Millions of people use Twitter, Facebook and other social networks. For these big internet companies - Twitter, Facebook, Amazon and Google – η συμπεριφορά αυτή είναι ένας θησαυρός, μια τεράστια συλλογή προσωπικών πληροφοριών που μπορεί να τους βοηθήσει να κατανοήσουν better who you are and ultimately show you what you dream and want to buy. But this is easier said than done. Their ability to discover this data depends on how good algorithms they use, and when we say "good algorithms" we mean algorithms that can understand how a human thinks. As we know, machines are not very good at this.
But a new algorithm developed at Stanford University could help companies change this reality by giving computers the ability to reliably interpret our data. It is called Neuronal Analysis Sentiment or NaSent for short. The algorithm seeks to improve current methods of written language analysis by drawing inspiration from the human brain.
NaSent is part of it scienceς των υπολογιστών που είναι γνωστή και ως βαθιά μάθηση (deep learning), ένα νέο πεδίο που επιδιώκει να αναπτύξει προletterτα που μπορούν να επεξεργάζονται δεδομένα με τον ίδιο τρόπο που ο το κάνει ο ανθρώπινος εγκέφαλος. Το κίνημα ξεκίνησε από τον ακαδημαϊκό κόσμο, αλλά έκτοτε έχει εξαπλωθεί και σε γίγαντες του web όπως η Google και το Facebook.
"We're trying to push deep learning of emotional comprehension closer to human competence levels - given that previous models have stabilized in terms of performance," he said. Richard Socher, the postgraduate student from Stanford University who developed it NaSent together with researchers artificial intelligence Chris Manning and Andrew Ng.
The goal, says Socher, says, is the development of algorithms that can work without ongoing human intervention. In the past, emotion analysis focused mainly on models that ignore the order of words or human experience. Though this works in very simple cases, it will never reach the human level of understanding.
Of course, despite the promising first tests, the algorithm needs improvements. For example, if he comes across phrases, words he has not come across before he seems to struggle. To strengthen it system, Socher and his team have started pulling more data from Twitter and other databases. They have also created one live demo where one can type his own phrases. The demo creates a tree structure and assigns a polarity tag to each word. If users think that NaSent is misinterpreting a particular word or phrase, they can give a different tag, as the Wired.