Method for analysing a product and system for analysing a product

Abstract

This invention relates to a method for analysing a product (P), the method comprising the following steps: via a server computer (3), receiving input data (23) through an input interface (21); acquiring one or more texts (R1, R2), representing reviews relating to the product (P), based on the input data (21); via a natural language processing (NLP) engine, processing the texts (R1, R2) so as to extract one or more opinions (31', 31'', 31''') on the product (P); via the NLP engine, generating one or more categories (32', 32'', 32''') representing recurring themes in the opinions (31', 31'', 31''') on the product; defining a match between the opinions (31', 31'', 31''') extracted and the categories (32', 32'', 32''') generated; for each category (32', 32'', 32''') generated, deriving summary data (33', 33'', 33''') representing the opinions (31', 31'', 31''') extracted and pertinent to the category (32', 32'', 32'''); and storing to a server memory a categories vector (32) containing the categories (32', 32'', 32''') generated, and a summary data vector (33) containing the summary data (33', 33'', 33''') pertinent to the categories (32', 32'', 32''') contained in the categories vector (32), in order to provide an operator (O) with the categories vector (32) and with the summary data vector (33).

Description

[0001] DESCRIPTION

[0002] METHOD FOR ANALYSING A PRODUCT AND SYSTEM FOR ANALYSING A PRODUCT

[0003] Technical field

[0004] This invention relates to a method for analysing a product and to a system for analysing a product.

[0005] In particular, this invention relates to a method, and a related system, for analysing data and information pertaining to a product so as to guide an operator - for example, a designer or engineer of an R&D department - in the process of revising (or improving) the product.

[0006] Background art

[0007] Known in the prior art of the research and development field are several processes for generating a product. Some processes involve, for example, the steps of defining, designing and producing the product. Typically, once the product has been produced, it is placed on the market. Furthermore, after being placed on the market, the product may be subjected to revision intended to improve or upgrade it. The purpose of product revision may be to eliminate or at least reduce any design errors. Another purpose of revision may be to improve or upgrade the product so as to augment or maintain its value over time. Product revision, i.e. improvement or upgrading, may be "on demand", that is, it may be carried out during particular stages of product life, or it may follow certain events. Furthermore, product revision may occur periodically, according to what is termed "ongoing improvement" of the product. Hereinafter, the terms "improvement", "upgrading" and "revision" with reference to the product may be used synonymously.

[0008] In practice, however, the revision of a product, whether "on demand" or "ongoing" is not always achieved and this may lead to the product failing or becoming obsolete in the market, resulting in loss of value. One of the reasons why product revision may not be achieved is that the process of improving a product is an onerous task carried out, for example, by designers, R&D engineers and / or sales operators, and involves considerable expenditure in terms of both time and money. In other cases, a product improvement process which is not correctly performed or guided may lead to unsatisfactory results, so the product does not meet market expectations or becomes old and obsolete and thus loses its value.

[0009] Disclosure of the invention

[0010] The aim of this disclosure is to provide a method for analysing a product and a system for analysing a product to overcome the above mentioned disadvantages of the prior art.

[0011] In particular, this disclosure has for an aim to provide a method for analysing a product and a system for analysing a product which are capable of guiding an operator in the process of improving the product based on data and information. Another aim of this disclosure is to provide a method for analysing a product and a system for analysing a product which provide an estimate of the results obtainable, for example, in terms of market satisfaction, by improving one or more aspects of the product.

[0012] A further aim of this invention is to provide a method for analysing a product and a system for analysing a product which provide an estimate of the market price for which the product could be sold if one or more aspects of the product were improved.

[0013] Yet another aim of this invention is to provide a method for analysing a product and a system for analysing a product which allow two or more products to be compared easily and objectively.

[0014] These aims are fully achieved by the method for analysing a product and the system for analysing a product of this disclosure as characterized in the appended claims.

[0015] In particular, this disclosure provides a method for analysing a product.

[0016] Advantageously, the method may comprise the following steps, carried out by a server computer: receiving input data through an input interface; and acquiring one or more texts, representing reviews relating to the product, based on the input data. Furthermore, the method may comprise the following steps, also carried out by the server computer: via a natural language processing (NLP) engine, processing the texts so as to extract one or more opinions on the product; and via the NLP engine, generating one or more categories representing recurring themes in the opinions on the product. In addition, the method may comprise the following steps, carried out by the server computer: defining a match between the opinions extracted and the categories generated; and for each category generated, deriving summary data representing the opinions extracted and pertinent to the category. To provide an operator with the categories vector and the summary data vector, the method may comprise the following step, carried out by the server computer: storing to a server memory a categories vector containing the categories generated, and a summary data vector containing the summary data pertinent to the categories contained in the categories vector.

[0017] More advantageously, the method may also comprise the following steps, carried out by a server computer: acquiring characteristic data relating to the texts extracted; receiving by the operator, through an input interface, a reprocess request containing one or more data items of the characteristic data, reprocessing the categories vector and / or the summary data vector based on the data contained in the reprocess request; and storing the reprocessed categories vector and / or summary data vector to the server memory in order to provide an operator with the reprocessed categories vector and / or summary data vector.

[0018] Additionally, the method may also comprise the following steps, carried out by a server computer: acquiring characteristic data relating to the texts extracted and a mean rating for the product; receiving from the client computer an improvement request containing one or more categories of the categories vector; processing the characteristic data as a function of the categories contained in the improvement request; determining an updated mean rating based on the processing of the characteristic data; and storing the updated mean rating to the server memory. The mean rating is a mean satisfaction rating for the product.

[0019] Furthermore, the method may also comprise the following steps, carried out by a server computer: acquiring a price of the product, a competitor price of a competitor product and a mean competitor rating representing a mean satisfaction rating for the competitor product; responsive to the improvement request, determining an updated price, based on the updated mean rating, on the mean competitor rating and on the competitor price; and storing the updated price to the server memory.

