IBM Watson understands human emotions and responds by knowing how the user feels



    The IBM Watson cognitive system is being improved. Specialists of the corporation significantly updated the "module for recognizing emotions", providing the system with the ability to recognize the emotional state of a person. Based on the analysis data, the system can respond to a person accordingly, knowing how he feels.

    Now three APIs are added at once (beta versions so far): Tone Analyzer , Emotion Analysis and Visual Recognition . The Text to Speech (TTS) service has been updated to provide new opportunities for emotions, as well as access to the Expressive TTS speech module API has been opened . All this significantly expands the boundaries of interaction between man and machine, and the machine was trained to empathize with human emotions, adding empathy.

    “We're continuing to improve the capabilities of the IBM Watson platform, providing the development community with the ability to create dynamic applications and services based on artificial intelligence,” said David Kenny, chief executive of IBM Watson. “We also simplify the platform, making the process of creating, training and implementing groundwork based on IBM Watson more understandable and rational,” he adds.

    IBM is introducing new tools and improving the Watson interface for working with the Node , Java , Python SDKs , as well as the recently introduced iOS Swift and Unity , which are part of the Watson portfolio. Developers have at their disposal Application Starter Kits, which allows developers to accelerate the development of applications and services based on Watson. All APIs are available on the IBM Watson Developer Cloud Bluemix platform.

    Tone Analyzer, Emotion Analysis, and Visual Recognition

    IBM Watson Tone Analyzer helps set the overall tone for text communication. This is one of the latest add-ins for the Watson Developer Cloud API and SDK, which is available in the Bluemix environment.

    Emotional coloring. The emotions of people described in the literature and familiar to us from childhood include anger, fear, expectation, surprise, joy, sadness, trust and others. Service developers have created a model for evaluating emotions in the text. Tone Analyzer divides all emotions into three categories: positive emotions, negative emotions and anger. Positive emotions include joy, optimism, inspiration, happiness, contentment. Negative - fear, disgust, denial, humiliation, guilt, despair. Anger is a kind of negative emotions with such components as aggression, rage, frustration, irritation, hostility.

    Social component. There is also a division into three types: openness, expression of consent, honesty (good faith). Openness is a person’s readiness for a new experience; expression of consent - the tendency to find a compromise and cooperation; honesty, honesty - the desire to act accordingly.

    Style of presentation . And there are three categories: analytics, confidence and caution. The analytical style of presentation shows the desire of the author of the message to analyze, comprehend the essence of things. Confidence reflects a person’s degree of confidence in something. And caution is the desire for a gradual development of events.

    Emotion analysiswas added by IBM to the AlchemyLanguage API package. It uses deep language processing and analysis to help the developer better understand other people's emotions. Developers can now identify positive and negative emotions, as well as highlight a wider range of emotions, including joy, sadness, fear, disgust, and anger. All this can be used, for example, to analyze user reviews, posts on social networks and ordinary comments. Companies can quickly determine the attitude of customers to their products by evaluating the ratio of positive and negative reviews. Accordingly, you can highlight the basic emotions of users and respond by changing a product or service.

    Visual recognitionprovides the ability to recognize and classify images. Before using the technology in any field of Visual Recognition, you can provide “training material”, on the basis of which the module will recognize images in the future.

    Other companies have analogues, but there are also significant differences. Here, the main goal of IBM is to give developers a tool that can be used for various tasks and train it on their own material, creating a custom image classifier. For example, a retailer can teach Visual Recognition to recognize the clothes of its production in pictures on social networks, analyzing the size of its audience.

    Expressive Text to Speech

    Work on this technology has been ongoing for over 12 years. Now it has become available to everyone in the form of the Watson TTS API. The cognitive system recognizes speech to a degree that was previously unavailable to machines. This includes the computer's “understanding” of the tone, context, and emotional coloring of speech. Based on such an analysis, the computer can respond accordingly.

    Previously, computers in such situations acted on the basis of a predetermined set of words. The appearance of one or more words from the “control set” caused the triggering of one of the algorithms, and the computer “answered”. Now the system of “text-based analysis” not only analyzes individual words, but also determines the emotional orientation of speech, its tone and semantic component. Based on the Watson TTS API, developers can create their own applications and services, simplifying and improving the principle of human-computer interaction.

    Currently, the IBM Watson platform has dozens of different services available on the IBM Watson Developer CloudBluemix platforms. The cognitive system community includes more than 80,000 developers, students, entrepreneurs, engineers and technologists.

    Also popular now: