You find all the information about integration here, including API access, UI components integration and more.
An aspect is an attribute of a product or service. For example, “battery life” is an aspect of cellular phones and tablets, and other products. “weight” is an aspect of many products of different types, like laptops, tables and many others. One of Aspectiva’s key elements in its analysis is the automatic identification of aspects discussed by users. Some aspects refer to specific product elements while others refer to different ways people use the products. For example, people sometimes argue that a certain product is good for travelling, for young children, for listening to classical music or that it’s perfect as a Christmas gift.
In Aspectiva’s analysis of user generated content, the content is broken down into aspects and each aspect is associated with a general sentiment score based on what people actually say. This provides an insightful view that allows users to get a lot of useful information at a glance, which otherwise would require reading through large volumes of texts.
The ‘in general’ (‘overall’) aspect is a subset of all product opinions while the total score reflects the percentage of positive opinions among all opinions about the aspect.
For the different aspects, each aspect is counted separately (see definition of Opinion) since each aspect mention that has a sentiment contributes an opinion about that aspect. For the same aspect, the algorithm picks the first mention in the review.
‘In general’ (‘Overall’) includes sentiment which is not associated with a specific aspect (or product attribute), but is expressed about the product in general. For example, when someone says “the product is great!”, this counts as “In general’, whereas if a user says “the battery life is so long”, there is a specific aspect in question – “battery life”, so in this case sentiment is counted as part of this aspect and not as part of “In general”.
An aspect score of a certain item takes into account all the opinions about that aspect and reflects the percentage of positive opinions among them. In a scale of 0 to 100, if an aspect score is 85, it means that 85% of the opinions about that aspect are positive. In a scale of 0 to 10, 7.5 means that 75% of the opinions about that aspect are positive.
Each review may bring up several aspects and expressing some sentiment about them. Each of these cases is counted by Aspectiva as an opinion. For example, a review may say that ‘the picture quality is fantastic but the sound quality is poor’. In this case, Aspectiva would pick up 2 opinions from the sentence (and possibly more opinions from other sentences in the review): one about aspect ‘picture quality’ and another one about ‘sound quality’.
A review is a text body written by some user that pertains to a specific item, as published in various websites, social media channels and other sources. An opinion is something that Aspectiva’s algorithm finds within reviews. It is when a user expresses sentiment about a certain aspect of an item (or a generally about the product). For example, ‘great sound’ is an opinion where a user says something positive about the sound.
Opinion is a positive or negative mention of an aspect. A single review may include more than one aspect, because a review can relate to different aspects. Per aspect, each review has at most 1 opinion (e.g. ‘battery lilfe’ can have at most 1 opinion in a certain review, even if it appears more than once, there is only 1 opinion deduced from it, contributing 1 mention).
An aspect mention is when a text contains some sentiment (either positive or negative) associated with an aspect. For example, ‘durable battery’ includes a mention of the aspect ‘battery’.
Aspectiva’s algorithms are based on unsupervised machine learning techniques that expand an initial sentiment detection capability. While the initial capability is based on lexicons, what the system automatically learns from actual texts is not limited to any predefined set. The system automatically learns sentiment expressions that might be specific to an aspect or have a different value depending on the aspect. For example, ‘high risk’ is negative while ‘high quality’ is positive.
Aspectiva’s employs advanced Natural Language Processing and Machine Learning techniques to automatically generate actionable insights from large volumes of user generated content.
In a nutshell, Aspectiva first gathers user generated content about a product from various sources. Aspectiva’s algorithms then analyze the texts and do 2 things: (1) identify what people are talking about regarding the product, or in other words – what are the product aspects brought up in the texts; (2) what sentiment is expressed in connection with each aspect (positive or negative and which expressions are used). The process is first done for each text separately, but a later phase aggregates results from per product and per category of products, and then applies statistical methods to filter the aspects and come up with an accurate set of aspects along with their associated sentiment scores and representative sentiment expressions.