[0020] Advantageously, the NLP engine is trained to extract one or more opinions on the product and / or to generate one or more categories representing recurring themes in the opinions on the product, for example through technical data of the product, research and development data, user manuals and guides, market surveys, competitive analysis reports, marketing data and / or technical data representing a category of products which the product belongs to.

[0021] More advantageously, the method may also comprise the following step,, carried out by a server computer: as a function of the input data, selecting an NLP engine from a plurality of NLP engines trained with different training data, to extract one or more opinions on the product and / or to generate one or more categories representing recurring themes in the opinions on the product.

[0022] Furthermore, the summary data may comprise the frequency with which each category recurs in the texts and / or, for each opinion corresponding to each category, a negative or positive assessment of the opinion, the negative assessment representing a negative perception the user has of the opinion, the positive assessment representing a positive perception the user has of the opinion.

[0023] This disclosure provides a computer program including instructions configured to execute the steps carried out by the server computer described above.

[0024] This disclosure provides an additional method for analysing a product.

[0025] Advantageously, the method may comprise the following steps, executed by a client computer: receiving input data through an input interface; transmitting the input data to a server computer to make the server computer acquire one or more texts, the texts representing reviews relating to the product; receiving from the server computer a categories vector containing one or more categories representing recurring themes in the texts; and receiving from the server computer a summary data vector containing the summary data pertinent to the categories contained in the categories vector, in order to provide an operator with the categories vector and with the summary data vector.

[0026] More advantageously, the method may comprise the following steps, executed by a client computer: receiving, through the input interface, a reprocess request containing one or more characteristic data items relating to the texts; transmitting the reprocess request to the server computer, and receiving from the server computer a categories vector and / or a summary data vector, both reprocessed to provide an operator with the reprocessed categories vector and summary data vector.

[0027] Additionally, the method may comprise the following steps, executed by a client computer: displaying through its own graphical interface a mean rating for the product, the mean rating representing a mean satisfaction rating for the product; receiving, through the input interface, an improvement request containing a category of the categories vector; transmitting the improvement request to the server computer; and receiving from the server computer an updated mean rating to provide the operator with the updated mean rating.

[0028] Additionally, the method may comprise the following steps, executed by a client computer: displaying through its own graphical interface a price of the product, the price representing a selling price of the product; and receiving from the server computer an updated price to provide the operator with the updated price.

[0029] This disclosure provides an additional method for analysing a product.

[0030] Advantageously, the method may comprise the following steps, executed by a client computer accessible by an operator or by a server computer, the client computer being configured to exchange data with the server computer: via the client computer, receiving input data through an input interface; via the client computer or the server computer, acquiring one or more texts based on the input data, the texts representing reviews relating to the product; via the client computer or the server computer, processing the texts so as to extract one or more opinions on the product; via the client computer or the server computer, generating one or more categories representing recurring themes in the opinions on the product; via the client computer or the server computer, defining a match between the opinions extracted and the categories generated; via the client computer or the server computer and, for each category generated, deriving summary data representing the opinions extracted and pertinent to the category; and via the client computer or the server computer, storing to a respective memory a categories vector containing the categories generated, and a summary data vector containing the summary data pertinent to the categories contained in the categories vector, in order to make the categories vector and the summary data vector available to the operator.

[0031] This disclosure provides a computer program including instructions configured to execute the above described steps of the method executed by a client computer accessible by an operator or by a server computer.

[0032] This disclosure provides a method for analysing data pertaining to a product.

[0033] Advantageously, the method may comprise the following steps, executed by a client computer: receiving input data through an input interface; transmitting the input data to a server computer to make the server computer acquire one or more texts, the texts representing reviews relating to the product; and receiving from the server computer a categories vector containing one or more categories representing recurring themes in the texts, and a summary data vector containing the summary data pertinent to the categories contained in the categories vector, in order to provide an operator with the categories vector and with the summary data vector. Furthermore, the method may comprise the following steps, executed by a server computer: receiving the input data from the client computer in order to acquire one or more texts representing reviews relating to the product; processing the texts so as to extract one or more opinions on the product; generating one or more categories representing recurring themes in the opinions on the product; defining a match between the opinions extracted and the categories generated; for each category generated, deriving summary data representing the opinions extracted and pertinent to the category; and storing to a server memory a categories vector containing the categories generated, and a summary data vector containing the summary data pertinent to the categories contained in the categories vector, in order to make the categories vector and the summary data vector available to the operator.

[0034] This disclosure provides a client computer for analysing a product.

[0035] Advantageously, the client computer may comprise an input interface, configured to receive input data. Furthermore, the client computer comprises a client communication system. The client communication system may be configured for transmitting the input data to a server computer to make the server computer acquire one or more texts, the texts representing reviews relating to the product. Additionally, the client communication system may be configured for receiving from the server computer a categories vector containing one or more categories representing recurring themes in the texts, and a summary data vector containing summary data pertinent to the categories contained in the categories vector. The client computer may also comprise a respective graphical interface configured for providing an operator with the categories vector and the summary data vector. This disclosure provides a server computer for analysing a product.

[0036] Advantageously, the server computer may comprise a communication system and / or an input interface, configured to receive input data. Furthermore, the server computer may comprise a processor, programmed for acquiring one or more texts, representing reviews relating to the product, based on the input data. Furthermore, the processor, via a natural language processing (NLP) engine, may be programmed to process the texts so as to extract one or more opinions on the product. Additionally, the processor, via the NLP engine, may be programmed to generate one or more categories representing recurring themes in the opinions on the product. In particular, the processor may be programmed to define a match between the opinions extracted and the categories generated, and, for each category generated, to derive summary data representing the opinions extracted and pertinent to the category. Furthermore, the client computer may comprise a client memory, which may be configured for storing a categories vector containing the categories generated, and a summary data vector containing the summary data pertinent to the categories contained in the categories vector. The client memory may also be configured to provide an operator with the categories vector and the summary data vector.

[0037] This disclosure provides a system for analysing a product.

[0038] Advantageously, the system may comprise a client computer and a server computer as described above.

