Using advanced Natural Language Processing (NLP) and Understanding techniques, the technology is able to provide an in-depth emotional analysis of any written content. This analysis can be accessed in a variety of ways - using a standalone WebApp, making requests via APIs, or using a Web Plugin.
The Deep Learning (DL) based algorithms running in the background provide both a real-time and objective analysis of the emotions being "exuded" from the content and what sort of an impact it may have on the user's psyche. Not only that, but the technology can also provide real-time word-level recommendations to a user if they so choose to modify the "score" of a particular emotion.
The in-house DL algorithms are novel interpretations of state-of-the-art architectures that have been optimized for deployment purposes. They are able to robustly capture both the global and local context, syntax and semantics. The technology provides basic spelling and grammar check, and topic modeling for the purposes of #-generation and SEOs, depending upon the user's needs.
Finally, a beta version of a WebApp has been released. The technology conducts sentiment and readability analysis for content written in Hindi. The development team is also continuously integrating new features.
Technology Features, Specifications and Advantages
The technology offers analysis of content at two scales:
- a general overview on the tonal scale: positive, negative and neutral; and
- a specific deep-dive on the emotional scale, where the content is classified into a set of 5 specific emotions, inspired by concepts of neuro-marketing.
This approach is ideally suited for real-world applications as both professionals and enthusiasts can benefit equally from the emotional “scores” provided. These scores can strongly equated to the intensity of the emotion. These scores have been produced by DL algorithms which have been trained on a huge dataset of 1 million+ data points, collected from many diverse sources.
After the analysis has been presented, the technology provides users the ability to dynamically modify the scores for a targeted emotion, by providing word level recommendations, all the while keeping the context and semantics unviolated. These recommendations are “smart” in the truest sense, and not just synonyms and antonyms, as they were built upon the language model.
Finally, the algorithms can achieve 91%+ accuracy scores for emotional and tonal analysis. The tasks of spelling check and #-generation are based upon more traditional machine learning techniques, but are highly effective. The Hindi product offering follows a similar pipeline and is showing good results of Hindi Readability and Sentiment Analysis. This is a first of its kind product built in Hindi.
The following areas are where our software technology can be deployed:
- Retail/ FMCG/ Direct-to-Consumer Brands: Content marketers and writers can analyze their marketing content and get smart recommendations for emotion-based selling to customers
- Corporate Communications: Human resource teams can use the software to create engaging content for their employee communications
The customers will benefit in two ways:
- A retail company, brand or an enterprise spends millions of dollars annually in outsourcing their content to agencies. Using the software technology, they can reduce costs and also augment sales by targeting customers emotionally.
- An average content marketer/writer consumes 3 hours to create a 500-word blog, 1 hour to do the research, 1 hour to write, and a third 1 hour (which is also the most crucial) to predict if the content is engaging enough and accordingly edit it. The software technology provides real-time analysis of the content-quality and engagement prediction. It also provides smart recommendations to make content engaging. Hence the technology is able to assist the content teams to save 30% of their time, thereby making the teams more productive.