least on the sentiment of the speech data and the context of the speech data comprises selecting the element of the selection matrix based at least on the total sentiment and the context of the speech data .4 . The computer-implemented method ( 100 ) according to any preceding claim, wherein the obtaining ( 101 ) the speech data from the user comprises continuously obtaining speech data from the user and the performing ( 102 ) sentiment analysis at least on the speech data further comprises continuously performing sentiment analysis on the continuously obtained speech data .5 . The computer-implemented method ( 100 ) according to any preceding claim, wherein the performing( 103 ) the context analysis comprises :obtaining context information related to the speech data and performing the context analysis based at least on the context information ; performing context analysis at least on the speech data ; and / or obtaining history information related to the speech data and performing the context analysis based at least on the history information .6 . The computer-implemented method ( 100 ) according to claim 5 , wherein the performing ( 103 ) context analysis at least on the speech data comprises : performing context analysis based on at least one keyword in the speech data ; and / or performing context analysis using a machine learning model .7 . The computer-implemented method ( 100 ) according to any preceding claim, wherein the content of the response comprises a pre-recorded audio response or data for generating an audio response .8 . The computer-implemented method ( 100 ) according to any preceding claim, wherein the providing the response to the user based on the selected element of the selection matrix comprises generating the response using a generative artificial intelligence model .9 . The computer-implemented method ( 100 ) according to any preceding claim, wherein the sentimentof the speech data is represented by a floating point number .

10. The computer-implemented method (100) according to any preceding claim, wherein the obtaining (101) speech data from the user comprises obtaining the speech data via a voice call and the providing (106) the response to the user comprises providing the response to the user via the voice call.

11. A computing device (500) , comprising at least one processor (501) and at least one memory (502) including computer program code, the at least one memory (502) and the computer program code configured to, with the at least one processor (501) , cause the computing device (500) to perform the method (100) according to any preceding claim.

12. A computer program product comprising program code configured to perform the method (100) according to any of claims 1 - 10 when the computer program product is executed on a computer.

Images