[0039] Brief description of the drawings

[0040] These and other features will become more apparent from the following description of a preferred embodiment, illustrated purely by way of non-limiting example in the accompanying drawings, in which:

[0041] - Figure 1 shows a system for analysing a product according to the invention, in a first embodiment;

[0042] - Figure 2 shows a system for analysing a product according to the invention, in a second embodiment; and

[0043] - Figure 3 shows a system for analysing a product according to the invention, in a third embodiment.

[0044] Moreover, hereinafter in this disclosure, directional terms such as "right", "left", "front", "rear", "top", "bottom", "upper", "lower", "lateral" etc. are used with reference to the accompanying drawings. Since components and / or elements and / or embodiments of this invention may be positioned and / or operated in several different orientations, the directional terms are used solely by way of nonlimiting example. Detailed description of preferred embodiments of the invention

[0045] Hereinafter in this disclosure, reference is made mainly to a method, and to a related system, for analysing a product, in particular, a physical product. It should be borne in mind, however, that the method and related system of this invention might also be applied to analyse a service, i.e. something which is not physical but which could be assimilated to a product, without departing from the scope of protection defined by the appended claims. In effect, for the purpose of this invention, a physical product and a non-physical product such as a service, for example, are totally analogous.

[0046] Figure 1 shows an example of a system 1 for analysing a product P.

[0047] The system 1 may comprise a client computer 2. The client computer 2 may be, for example, a PC (personal computer) or a mobile device such as, for example, a smartphone or a tablet, accessible to an operator O. The client computer 2 may comprise an input interface 21, allowing the operator O to enter input data 23. The input interface 21 may be, for example, a keyboard, a touch screen or a reader of a digital storage medium such as, for example, a CD, a DVD or a USB flash drive. The client computer 2 may comprise its own graphical interface 22, configured to make data and information available to the operator O. Its own graphical interface 22 may be, for example, a screen for displaying data and information for the operator O to see. The client computer 2 may comprise a client communication system to allow, for example, information, data, commands and / or instructions to be transmitted to, and received from, a server computer 3. In particular, the client computer 2 and the server computer 3 may be connected to each other by a private communication network or a public communication network such as the Internet, according to known communication standards.

[0048] The system 1 may comprise the server computer 3. The server computer 3 may comprise a server communication system to allow, for example, information, data, commands and / or instructions to be transmitted to, and received from, a client computer 2. Furthermore, the server computer 3 may be connected, for example by the server communication system, to one or more remote servers 4, by a private communication network or a public communication network such as the Internet, according to known communication standards.

[0049] The server computer 3 may "host" an algorithm, in particular, an artificial intelligence (Al) algorithm. By "hosting" is meant that the Al algorithm runs on the server computer, which offers to a client, such as, for example, the client computer or an operator having direct access to the server computer, the processing services offered by the Al algorithm itself. In particular, the Al algorithm may be, for example, a natural language processing (NLP) engine. In particular, the Al algorithm may be a large language model (LLM) or a multimodal large language model (MLLM). Furthermore, the algorithm may have an architecture of the pretrained generative transformer (GPT) type.

[0050] With reference to the example shown in Figure 1 , each remote server 4 may contain one or more texts R1 , R2. The texts R1, R2 may represent, for example, reviews by one or more users II, relating to the product P to be analysed. Furthermore, the texts R1, R2 may also represent, for example, comparative analyses, interviews and / or surveys, conducted by operators of the manufacturers of the product P, in order to subsequently conduct an analysis of the product P, with a view to revising and improving the product P. Each remote server 4 may contain one or more characteristic data items DC relating to each text R1 , R2. For example, each text R1, R2 may be associated with a respective rating V1, V2. Each rating V1, V2 may be a rating assigned by the user II and representing a summary of the user's satisfaction with the product P. For example, the higher the rating V1, V2, the more the user is satisfied with the product P. Additionally, the characteristic data DC may comprise, for example, information about the user II who left the review, such as, for example, the user's origin and / or age, and / or a date the text R1 , R2 was created.

[0051] Furthermore, each remote server 4 may contain product data DP such as, for example, a mean rating VM representing the average of the ratings V1, V2, a price S, representing a selling price of the product P, in particular the price for which the product P is sold on the remote server 4, information about the manufacturers of the product P and / or the country of origin of the product P.

[0052] Each remote server 4 may contain information and data like those described above also for one or more competitor products PC. Such information and data may be very useful to analyse the product P, for example by comparing it with the competitor products PC.

[0053] In the example shown in Figure 1 , the remote server 4 comprises information and data about a competitor product PC. For example, each remote server 4 may contain one or more competitor texts RC1 , RC2. The competitor texts RC1, RC2 may represent reviews, left by a respective user II, relating to the competitor product PC. Each remote server 4 may contain one or more characteristic data items DCC relating to each competitor text RC1, RC2. For example, each competitor text RC1, RC2 may be associated with a respective competitor rating VC1, VC2. Each competitor rating VC1 , VC2 may be a rating assigned by the user II and representing a summary of the user's satisfaction with the competitor product PC. For example, the higher the competitor rating VC1, VC2, the more the user is satisfied with the competitor product PC. Additionally, other characteristic competitor data DCC relating to each competitor text RC1 , RC2 may comprise, for example, information about the user II who left the review, such as, for example, the user's origin and / or age, and / or a date the competitor text RC1 , RC2 was created.

[0054] Furthermore, each remote server 4 may contain competitor product data DPC such as, for example, a mean rating VMC representing the average of the competitor ratings VC1, VC2, a competitor price SC, representing a selling price of the competitor product PC, in particular the price for which the product competitor PC is sold on the remote server 4, information about the manufacturers of the competitor product PC and / or the country of origin of the competitor product PC.

[0055] Again with reference to Figure 1 , the operator O, through the input interface 21, may provide the client computer 2 with input data 23. The input data 23 may relate to the product P to be analysed. Furthermore, the input data 23 may also relate to one or more competitor products PC, to be compared with the product P, and / or to one or more remote servers 4 from which to extract the above described information and data, in particular, the texts R1, R2, the characteristic data DC, the competitor texts RC1, RC2, the characteristic competitor data DCC, the product data DP and the competitor product data DPC. The input data 23 may be transmitted by the client computer 2 to the server computer 3.

