Transcript
UNIVERSITY OF JYVÄSKYLÄ
IBM Watson
Report is prepared by: (Postdoc researcher) Dr.
Oleksiy Khriyenko (WISE Master’s Program student) Chinh Nguyen Kim (WISE Master’s Program student) Przemyslaw Marek MIT Department University of Jyväskylä
UNIVERSITY OF JYVÄSKYLÄ
IBM Watson Watson is an artificially intelligent cognitive computer system capable of processing large amounts of unstructured data and answering to queries posed in natural language. Applications: In business environment Watson Analytics can be fed with unstructured data and asked in natural language to find connections. Watson can talk with children, answering the typical questions with the level adjusted to comprehensive level of a child. Watson for Cyber Security project is aimed to create a cognitive system able to respond to the security threats. Watson Health is aimed to provide support for physicians by offering treatment and analyzing patient’s symptoms. Natural language processing – the ability for software to understand the intent and the meaning of the question asked by a human Tradeoff analytics – providing optimized solutions to conflicting objectives. Etc.
2
UNIVERSITY OF JYVÄSKYLÄ
IBM Watson Now Watson is available as a set of open APIs and SaaS products.
SaaS products
Services at IBM Bluemix Cloud
Watson Virtual Agent o Watson Explorer o Watson Analytics o Watson Knowledge Studio
o
o
o o o o o o o o o o o o
AlchemyLanguage Conversation Document Conversion Language Translator Natural Language Classifier Personality Insights Retrieve and Rank Tone Analyzer Speech to Text Text to Speech Visual Recognition AlchemyData News Tradeoff Analytics
Language
Speech Vision Data Insights 3
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Watson Virtual Agent is an automated customer services system that can: provide answers and take action in a cognitive, conversational way. be customized to fit specific business needs, provide custom content and match your business brand. analyzes customer’s needs base on customer's engagement with the system. Link: https://www.ibm.com/blogs/watson/2016/09/introducing-watson-virtual-agent/ https://www.ibm.com/marketplace/cloud/cognitive-customer-engagement/us/en-us https://www.ibm.com/watson/developercloud/doc/virtual-agent/wva_overview.shtml
4
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Components: o o o o
o o
Configuration tool that you can use to configure how you want the bot to respond to various types of requests. Set of intents representing actions a customer might request (such as "Pay my bill" or "Tell me your business hours"). Natural-language model trained to identify intents based on customer input. Conversation dialog flow that can handle some complex intents by prompting for additional information, generate replies, and trigger events to be handled by your application. You can implement your own custom dialog using the Watson Conversation service tools. Chat widget you can embed in your web page, and an SDK you can use to develop a custom chat widget. Set of APIs you can use to integrate your application with the virtual agent. JavaScript SDK is used to develop an application that interacts with Watson Virtual Agent and REST APIs on IBM Bluemix and can be used for customization.
5
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Architecture:
Conversation service An instance of the Watson Conversation service. The Conversation service defines intents and provides the underlying cognitive processing and dialog flow for the chat bot. You do not need to interact directly with the Conversation workspace unless you want to implement a custom dialog.
6
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Architecture:
Bot A bot built on the Conversation service, including a set of intents and dialog. The bot is trained to recognize intents related to customer engagement, such as basic information queries and bill paying. The provided bot configuration tool enables you to configure company-specific information that can be provided in response to user queries, and to configure the response to each customer intent.
7
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Architecture:
Your company website Your customer-facing business application, which handles communication with the Watson Virtual Agent bot and with your systems of record (such as customer databases or billing systems).
8
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Architecture:
Chat window The virtual agent chat interface, which customers use to converse with the bot. You can use the provided chat widget, with or without customization, or you can use the client SDK to implement your own chat widget.
9
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Use-cases: Telco Industry Customer Support Problem: The average Telco company receives 60 million calls per year from customers requiring help or advice. When the average cost to service each call is between $5-$10, how do you manage that cost-effectively? Solution: Watson Virtual Agent supports your Postpaid Wireless customers. Specially trained on Telco content, it deflects contacts from higher cost channels and can answer common industry questions relating to billing, device, service management and more.
Use-cases: Cross-Industry Customer Service Problem: Most of the questions your customers have come up time and again. Having live agents respond to these is a waste of expensive, talented resource but finding the right digital approach that's effective and appreciated by customers has been elusive. Solution: By deploying Watson Virtual Agent on the front-line of customer support you can offer customers a cognitive, conversational self-service engine that can provide answers and take action through a variety of channels at scale.
