Using machine learning and natural language processing to measure consumer reviews for product attribute insights

- Advertisement -
- Advertisement -
- Advertisement -


Credit: Pixabay/CC0 Public Domain

Researchers from Western University, SUNY Buffalo State College, University of Cincinnati, and City University of Hong Kong revealed a brand new paper within the Journal of Marketing that presents a methodological framework for managers to extract and monitor info associated to merchandise and their attributes from consumer reviews.

Understanding how concrete product attributes kind higher-level advantages for can profit varied company groups. Concrete, or “engineered attributes” refer to technical specs and product options. For instance, within the context of pill computer systems, such attributes embody RAM, CPU, weight, and display screen decision. Understanding how mixtures of those lower-level attributes kind higher-level advantages, or “meta-attributes,” for customers, similar to Hardware and Connectivity, can present managers with actionable insights. Sales groups want to perceive the higher-level product advantages that drive consumer shopping for conduct. Product design groups should talk with engineering and manufacturing to perceive the relationships between the product’s technical specs and its perceived advantages. Engineering groups want to have the option to estimate the trade-offs of technical subcomponents to construct the product mannequin that fulfills the extra summary advantages related to the product’s meta-attributes.

The conventional technique of surveys might be time-consuming and might yield inconsistent outcomes throughout completely different sampling durations. Thus, there stays a big hole in concept and observe: How can the hyperlink between engineered attributes and meta-attributes be uncovered straight from consumer enter to inform managerial selections? 

To fill this hole, the analysis group devised a methodological framework primarily based on and processing to acquire an embedded illustration of product attributes. Specifically, embedded illustration describes (represents) textual knowledge similar to particular person product attributes utilizing the phrases that encompass such textual knowledge (i.e., the contextual info) in consumer reviews. The illustration is quantified utilizing neural networks that allow mathematically measurement of the levels of similarity between varied product attributes primarily based on how they’re described by customers themselves (i.e., the contextual info), thus revealing similarities and variations within the attributes’ utilization by customers.

From this embedded illustration, the mannequin then identifies multi-level clusters of product attributes that mirror the degrees of summary product advantages. “In other words,” says Wang, “this new technique algorithmically extracts customers’ personal phrases within the reviews they write to quantify particular contexts which can be expressed in relation to particular person product attributes.

This then permits grouping the product attributes collectively primarily based on their contextual similarities to uncover higher-level advantages that may affect consumer satisfaction or dissatisfaction with a product.” The sentiments related to these meta-attributes are used to consider objects of managerial curiosity, similar to a product or model, and then can go deeper to study which engineered attributes primarily drive consumer sentiments in relation to the meta-attributes.  

The analysis makes three foremost contributions. First, it gives a methodological framework for managers to extract and monitor info associated to merchandise and their attributes from consumer reviews. As He explains, “Because our framework exploits the contexts surrounding product attributes expressed in consumer reviews, managers can use it to directly monitor how meta-attributes evolve within brands and to compare brands within a product category to inform their product-related decisions. We provide validations that our hierarchical structure of meta-attributes adequately approximates consumers’ underlying -writing behaviors.” Second, the analysis extends evaluation of consumer reviews by demonstrating hierarchical sentiment evaluation, which aggregates sentiment scores related to particular person attributes primarily based on an attribute hierarchy.

Starting on the assessment stage, sentiment scores might be aggregated upwards to yield insights for varied models of study, similar to SKU, product collection, and manufacturers. “Using hierarchical sentiment analysis, managers can go beyond relying on review ratings, which only describe products as a whole and cannot be accredited to specific product attributes. We demonstrate that this flexible approach to sentiment analysis can generate tailored dashboards and perceptual maps from consumer reviews that can inform managerial decisions,” says Curry.  

Third, the examine makes use of consumer reviews of tablets to present a sensible demonstration of the tactic. In specific, it analyzes consumer sentiments about Hewlett-Packard and Toshiba to discover potential the reason why these manufacturers in the end discontinued their pill product strains. Ryoo explains that “Using our attribute hierarchy, we consider their meta-attributes and then drill down to the extent of engineered attributes to discover that the restricted variety of apps accessible for HP’s tablets and the thickness and weight of Toshiba’s tablets have been the principle drivers of customers’ unfavorable sentiments concerning the merchandise.

We then analyze the meta-attributes of market-leading manufacturers Samsung and Apple to discover potential drivers of their successes.” Berger et al. note that “for knowledge to be helpful, researchers have to be in a position to extract underlying perception—to measure, monitor, perceive, and interpret the causes and penalties of market conduct.” In this sense, this technique is extremely helpful for creating advertising and marketing methods as a result of it gives useful insights into the relationships between product attributes and consumer valuations.


Consumer-created social media visuals seize consumer model perceptions


More info:
Xin (Shane) Wang et al, Attribute Embedding: Learning Hierarchical Representations of Product Attributes from Consumer Reviews, Journal of Marketing (2021). DOI: 10.1177/00222429211047822

Citation:
Using machine learning and natural language processing to measure consumer reviews for product attribute insights (2021, November 23)
retrieved 23 November 2021
from https://techxplore.com/news/2021-11-machine-natural-language-consumer-product.html

This doc is topic to copyright. Apart from any honest dealing for the aim of personal examine or analysis, no
half could also be reproduced with out the written permission. The content material is offered for info functions solely.





Source hyperlink

- Advertisement -

More from the blog

Early Black Friday 2021 gaming offers: video games, equipment, and more

Gamers are served up every kind of nice offers all year long. It’s not unusual for just lately launched video games...

Wrap your Pixel 6 and Pixel 6 Pro in these unique cases

If you like to hold money or a few your most essential playing cards with...

What Peloton’s growing pains mean for at-home fitness

When gyms and fitness boutiques closed their doorways in early 2020, it was fitness tech’s second to shine. Stuck at house,...