[0056] The input data 23 may comprise, for example URL (Uniform Resource Locator) addresses or references of one or more remote servers 4 from which to extract the above described information and data to analyse the product P.

[0057] Furthermore, the input data 23 may comprise, for example URL (Uniform Resource Locator) addresses or references of web pages, contained in a respective remote server 4 from which to extract the above described information and data to analyse the product P.

[0058] Additionally, the server computer 3 may be configured, based on the input data 23, to autonomously identify the remote servers 4 and, in particular, the texts R1, R2, the characteristic data DC, the competitor texts RC1, RC2, the characteristic competitor data DCC, the product data DP and / or the competitor product data DPC in each remote server 4 identified. For example, the input data 23 may comprise an image of the product P to be analysed. The server computer 3 may be configured and programmed, based on the image of the product P, to autonomously identify the remote servers 4 containing the information for analysing the product P.

[0059] Once the server computer 3 has received the input data 23, the server computer 3 may extract from the remote servers 4 the texts R1, R2, the characteristic data DC, the competitor texts RC1 , RC2, the characteristic competitor data DCC, the product data DP and / or the competitor product data DPC.

[0060] After extracting, the server computer 3, via the NLP engine, may process the texts R1, R2, so as to extract one ore more opinions 3T, 31", 3T" on the product P. The opinions 3T, 31", 3T" represent the key concepts contained in a review. The opinions 3T, 31", 3T" are, in practice the textual and conceptual summaries of what the review contains.

[0061] To optimize the efficacy of the step of processing the texts R1, R2, in particular in the step of extracting the opinions 3T, 31", 3T" on the product P and / or generating the categories 32', 32", 32"' representing recurring themes in the opinions 3T, 31", 3”’ on the product P, the NLP engine may be trained through an approach based on reinforced learning. This method may advantageously improve precision in extracting the opinions 3T, 31", 3T". The reinforced learning technique makes significant use of human feedback (Reinforcement Learning from Human Feedback, RLHF): an expert supervisor intervenes to validate or correct the outputs of the NLP engine, thereby contributing to an ongoing process of perfecting and adapting. Alternatively or in addition, an automated "reinforcement learning" process may be implemented, whereby the outputs generated by the NLP engine are compared with a set of predefined responses or reference parameters are compared with an expected set of outputs, allowing the NLP engine to self-assess and autonomously adjust its performance. The aim is to make the NLP engine more and more effectual and precise in recognizing and categorizing the opinions regarding the product, thus improving the quality and relevance of the information extracted for subsequent stages in product analysis and development. The NLP engine may be trained using different data sources such as, for example:

[0062] • product technical specifications, such as detailed information about the design, materials, features and technologies used in the products;

[0063] • R&D documentation, such as studies, research reports and development notes, which give details of the product creation process;

[0064] • marketing and branding materials, such as promotional content, branding guide lines and marketing strategies for products of a specific brand;

[0065] • user manuals and guides, such as documentation providing instructions for use, safety guide lines and maintenance recommendations;

[0066] • market surveys, consumer reviews and feedback, such as data collected with regard to the experiences of consumers, reviews and feedback; and / or

[0067] • competitor analysis reports, such as information about direct competitors, including the comparative analysis of similar products on the market.

[0068] The training data may be supplied to the NLP engine, in particular when it has a GPT type architecture, for example in the form of natural language, in the form of images and / or videos and / or in the form of audio, in any electronic format. Once trained and optimized, the NLP engine may be integrated, for example, in the server computer 3. The server computer 3 may thus host a plurality of NLP engines trained with different data for different products, for example. Each engine of the plurality of NLP engines may thus be optimized for different products, for different product categories and at different specialization levels.

[0069] The server computer 3 may choose which NLP engine of the plurality of NLP engines to use to extract the opinions 3T, 31”, 3T” on the product P and / or to generate one or more categories 32’, 32”, 32”’ representing recurring themes in the opinions 3T, 31”, 3T” on the product P, so as to use, for example, the NLP engine most suitable for a certain product P to be analysed. The choice may be based on the input data 23. For example, if the input data 23 contains product identification data, the server computer 3 may choose the NLP engine most suitable for the given product P. Furthermore, the input data 23 may, for example, contain an indication from the operator O as to the NLP engine the server computer 3 must use to process the texts R1 , R2.

[0070] After the step of processing, the server computer 3 may generate, through the NLP engine, one or more categories 33', 33", 33"'. The categories 33’, 33”, 33”’ represent recurring themes in the opinions 3T, 31”, 3T” on the product. Using the categories 33’, 33”, 33’” to categorize the opinions 3T, 31”, 3T” may advantageously reduce the number of variables to be considered to analyse the product P, while preserving the variety and richness of the results obtained in the step of extracting the opinions 3T, 31”, 3T”. Furthermore, the step of generating the categories 32’, 32”, 32’” through the correctly trained NLP engine may be based, for example, on the intrinsic meaning of the opinions 3T, 31”, 3T” and their specific purpose in the context of the product P.

[0071] Also in the case of this step of generating the categories 32', 32", 32'", the NLP engine may be trained through a reinforcement learning process, as described above, so as to fine tune the quality of the categories 32', 32", 32'" generated in connection with the opinions 3T, 31", 3T" extracted.

[0072] After the step of generating, the sever computer 3 may define a match between the opinions 32’, 32”, 32’” extracted and the categories 32’, 32”, 32’” generated. For example, the opinions 3T, 31", 3T" extracted may be grouped on the basis of their semantic meaning, their similarity and / or relevance. In particular, the sever computer 3 may define a match between the texts R1, R2, the opinions 32’, 32”, 32’” extracted and the categories 32’, 32”, 32’” generated. This may, advantageously, ensure easy access to the information contained in the texts R1, R2 of the reviews. In effect, the method of this invention may advantageously use a "retrieval based" approach for searching and retrieving the information and data. This implies that in each step of the method described herein, and at any time during the process of analysing the product P, it is always possible to trace the origin of the information and data, be it primary information or data collected directly from the remote servers 4, or a result derived from the processing performed when the method is carried out. This may ensure the consistency of the information and data and the accuracy of the results.