10
UNIVERSITY OF JYVÄSKYLÄ
Watson Virtual Agent Requirements: Software The following list specifies the minimum required browser software for Watson Virtual Agent: o Chrome, latest version for your operating system o Firefox, latest version for your operating system and ESR 38 o Internet Explorer, version 11 Hardware There are no hardware requirements for Watson Virtual Agent.
Pricing:
11
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Watson Explorer is an cognitive search and content analysis platform that: Gives access to insights from all the data that can be used drive business performance and growth. Search and analyze structured, unstructured, internal, external and public content to uncover trends and patterns that improve decision-making, customer service and return-on-investment. Leverage built-in machine learning, natural language processing and next-gen APIs to unlock hidden value in ALL data.
Link:
http://www.slideshare.net/VirginiaFernandez11/ibm-watson-explorer-explore-analyze-andinterpret-information-for-better-business-outcomes https://www.youtube.com/watch?v=72goR_p4NwI https://www.youtube.com/watch?v=dKKbYzDLXHo http://www.ibm.com/support/knowledgecenter/SS8NLW_10.0.0/watsonexplorer_10.0.0.html https://www.ibm.com/marketplace/cloud/content-analytics/us/en-us
12
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
13
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
Connector framework
which allows Watson Explorer to tap into virtually any application or data management system to extract data for indexing, analysis, interpretation and visualization. A sophisticated security model enables Watson Explorer to map the access permissions of each and later enforce these permissions. The connector framework also allows rapid creation of new connectors for additional data sources. 14
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
Indexing, search and analytics
Here, information is transformed and processed using a number of different analytic tools, including content conversion, text analytics, entity extraction and, for clients that have licensed Watson Explorer Advanced edition, content analytics. These processes ensure that the resulting index will yield highly enriched results and relevancy, and provides the needed structure for navigation and visualization. 15
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
Indexing, search and analytics
Watson Explorer’s search combines content and data from many different systems throughout the enterprise and presents it to users in a single view, dramatically reducing the amount of time spent looking for information and increasing their ability to work smarter. Explorer’s 360-degree information applications deliver data, analytics and cognitive insights relevant to the user’s role, context and current activities. 16
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
Explorer management and application development
This layer includes tools, options and templates that simplify developing, configuring, deploying and managing solutions, as well as user profile management, authentication, security and query routing to external sources. Personalization capabilities ensure that each user receives relevant content based on his or her role and access rights in the organization.
17
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
Explorer management and application development
For each standard feature, Watson Explorer provides an easily adaptable template to create custom configurations, which gives administrators and developers the power to deliver features and functionality tailored to their own environment. 18
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Architecture:
Explorer management and application development
Advanced Edition capabilities in this layer include the Content Analytics Studio, Content Miner, and Solutions Gallery for developing, using and managing content analytics solutions. Watson Developer Cloud services may also be accessed from the management and application development layer to add cognitive and information analysis capabilities to Watson Explorer applications.
19
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer Use-cases: Gemstone Medical Demo A hypothetical scenario based on real data from the FDA MedWatch databases to show how effectively Watson Explorer (WE) platform can be used to avoid paying a 250 millions dollar legal claim and the write-off of an entire product line, told from the point of view of a Product Manager (PM). Action flow: o
PM finds an increasing trend in the number of FDA Adverse Event Reports relating to the product (this trend report is constructed by WE from the data it gathered from FDA public Medical Devices Adverse Event Reporting Database) and clicks on the widget to get into the details of the problem.
o
PM uses WE’s Concept Discovery tool to learn at a high level what the reports are about (which concepts are the most relevant). This tool employs NLP techniques to analyze structured and more importantly unstructured text data from the reports. He is also able to navigate between concepts and see the trend relating to these concepts. This pattern and trend discovery allows him to quickly make predictions of the problem from more than 200 reports and take action.
o
PM wants to know the supplier of a component of the product. Instead of having to search the internal database and the Internet, he ask WE a question in natural language. WE again use NLP techniques and machine learning from the cloud server to response to the question with a short and precise answer. The data used to derive the answer are incorporated from both internal and external source. WE customizable widgets provides insights on important aspects of the data entity, which is the supplier, and from there, PM can identify potential risks and take preemptive actions.
