Most businesses understand that social sentiment analysis is a crucial part of the Social Listening that can help them make better products and see greater returns on investment. It’s important that your Social Listening tool provides you with an ability to parse what people are saying and how they feel. In other words, in order to get the most out of your Social Listening tool, you must be doing precise social sentiment analysis.
Every day, internet users write 500 million tweets and 55 million Facebook status updates, in addition to hundreds of millions of posts on other sites such as Instagram, Sina Weibo, VKontakte, YouTube, and Reddit. That is a lot of content to keep up with. Identifying what people are talking about and what language they’re speaking is difficult — but understanding how they feel (the sentiment of a given mention or set of mentions) is an even greater challenge.
Yet, the importance of capturing this kind of data can’t be overstated. How do consumers feel about the new car model your auto brand just launched? What are the most heated complaints with your airline’s New York to Los Angeles service? What’s the best part about the new potato chip line your competitor just launched? At Synthesio, we pride ourselves on providing this level of social sentiment analysis to our customers, knowing that the answers unlock competitive advantages and higher ROI.
Synthesio is the industry leader in Natural Language Processing (NLP), a suite of automated systems and methodologies that enables clients to quickly pull actionable, human insights out of massive quantities of data, and answer complex business questions. Synthesio’s NLP services consist of four core tools. This is the first blog in our NLP series that will take a deep dive into what each of these four core tools is, how they work and how they help our customers. For today’s post, we are going to discuss Automatic Social Sentiment Analysis (ASA).
What is Automatic Social Sentiment Analysis?
Automatic Social Sentiment Analysis uses machine learning, sentiment dictionaries, and linguistic rules to deliver sentiment identifications on the mentions in client listening dashboards. Our sentiment dictionaries, together with machine learning, allow us to detect an opinionated block of text (e.g. words or sequences) and assign a sentiment charge: negative or positive. Linguistic rules coded into our system contain specific words (adverbs, negatives, and others) that increase, decrease or invert the sentiment values of these blocks.
For agglutinative languages, such as Turkish, we identify a word’s core unit of meaning and its morphological add-ons for negation detection. This negation is sometimes seen in English, such as in the words “breakable” vs. “unbreakable”, “credible” vs. “incredible”, etc. For character-based languages that do not mark word boundaries (e.g. Chinese and Japanese), we use a machine learning process known as tokenization that splits sentences into multiple words – making it possible to identify sentiment.
Each piece of content that we pull into our system is split into semantic sequences. These splits are executed based on specific words – such as “but” or punctuations. For each semantic sequence, we detect the presence of sentiment blocks and apply the linguistic rules that increase, invert or decrease sentiment value. Then, using machine learning tools, we merge all the information from the semantic blocks and define a global score: positive, negative or neutral.
What does this mean for you? Instead of having to sort through a long list of mentions to figure out what people are saying and how they feel, our system automatically delivers statistics on consumer social sentiment around any given keyword, topic, product or brand of interest. And if you ever want to dive deeper into the raw mentions fueling the stats (e.g. to see what some of the actual positive or negative comments are), they are always just a click away.
Currently, Synthesio offers Automatic Social Sentiment Analysis support for 21 languages, which accounts for more than 95% of all client conversations across the web, and is more than any other platform in our industry.