[0073] In this case, too, the NLP engine may be trained through a reinforcement learning process, as described above, so as to improve the match between opinions 32', 32", 32'" which share the same semantic area (especially when applied to the sector of the product being analysed) and the categories 32', 32", 32'" generated. After the step of defining a match, the server computer 3 may, for each category 32’, 32”, 32”’ generated, derive summary data 33’, 33”, 33’” representing the opinions 31’, 31”, 3T” extracted. For example, the summary data 33’, 33”, 33’” may comprise the frequency with which each category 32', 32", 32" recurs in the opinions 31 , 31", 3T", that is, in the texts R1, R2. Furthermore, the summary data 33', 33", 33'" may comprise, for example for each opinion 3T, 31", 31" corresponding to each category 32', 32", 32'", a negative or positive assessment. In particular, the assessment may be negative, positive or neutral. The negative, positive or neutral assessment represents a perception, respectively negative, positive or neutral, the user has of the key concept contained within a review. For example, a key concept, that is, an opinion 3T, 31", 3T", may be positively perceived and put in a review by one user, while the same key concept may be perceived and put in another review negatively or neutrally by another user.

[0074] Also for each summary data item 33', 33", 33'" it may always be possible to trace the related category 32', 32", 32'", the related opinion 3T, 31", 3T" and / or the related text R1 , R2.

[0075] After the step of deriving, the server computer 3 may transmit to the client computer 2 a categories vector 32 containing the categories 32', 32", 32'" generated, in particular all the categories 32', 32", 32'" or only some of them, and a summary data vector 33 containing the summary data 33', 33", 33'" relating to the categories 32', 32", 32'" contained in the categories vector 32. In particular, the server computer 3 may transmit to the client computer 2, a portion or all of the information and data extracted from the remote servers 4 and, in particular, the texts R1, R2, the characteristic data DC, the competitor texts RC1, RC2, the characteristic competitor data DCC, the product data DP and / or the competitor product data DPC.

[0076] Advantageously, analysing the categories 32', 32", 32'" generated and the related summary data 33', 33",. 33'" may allow the product P to be analysed based on objective information and data extracted from the reviews of users II by processing the texts R1 , R2. Such an analysis may guide the operator O during a process of improving the product P. The categories 32', 32", 32'" generated and the related summary data 33', 33", 33'" may be made available to the operator in different graphical forms such as, for example, graphs, lists, tables but also in natural language texts. In effect, the categories 32', 32", 32'" generated and the related summary data 33', 33", 33"' may be processed by an NLP engine to generate a natural language text containing information about the categories 32', 32", 32'" generated and the related summary data 33', 33", 33'" so as to guide the operator O, for example during a process of improving the product P.

[0077] More advantageously, the steps of processing, generating and defining may be repeated also for the information and data relating to competitor products PC.

[0078] With reference to the example shown in Figure 1, the server computer 3 may, via an NLP engine (or a GPT algorithm), process the competitor texts RC1, RC2 so as to extract one or more competitor opinions 34', 34" on the competitor product PC. The competitor opinions 34', 34" represent the key concepts contained in a review about the competitor product PC.

[0079] After the step of processing, the server computer 3 may generate, through the NLP engine (or the GPT algorithm) one or more categories 33', 33", 33" based also on the competitor opinions 34', 34" extracted'. Advantageously, to facilitate comparing the product P with the competitor products PC, it may be important to use the same categories 33', 33", 33'" to categorize both the opinions 3T, 31", 3T" and the competitor opinions 34', 34".

[0080] After the step of generating, the sever computer 3 may define a match between the competitor opinions 34’, 34”, extracted and the categories 32', 32", 32" generated (based on the opinions 3T, 31”, 3T" and, optionally, based also on the competitor opinions 34', 34"). In particular, the sever computer 3 may define a match between the texts R1, R2, the opinions 32’, 32”, 32’” extracted and the categories 32’, 32”, 32’” generated.

[0081] After the step of defining a match, the server computer 3 may, for each category 32’, 32”, 32’” generated, derive summary data 33’, 33”, 33’” representing also the competitor opinions 34’, 34” extracted. For example, the summary data 33’, 33”, 33’” may comprise the frequency with which each category 32', 32", 32" recurs also in the competitor opinions 34', 34", that is, in the competitor texts RC1 , RC2. Furthermore, the summary data 33', 33", 33'" may comprise, for example for each opinion 3T, 31", 3T" corresponding to each category 32', 32", 32'", a negative or positive assessment, in particular a negative, positive or neutral assessment.

[0082] Also for the summary data 33', 33", 33'", representing the competitor opinions 34', 34", it may always be possible to trace the related category 32', 32", 32'", the related competitor opinion 34', 34", and / or the related competitor text RC1, RC2. Advantageously, analysing the categories 32', 32", 32'" generated and the related summary data 33', 33",. 33'" may allow the product P to be analysed, in particular by comparing it with competitor products PC, based on objective information and data extracted from the reviews of users II by processing the texts R1, R2 and the competitor texts RC1, RC2. Such an analysis may guide the operator O during a process of improving the product P.

[0083] Again with reference to the example shown in Figure 1 , the operator O may, through the input interface 21 , transmit a reprocess request 24 to the client computer 2. The reprocess request 24 may be transmitted by the operator O, for example by selecting a request from a plurality of reprocess requests 24 in a dropdown menu displayed on a respective graphical interface 22. The reprocess request 24 may contain one of the characteristic data items DC. The reprocess request 24 may be transmitted to the server computer 3. After receiving the reprocess request 24, the server computer 3 reprocesses the categories vector 32 and / or the summary data vector 33 based on the data item contained in the reprocess request 24. Next, the server computer 3 may transmit the reprocessed categories vector 32 and / or summary data vector 33 to the client computer 2. The client computer 2 may, through its own graphical interface, display the reprocessed categories vector 32 and / or summary data vector 33.