Demo: https://www.youtube.com/watch?v=1tmeqtl9TwE 20
UNIVERSITY OF JYVÄSKYLÄ
Watson Explorer System Requirements for Foundational Components: Operating Systems
Prerequisite Runtime Environment
Hardware
Prerequisite Software Installation
21
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Watson Analytics is an smart data analysis and visualization service that: Can be used to quickly discover patterns and meaning in data through data visualization. Provides guided data discovery, automated predictive analytics and cognitive capabilities such as natural language dialogue to interact with data conversationally. Is from the cloud. Link: http://www-03.ibm.com/software/products/en/watson-analytics https://www.ibm.com/analytics/watson-analytics/us-en/ https://www.ibm.com/marketplace/cloud/watson-analytics/us/en-us https://www.youtube.com/watch?v=xBoem605XQ4
Use-cases: Analyze Sales-effectiveness A demo scenario where a Sales Enablement Manager are given the task to analyze Sales and Training data. Through conversational interaction with Watson Analytics and the tool’s powerful data visualization, the SEM is able to identify the correlation between factors that results in the most effective sale people. Link: https://www.ibm.com/communities/analytics/watson-analytics-blog/analyze-sales-effectivness/ Demo: https://www.youtube.com/watch?v=s2aP5LY1wSQ 22
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
The SEM is given a data sheet and import it to WA.
23
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
After successfully import that data, the SEM is presented with “Starting Points” screen where a set of un-biased questions which the data can answer is shown. These questions are constructed base on cognitive analysis of the data by WA to determine which aspects of the data are the most likely to yield valuable information.
24
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
Upon clicking one of the suggested questions (“What are the value of Sum of Earnings for each Region?”), the SEM is presented with bubble chart. This is a straightforward but effective data visualization as WA automatically select the presentation format that best suited to illustrate these date and their relationship.
25
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
SEM tries to ask his own question in the form of natural language: “show me average of attainment across hire source”. WA then analyzes the given question and suggests a list of more formal questions ranked by relevancy/similarity. The top candidate is “How do the values of Average of Attainment compare by Hire Source”. Selecting this suggestion leads to another Discovery screen. This time, with a bar chart.
26
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: Next, WA’s powerful analytical capabilities are demonstrated as it allows the SEM to easily integrate more data dimensions into the sampling. o From the response for the SEM’s next question “show average of attainment and cultural fit compare by geo” (which is then transformed to “How do the values of Average of Attainment and Cultural Fit compare by Geo”), WA’s abilities to perform drill down and roll up operations are presented. o On each operation, the visualization is elegantly changed to present the data cube in a complete and cohesive manner. o
27
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
Integrating Payee Role
28
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
Integrating Geo
29
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
Answer for the 2nd question
30
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
Drill down operation performed
31
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
With the suggestion from the widgets, the SEM proceeds with a more general question “What drives Average of Attainment”. Here, it is no longer about performing a requested analytic operation or mere data visualization. Instead, WA’s cognitive capability comes into play to identify patterns of correlation between data dimensions.
32
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
WA is also able of deriving predictive models. In the demo, it is demonstrated by a series of decision rules that affect the value of Average of Attainment. It is worth noting that different from the answer for the question “What drives Average of Attainment” which mostly consider linear correlations, decision rules also take into account nonlinear values. As shown in the demo, the value of Sale Training Attended, too, is noted by the SEM as relating to Average of Attainment.
33
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: Last but not least, WA also provides a user-friendly Display function which the SEM in the demo uses to generate a report of his discovery with every piece of evidences clearly presented. o These results can then be easily communicated to other people using WA built-in sharing tool. o
34
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Use-cases: Analyze Sales-effectiveness Action flow: o
And let’s email it!
35
UNIVERSITY OF JYVÄSKYLÄ
Watson Analytics Requirements: Software Supported Browsers: o Apple Safari 9+ o Google Chrome 51+ o Microsoft Internet Explorer 11 o Mozilla Firefox 47+ and ESR 45+ Hardware Users need a workstation or mobile device that runs one of the supported web browsers.
Pricing:
36
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Watson can be taught to extract meaningful information from unstructured text. User can create annotators that later will be used by Watson to discover relationships in unstructured data.
Watson Knowledge Studio is a cloud-based application that enables developers and domain experts to collaborate and create custom annotator components for unique industries. These annotators can identify mentions and relationships in unstructured data and be easily administered throughout their lifecycle using one common tool. Annotator components can be deployed directly to IBM Watson Explorer and Alchemy Language on IBM Watson Developer Cloud.