[0084] Described below is a numerical example of how the categories vector 32 and / or summary data vector 33 can be reprocessed. The reprocess request 24 may, for example, be that to display the categories 32', 32", 32'" and the related summary data 33', 33", 33'" considering only the reviews, that is, the texts R1, R2 which gave a rating of 3. Thus, the reprocess request 24 may contain the rating 3 as summary data item. This request is sent by the client computer 2 to the server computer 3 which takes into consideration only the texts R1, R2 with a rating of 3, and reprocesses the categories vector 32 by entering only the categories 32', 32", 32'" relating to the texts R1 , R2 considered. Next, the server computer 3 reprocesses the summary data (in effect, the frequency of the categories 32', 32", 32'" may change if a different set of texts R1 , R2 is considered) and reprocesses the summary data vector 33 with the reprocessed data pertaining only to the categories of the categories vector 32. The reprocessed categories vector 32 and summary data vector 33 are transmitted by the server computer 3 to the client computer 2, which displays them through its own graphical interface 22.

[0085] In another example, the operator O may, through the input interface 21, transmit a reprocess request 24 to the client computer 2. The client computer 2 reprocesses the categories vector 32 and / or the summary data vector 33 based on the data item contained in the reprocess request 24. In particular, this is possible when all the data necessary for reprocessing are transmitted by the server computer 3 to the client computer 2. Next, the client computer 2 may, through its own graphical interface, display the reprocessed categories vector 32 and / or summary data vector 33.

[0086] Again with reference to the example shown in Figure 1 , the operator O may, through the input interface 21, transmit an improvement request 25 to the client computer 2. The reprocess request 24 may be transmitted by the operator O, for example by selecting a request from a plurality of improvement requests 25 in a dropdown menu displayed on a respective graphical interface 22. The improvement request 25 may contain a category of the categories vector 32. The improvement request 25 may be transmitted to the server computer 3. After receiving the improvement request 25, the server computer 3 processes the characteristic data DC as a function of the category contained in the improvement request 25. After processing, the server computer 3, may determine an updated mean rating VM based on the processing of the characteristic data DC. The updated mean rating VM may, optionally, be transmitted by the server computer 3 to the client computer 2. The client computer 2 may, through its own graphical interface 22, display the updated mean rating VM. Through the improvement request 25, the operator O has the possibility of selecting one or more categories of the categories vector 32 and observe the hypothetical effect that the modification to the product P, in connection with the selected categories 32', 32", 32"', would have on the mean rating VM, which represents a mean satisfaction rating for the users II of the product.

[0087] Described below is a numerical example of how the characteristic data DC can be reprocessed as a function of the categories contained in the improvement request 25, and how an updated mean rating VM can be determined, based on the processing of the characteristic data DC. For example, hypothetically, the cause of the difference between the rating V1, V2 relating to a review and the maximum rating V1, V2 assignable to that review might be caused, in equally distributed manner, to the categories 32', 32", 32"', in particular the negative ones, corresponding to the text R1, R2. For example, if the text R1, R2 has a rating V1, V2 of 2 / 5 and 3 negative categories 32', 32", 32'" are found, it may be assumed that each of these has an impact of 1 point on the rating V1, V2 of the review. Therefore, assuming that the product P is improved in connection with only one of the 3 negative categories 32', 32", 32'", the rating V1, V2 of the text R1 , R2 might, hypothetically, rise to 3 / 5, whilst acting on two of the negative categories 32', 32", 32" the rating V1, V2 might, might, hypothetically, rise to 4 / 5. Iterating these calculations for all the texts R1, R2 corresponding to the category of the improvement request allows recalculating the updated mean rating VM as the average of the updated and non-updated ratings V1, V2.

[0088] In another example, the operator O may, through the input interface 21, transmit an improvement request 25 to the client computer 2. After receiving the improvement request 25, the client computer 2 may autonomously process the characteristic data DC as a function of the category contained in the improvement request 25. After processing, the client computer 2, may determine an updated mean rating VM based on the processing of the characteristic data DC. Next, the client computer 2 may, through its own graphical interface 22, display the updated mean rating VM.

[0089] Again with reference to the example shown in Figure 1 , the server computer 3 may extract from the remote servers 4 the price S for the product P, the competitor price SC for the competitor product PC, and the mean competitor rating VMC, representing a mean satisfaction rating for the competitor product PC. Responsive to receiving the improvement request 25, the server computer 3 may determine an updated price S, based on the updated mean rating VM, on the mean competitor rating VMC and on the competitor price SC. Furthermore, the server computer 3 may then transmit the updated price S to the client computer 2. The client computer 2 may, through its own graphical interface 22, display the updated price S. Through the improvement request 25, the operator O has the possibility of selecting one or more categories of the categories vector 32 and observe the hypothetical effect that the modification to the product P, in connection with the selected categories 32', 32", 32'", would have on the selling price.

[0090] Described below is a numerical example of how the updated price S may be determined, based on the updated mean rating VM, on the mean competitor rating VMC and on the competitor price SC. For example, let's assume a product P has a price S of €200 and a mean rating VM of 2 / 5. Let's also assume that a competitor product PC has a competitor price SC of €500 and an mean competitor rating VMC of 5 / 5. By carrying out the step described above, we may obtain a hypothetical updated mean rating VM of 3 / 5. If we place in relation to each other (for example, in linear relation to each other) the price S, the mean rating VM, the competitor price SC and the mean competitor rating VMC, the updated price S determined might be €300. Furthermore, for example, by raising the updated mean rating VM to 4 / 5, the updated price S determined might be €400.

[0091] In another example, the operator O may, through the input interface 21, transmit an improvement request 25 to the client computer 2. After receiving the improvement request 25, the client computer 2 may autonomously determine an updated price S, based on the updated mean rating VM, on the mean competitor rating VMC and on the competitor price SC. Next, the client computer 2 may, through its own graphical interface 22, display the updated price S.

[0092] As shown solely by way of an example in Figure 2, the server computer 3 may acquire the texts R1, R2, the characteristic data DC, the competitor texts RC1 , RC2, the competitor characteristic data DCC, the product data DP and / or the competitor product data DPC from the client computer 2 through the client communication system alternatively or in addition to acquiring this information also from the remote server 4, as shown in Figure 1.

[0093] In another example, shown in Figure 3, the system 1 may comprise a single computer, which may function as a server computer 3 although characterized by accessibility typical of a client computer 2. In this case, the single computer, i.e. the server computer 3, may be configured to be able to acquire data and information, in the ways described above, for example, comprising an input interface 21. Furthermore, the single computer, i.e. the server computer 3, may be configured to make data and information available to the operator O, in the ways described above, for example, comprising its own graphical interface 22. That way, the single computer, i.e. the server computer 3, may offer the processing services typical of the server computer 3. Furthermore, the single computer, i.e. the server computer 3, may offer the operator O the accessibility typical of the client computer 2. In this case, the single computer, i.e. the server computer 3, may be directly accessible to the operator O and may acquire the texts R1, R2, the characteristic data DC, the competitor texts RC1, RC2, the competitor characteristic data DCC, the product data DP and / or the competitor product data DPC from the operator O through the input interface 21 alternatively or in addition to acquiring this information from the remote server 4, as shown in Figure 1 , and / or from the client computer 2, as shown in Figure 2.

[0094] This invention has been described by way of non-limiting illustration with regard to its preferred embodiments but it is understood that variants and / or modifications can be made by experts in the trade without thereby departing from the scope of protection afforded by the claims appended hereto.

Claims

CLAIMS1. A method for analysing a product (P), the method comprising the following steps: via a server computer (3),- receiving input data (23) through an input interface (21);- acquiring one or more texts (R1 , R2) representing reviews relating to the product (P), based on the input data (21);- via a natural language processing (NLP) engine, processing the texts (R1, R2) so as to extract one or more opinions (31’, 31”, 3T”) on the product (P);- via the NLP engine, generating one or more categories (32’, 32”, 32’”) representing recurring themes in the opinions (3T, 31”, 31’”) on the product;- defining a match between the opinions (31’, 31”, 3T”) extracted and the categories (32’, 32”, 32’”) generated;- for each category (32’, 32”, 32’”) generated, deriving summary data (33’, 33”, 33’”) representing the opinions (31’, 31”, 31’”) extracted and pertinent to the category (32’, 32”, 32’”); and- storing to a server memory a categories vector (32) containing the categories (32’, 32”, 32’”) generated, and a summary data vector (33) containing the summary data (33’, 33”, 33’”) pertinent to the categories (32’, 32”, 32’”) contained in the categories vector (32), in order to provide an operator (O) with the categories vector (32) and with the summary data vector (33).

2. The method according to claim 1, comprising the following steps:- acquiring characteristic data (DC) relating to the texts (R1, R2) extracted;- receiving by the operator (O) through an input interface (21) a reprocess request (24) containing one or more data items of the characteristic data (DC);- reprocessing the categories vector (32) and / or the summary data vector (33) based on the data contained in the reprocess request (24); and- storing the reprocessed categories vector (32) and / or summary data vector (33) to the server memory.

3. The method according to any one of the preceding claims, comprising the following steps:- acquiring characteristic data (DC) relating to the texts (R1, R2) extracted and amean rating (VM) for to the product (P), the mean rating (VM) representing a mean satisfaction rating for the product (P);- receiving from the client computer (2) an improvement request (25) containing one or more categories (32’, 32”, 32”’) of the categories vector (32);- processing the characteristic data (DC) as a function of the categories contained in the improvement request (25);- determining an updated mean rating (VM) based on the processing of the characteristic data (DC); and- storing the updated mean rating (VM) to the server memory.

4. The method according to claim 3, comprising the following steps:- acquiring a price (S) of the product (P), a competitor price (SC) of a competitor product (PC) and a mean competitor rating (VMC) representing a mean satisfaction rating for the competitor product (PC); responsive to the improvement request (25), determining an updated price (S), based on the updated mean rating (VM), on the mean competitor rating (VMC) and on the competitor price (SC); and- storing the updated price (S) to the server memory.

5. The method according to any one of the preceding claims, wherein the NLP engine is trained to extract one or more opinions (3T, 31”, 3T”) on the product (P) and / or to generate one or more categories (32’, 32”, 32’”) representing recurring themes in the opinions (3T, 31”, 3T”) on the product (P), through technical data of the product (P), research and development data, user manuals and guides, market surveys, competitive analysis reports, marketing data and / or technical data representing a category of products which the product (P) belongs to.

6. The method according to any one of the preceding claims, comprising the following step:- as a function of the input data (23), selecting an NLP engine to extract one or more opinions (3T, 31”, 3T”) on the product (P) and / or to generate one or more categories (32’, 32”, 32’”) representing recurring themes in the opinions (3T, 31”, 3T”) on the product (P) from a plurality of NLP engines trained with differenttraining data.

7. A computer program including instructions configured to perform the steps of the method according to any one of claims 1 to 6.

8. A method for analysing a product (P), the method comprising the following steps, performed by a client computer (2):- receiving input data (23) through an input interface (21);- transmitting the input data (23) to a server computer (3) to make the server computer (3) acquire one or more texts (R1, R2), the texts (R1 , R2) representing reviews relating to the product (P);- receiving from the server computer (3) a categories vector (32) containing one or more categories (32’, 32”, 32”’) representing recurring themes in the texts (R1, R2); and- receiving from the server computer (3) a summary data vector (33) containing the summary data (33’, 33”, 33’”) pertinent to the categories (32’, 32”, 32’”) contained in the categories vector (32), in order to provide an operator (O) with the categories vector (32) and with the summary data vector (33).

9. The method according to claim 8, comprising the following steps:- receiving, through the input interface (21), a reprocess request (24) containing one or more characteristic data items (32’, 32”, 32’”) relating to the texts (R1, R2);- transmitting the reprocess request (24) to the server computer (3); and- receiving from the server computer (3) a categories vector (32) and / or a summary data vector (33), both reprocessed to provide an operator (O) with the reprocessed categories vector (32) and summary data vector (33).