Watson Knowledge Studio offers the participation in a semi-supervised machine learning process with Watson as the learning agent (for a fee). Link: https://www.ibm.com/marketplace/cloud/supervised-machine-learning/us/en-us https://www.ibm.com/blogs/watson/2016/06/alchemy-knowledge-studio/ https://www.ibm.com/watson/developercloud/doc/wks/wks_overview.shtml https://www.youtube.com/watch?v=xBoem605XQ4
37
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Workflow:
Based on a set of domain-specific source documents, the team creates a type system that defines entity types and relation types for the information of interest to the application that will use the model. 2) A group of two or more human annotators annotate a small set of source documents to label words that represent entity types, words that represent relation types between entity mentions, and to identify coreferences of entity types. Any inconsistencies in annotation are resolved, and one set of optimally annotated documents is built, which forms the ground truth. 3) The ground truth is used to train a model. 4) The trained model is used to find entities, relations, and coreferences in new, neverseen-before documents. 1)
38
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Use-cases: Teach Watson with Watson Knowledge Studio The demo aims at presenting the feature of teaching Watson using Watson Knowledge Studio (WKS). It is emphasized that the user is teaching Watson – a machine – rather than programming it and the whole process of constructing learning models, which would normally require advanced qualifications and time, is withheld from the user.
Demo: https://www.youtube.com/watch?v=XBwpU97D5aE
39
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Use-cases: Teach Watson with Watson Knowledge Studio Action flow: o
A demo service on Watson Developer Cloud which aims to extract relationships in text documents using WKS models is used to analyze a car crash report. First, a model belonging to the general news domain is applied. In the analysis result, several words in the document are matched with certain entity types. It is noticeable that the accuracy of the matching is not high as “Ford Escape XLT”, “Ford Escape” and “Ford” are identified as “ORGANIZATION’ and “Qin” is identified as “PERSON”.
Result from the English News (KLUE3) model
40
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Use-cases: Teach Watson with Watson Knowledge Studio Action flow: o
Next, WKS is opened with the report used as a source document. Here, the training process is demonstrated as simple steps on a user-friendly interface: words are assigned to pre-defined entity types simply by clicking on the word and then on the type; and relationships between entities can be specified by connecting their wordinstances in the document.
Assigning words with entity types
41
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Use-cases: Teach Watson with Watson Knowledge Studio Action flow:
Specifying relationships
42
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Use-cases: Teach Watson with Watson Knowledge Studio Action flow: o
The demo service is used once again to demonstrate the difference when using the new custom model. This model, namely English Traffic Incident Report, is said to be constructed in 3 weeks by a team of 4 non-NLP-specialist people. The analysis result this time is much better. It is also noted that the team did not explicitly teach Watson about every car and manufacture but it can infer from the model that “BYD” is a car manufacturer and “Qin” is one of its models.
Assigning words with entity types Result from the custom model
43
UNIVERSITY OF JYVÄSKYLÄ
Watson Knowledge Studio Requirements: Software There are no software requirements for Watson Knowledge Studio. Hardware There are no hardware requirements for Watson Knowledge Studio.
Pricing:
44
UNIVERSITY OF JYVÄSKYLÄ
AlchemyLanguage AlchemyLanguage is a collection of APIs that offer text analysis through natural language processing. The AlchemyLanguage APIs can analyze text and help you to understand its sentiment, keywords, entities, high-level concepts and more. Clients can train their own custom model in a specific domain using Watson Knowledge Studio. Business can use Watson abilities to understand the content and context of text in webpages, news articles and blogs.
Available functions: o o o o o o o
Entity Extraction Sentiment Analysis Emotion Analysis Keyword Extraction Concept Tagging Relation Extraction Taxonomy Classification
o o o o o o
Author Extraction Language Detection Text Extraction Microformats Parsing Feed Detection Linked Data Support
Link: https://www.ibm.com/watson/developercloud/alchemy-language.html 45
UNIVERSITY OF JYVÄSKYLÄ
AlchemyLanguage Use: Watson Oncology Cognitive system designed to support oncologists as they consider treatment options for their patients. o Watson is trained to interpret patients’ clinical information and apply latest research and decades of specialists' experience in cancer treatment o The first commercial application in Memorial Sloan Kettering a cancer treatment research institution o
Programing interface: Available SDKs (Node, Java, Python, iOS) o Input data can have a form of HTML document, plain text or URL o
Pricing model: Free – 1000 API calls per day o Standard – 0.007 USD per event 1 - 250,000 calls, 0.001 USD per event for 250,001 - 5,000,000 calls and 0.0002 USD for next calls o Advanced, same pricing per API call as in standard plan, with additional fee of 3 500 USD/Custom Model Instance per Month o
46
UNIVERSITY OF JYVÄSKYLÄ
Conversation Conversation adds a natural language interface to your application to automate interactions with your end users. Common applications include virtual agents and chat bots that can integrate and communicate on any channel or device. Train Watson Conversation service through an easy-to-use web application, designed so you can quickly build natural conversation flows between your apps and users, and deploy scalable, cost effective solutions.
47
UNIVERSITY OF JYVÄSKYLÄ
Document Conversion Document Conversion service converts a single HTML, PDF, or Microsoft Word™ document into a normalized HTML, plain text, or a set of JSON-formatted Answer units that can be used with other Watson services.
48
UNIVERSITY OF JYVÄSKYLÄ
Language Translator Language Translator With Watson Language Translator you can: dynamically translate news, patents, or conversational documents; instantly publish content in multiple languages; allow your, for example, French-speaking staff to instantly send emails in English.
49
UNIVERSITY OF JYVÄSKYLÄ
Natural Language Classifier Natural
Language
Classifier
service applies cognitive computing techniques to return the best matching classes for a sentence or phrase. For example: you submit a question and the service returns keys to the best matching answers or next actions for your application. you create a classifier instance by providing a set of representative strings and a set of one or more correct classes for each training. after training, the new classifier can accept new questions or phrases and return the top matches with a probability value for each match.
50
UNIVERSITY OF JYVÄSKYLÄ
Personality Insights Personality Insights uncover a deeper understanding of people's personality characteristics, needs, and values to drive personalization. Extracts and analyzes a spectrum of personality attributes to help discover actionable insights about people and entities, and in turn guides end users to highly personalized interactions. Processes linguistic input such a text messages, emails, posts, tweets to provide more customized answers and predict social behavior of the customers.
The service is based on psychology of language in combination with data analytics algorithms. The algorithm is trying to extract personality characteristics from social media activity, providing three models of personality: Big Five (Agreeableness, Conscientiousness, Extraversion, Emotional Range, Openness) o Needs (Excitement, Harmony, Curiosity, Ideal, Closeness, Self-expression, Liberty, Love, Practicality, Stability, Challenge, Structure) o Values (Self-transcendence / Helping others, Conservation / Tradition, Hedonism / Taking pleasure in life, Self-enhancement / Achieving success, Open to change / Excitement) o
Link: https://www.ibm.com/watson/developercloud/personality-insights.html 51
UNIVERSITY OF JYVÄSKYLÄ
Personality Insights Service assign percentile for each personality characteristic. Based on personality characteristics service tries to evaluate the likelihood to pursue different products, services or activities and consumer preferences. It is recommended to provide between 1200 and 3000 words of input.
The main applications: Targeted marketing - business can create personalized offer to customers, based on personal characteristics. o Customer acquisition – personality portrait can help identify which people are likely to respond to certain marketing campaigns. o Customer care – with better understanding of customers and treating them as individuals, business can improve communication and personalize message exchange. o
Possible applications: o o o o o
People with high openness are more likely to try new products or activities and respond to product design, while people with low openness value other traits. Owners of different car types (compacts, powerful cars, convertibles) differ in personality. Music and movie preferences highly correlate with personality. Openness correlates with more frequent dinning out. Personality characteristic influence life expectancy, mortality and divorce rate. 52
UNIVERSITY OF JYVÄSKYLÄ
Personality Insights Availability: Request languages: Arabic (not available in a premium plan), English, Japanese, Spanish o Respond languages: Arabic, English, Japanese, Spanish, Brazilian, Portuguese, French, German, Italian, Korean, Simplified Chinese, Traditional Chinese o
Programing interface: Input is provided via REST API post call o Output can be requested in JSON or CSV o Service provides JavaScript that enable graphic visualization of a profile o SDKs are available for many popular programming languages and platforms, including Node.js, Java, Python, and Apple® iOS. o
Pricing model: Service output consist a three of cognitive and social characteristics. o Free tier – first 100 API calls per month offers three personality models: Big 5, Values and Needs. Consumption preferences are free until March 2017. o 0.02 USD per API call for first 100 000 calls. 0.01 USD for 100 001 – 250 000 calls, 0.005 for 250 001 and greater call o Premium plan targeted for customers with high security requirements, who handle sensitive data. The plan offers isolated computing model. o
53
UNIVERSITY OF JYVÄSKYLÄ
Retrieve and Rank Retrieve and Rank service helps users find the most relevant information for their query by using a combination of search and machine learning algorithms to detect "signals" in the data. Built on top of Apache Solr, developers: load their data into the service, train a machine learning model based on known relevant results, then leverage this model to provide improved results to their end users based on their question or query.