10. The method according to claim 8 or 9, comprising the following steps:- displaying through the graphical interface (22) a mean rating (VM) for the product (P), the mean rating (VM) representing a mean satisfaction rating for the product (P);- receiving, through the input interface (21), an improvement request (25) containing a category of the categories vector (32);- transmitting the improvement request (25) to the server computer (3); and- receiving from the server computer (3) an updated mean rating (VM) to provide the operator (O) with the updated mean rating (VM).

11. The method according to claim 10, further comprising the following steps:- displaying through the respective graphical interface (22) a price (S) of the product (P), the price (S) representing a selling price of the product (P); and- receiving from the server computer (3) an updated price (S) to provide the operator (O) with the updated price (S).

12. A method for analysing a product (P) comprising the following steps, performed by a client computer (2) accessible by an operator (O) or by a server computer (3), the client computer (2) being configured to exchange data with the server computer (3):- receiving input data (23) in the client computer (2) through an input interface (21);- via the client computer (2) or the server computer (3), acquiring one or more texts (R1, R2) based on the input data (21), the texts (R1 , R2) representing reviews relating to the product (P);- via the client computer (2) or the server computer (3), processing the texts (R1, R2) so as to extract one or more opinions (3T, 31”, 3T”) on the product (P);- via the client computer (2) or the server computer (3), generating one or more categories (32’, 32”, 32”’) representing recurring themes in the opinions (3T, 31”, 3T”) on the product (P);- via the client computer (2) or the server computer (3), defining a match between the opinions (3T, 31”, 3T”) extracted and the categories (32’, 32”, 32’”) generated;- via the client computer (2) or the server computer (3) and for each category (32’, 32”, 32’”) generated, deriving summary data (33’, 33”, 33’”) representing the opinions (3T, 31”, 3T”) extracted and pertinent to the category (32’, 32”, 32’”); and- via the client computer (2) or the server computer (3) storing to a respective memory a categories vector (32) containing the categories (32’, 32”, 32’”) generated, and a summary data vector (33) containing the summary data (33’,33”, 33”’) pertinent to the categories contained in the categories vector (32), in order to make the categories vector (32) and the summary data vector (33) available to the operator (O).

13. A computer program including instructions configured for executing the steps of the method according to claim 12.

14. A method for analysing data pertaining to a product (P), the method comprising the following steps: via a client computer (2),- receiving input data (23) through an input interface (21);- transmitting the input data (23) to a server computer (3) to make the server computer (3) acquire one or more texts (R1, R2), the texts (R1 , R2) representing reviews relating to the product (P); and- receiving from the server computer (3) a categories vector (32) containing one or more categories (32’, 32”, 32’”) representing recurring themes in the texts (R1, R2), and a summary data vector (33) containing the summary data (33’, 33”, 33’”) pertinent to the categories contained in the categories vector (32), in order to provide an operator (O) with the categories vector (32) and with the summary data vector (33); via a server computer (3),- receiving the input data (23) from the client computer (2) in order to acquire one or more texts (R1, R2) representing reviews relating to the product (P);- processing the texts (R1 , R2) so as to extract one or more opinions (3T, 31”, 31’”) on the product (P);- generating one or more categories (32’, 32”, 32’”) representing recurring themes in the opinions (31’, 31”, 31’”) on the product (P);- defining a match between the opinions (31’, 31”, 3T”) extracted and the categories (32’, 32”, 32’”) generated;- for each category (32’, 32”, 32’”) generated, deriving summary data (33’, 33”, 33’”) representing the opinions (31’, 31”, 31’”) extracted and pertinent to the category (32’, 32”, 32’”); and- storing to a respective server memory a categories vector (32) containing the categories (32’, 32”, 32’”) generated, and a summary data vector (33) containingthe summary data (33’, 33”, 33”’) pertinent to the categories (32’, 32”, 32’”) contained in the categories vector (32), in order to make the categories vector (32) and the summary data vector (33) available to an operator (O).

15. A client computer (2) for analysing a product (P), comprising:- an input interface (21), configured for receiving input data (23);- a client communication system, configured for- transmitting the input data (23) to a server computer (3) to make the server computer (3) acquire one or more texts (R1, R2), the texts (R1, R2) representing reviews relating to the product (P); and- receiving from the server computer (3) a categories vector (32) containing one or more categories (32’, 32”, 32’”) representing recurring themes in the texts (R1 , R2), and a summary data vector (33) containing the summary data (33’, 33”, 33’”) pertinent to the categories (32’, 32”, 32’”) contained in the categories vector (32); and- a respective graphical interface (22) configured for providing an operator(O) with the categories vector (32) and the summary data vector (33).

16. A server computer for analysing a product, comprising:- a server communication system and / or an input interface (21), configured for receiving input data (23);- a processor, programmed for- acquiring one or more texts (R1 , R2), representing reviews relating to the product (P), based on the input data (23);- processing the texts (R1, R2) via a natural language processing (NLP) engine, so as to extract one or more opinions (31’, 31”, 31’”) on the product (P);- generating, through the NLP engine, one or more categories (32’, 32”, 32’”) representing recurring themes in the opinions (3T, 31”, 31’”) on the product(P);- defining a match between the opinions (3T, 31”, 31’”) extracted and the categories (32’, 32”, 32’”) generated;- deriving, for each category (32’, 32”, 32’”) generated, summary data (33’, 33”, 33’”) representing the opinions (31’, 31”, 31’”) extracted and pertinent to the category (32’, 32”, 32’”); and- a server memory, configured for storing a categories vector (32) containing the categories (32’, 32”, 32”’) generated, and a summary data vector (33) containing the summary data (33’, 33”, 33’”) pertinent to the categories (32’, 32”, 32’”) contained in the categories vector (32), and for providing an operator (O) with the categories vector (32) and with the summary data vector (33).

17. A system (1) for analysing a product (P), comprising:- a client computer (2) according to claim 15; and- a server computer (3) according to claim 16.

Images