54
UNIVERSITY OF JYVÄSKYLÄ
Tone Analyzer Tone Analyzer leverages cognitive linguistic analysis to identify a variety of tones at both the sentence and document level. This insight can then used to refine and improve communications. It detects three types of tones, including: emotion (anger, disgust, fear, joy and sadness), social propensities (openness, conscientiousness, extroversion, agreeableness, and emotional range), and language styles (analytical, confident and tentative) from text.
55
UNIVERSITY OF JYVÄSKYLÄ
Speech to Text Speech to Text service converts the human voice into the written word. This easy-to-use service uses machine intelligence to combine information about grammar and language structure with knowledge of the composition of an audio signal to generate an accurate transcription. It uses IBM's speech recognition capabilities to convert speech in multiple languages into text. The transcription of incoming audio is continuously sent back to the client with minimal delay, and it is corrected as more speech is heard. Additionally, the service now includes the ability to detect one or more keywords in the audio stream. The service is accessed via a WebSocket connection or REST API.
Link: https://www.ibm.com/watson/developercloud/speech-to-text.html Available Languages:
English (US), English (UK), Japanese, Arabic (MSA, Broadband model only), Mandarin, Portuguese (Brazil), Spanish, French (Broadband model only)
Pricing: First thousand minutes per month are free. Additional minutes are 0.02 USD per minute.
56
UNIVERSITY OF JYVÄSKYLÄ
Text to Speech Text to Speech designed for streaming low-latency synthesis of audio from written text. The service synthesizes natural-sounding speech from input text in a variety of languages and voices that speak with appropriate cadence and intonation. Watson Text to Speech provides a REST API to synthesize speech audio from an input of plain text. Multiple voices, both male and female, are available. The Text to Speech service now enables developers to control the pronunciation of specific words.
Link: https://www.ibm.com/watson/developercloud/text-to-speech.html Available Languages:
Brazilian Portuguese, US English, UK English, French, German, Japanese, Italia, Castilian Spanish, North American Spanish
Pricing:
First million characters per month are free. Additional characters are 0.02 USD per
thousand.
57
UNIVERSITY OF JYVÄSKYLÄ
Visual Recognition Visual Recognition finds meaning in visual content. Allows to: Analyze images for scenes, objects, faces, and other content. Choose a default model off the shelf, or create your own custom classifier. Find similar images within a collection. Develop smart applications that analyze the visual content of images or video frames to understand what is happening in a scene.
Link: http://www.ibm.com/watson/developercloud/visual-recognition.html Features: o
General Classification Generate class keywords that describe the image. Use your own images, or extract relevant image URLs from publicly accessible webpages for analysis.
o
Visual Training Create custom, unique visual classifiers. Use the service to recognize custom visual concepts that are not available with general classification.
o
Face Detection Detect human faces in the image. This service also provides a general indication of age range and gender of faces.
o
Similar Image Search (BETA) Upload and search through image collections to find visually similar images.
58
UNIVERSITY OF JYVÄSKYLÄ
Visual Recognition Pricing model: Free: 250 images per day and custom classifier trained using 1000 images o Face detection: 0.004 USD per image o Image classification: 0.002 USD per image o Custom Classifier training: 0.25 USD per training image, 0.004 USD per image per class o
59
UNIVERSITY OF JYVÄSKYLÄ
AlchemyData News AlchemyData News indexes 300k English articles of news and blog content each day and is enriched with natural language processing and visual recognition to allow for highly targeted search and trend analysis. Now you can query the world's news sources and blogs like a database. With AlchemyData News you can: Retrieve articles that match specific sentiment, keyword, taxonomy, and more; Identify key events like acquisitions or personnel changes; Create trend lines all with a single API call
60
UNIVERSITY OF JYVÄSKYLÄ
Tradeoff Analytics Tradeoff Analytics helps people make better choices while taking into account multiple, often conflicting, goals that matter when making that choice. The service can be used to help make complex decisions like what mortgage to take, and also for helping with more everyday ones like which laptop to purchase. Tradeoff Analytics uses Pareto filtering techniques in order to identify the optimal alternatives across multiple criteria. It then uses various analytical and visual approaches to help the decision maker explore the tradeoffs within the optimal set of alternatives. This insures that the chosen option will adhere to the goals and criteria that matter for the decision maker.
61
