Thursday, November 29, 2018

We are second again

We have made it again with the Alquist bot. Alquist competed with the Alana bot from the Heriot-Watt University, Edinburgh, Scotland and the winning Gunrock bot from the University of California Davis Davis, CA, USA.

A $500,000 prize was awarded to the winning team. We are bringing back $100,000 in prize money, and Alana receives $50,000, The challenge for additional, a $1 million research grant has not been awarded yet. It will take some time to make 20 minutes long chat. Just imagine how difficult it will be to get to a bar and talk to a stranger for 20 min.

The Alexa Prize is a $3.5 million challenge for university teams to advance human-computer interaction. Similarly, as last year the goal was to develop the best social bot conversing coherently and engagingly with humans on a range of current events and favorite topics such as entertainment, sports, politics, technology, and fashion. We continued this year with the Alquist II starting from the beginning of 2018 when the Amazon Alexa Prize was announced. We submitted our proposal, and we have made it in between eight semifinalists who were selected from more than a hundred teams from 15 countries.  Amazon has awarded us with a $250,000 research grant, Alexa-enabled devices, and free Amazon Web Services (AWS) to support our development efforts.

The research grant was significant support for our team The team leader as last year was Jan Pichl who is pursuing the third year of his Ph.D. program at the faculty of Electrical Engineering in Conversational AI. This year charged with enthusiasm the team decided to drop the first version of Alquist and started from scratch with an entirely superior, new design. We built on the latest neural network technology in combination with a small number of rules to conduct the dialogs. Alquist II knows how to react to most conversational utterances, but it excels in 26 selected topics. A great deal in the quality improvement came from a large number of users conversing with our bot. Each conversation helps to understand better the complexity and select the best matching reply.

In-depth knowledge is required to create an exciting and entertaining conversation. Where to get the content? The web is an endless source of interesting facts, but mostly in a written text. When played back it feels too long and a little tedious. To make the conversation natural, we had to solve this problem. The large part of work went to the knowledge acquisition and processing. If you are a lucky owner of Alexa device, you can test the Alquist abilities, just say let’s chat with Alquist.

The Alexa Price winners announcement was part of the AWS re:invent conference in Las Vegas. The finalists were invited. We all have enjoyed a grand celebration as well as the gathering.

Amazon is investing a lot in the development of intelligent conversational gadgets led by Alexa. The experts predict that the most natural way for communication, the speech will become in the nearest future an additional channel to control appliances, access knowledge, etc. It is fascinating and inspiring to find our team between the leading groups in the world working on the latest technology with an exciting vision. We wish our success will attract new students to join our team an pursue our adventure next year. Let us know!

Friday, August 31, 2018

We are the Alexa Prize finalists again

We have made it to the Alexa Prize 2018 finals again with our social bot Alquist. Our competitors are the Alana bot from the Heriot-Watt University, Edinburgh, Scotland and the Gunrock bot from the UC Davis, Davis, CA.

It was almost exactly one year ago I wrote the last blog. That time we were excited to get to the Alexa Prize 2017 finals, and we celebrate today again, we made it to the finals with the Alquist team again. It was a hectic time.

We have completely redesigned our bot. This year when we started the semifinals, we experienced problems with data to train the new AI. As the number of interactions was growing, we were increasing the training sets and improving the accuracy. We have augmented the dialog acts classifier processing every new user utterance. It is using the convolutional neural network and classifies the utterances to around thirty classes. The significant change in the overall architecture is the dialog manager. Last year we used a rule-based approach. It was great for cooperative users, but once the user did something unexpected we had troubles. It was also a very laborious process to write the rules. We ended with hundreds and hundreds of rules. It was also challenging to update or enhance the dialogs.  The latest Alquist uses hybrid dialog management. We have reduced the rule-based decision to a minimum and made the principal part controlled by an LSTM neural network. We have many LSTM models for different sub-dialogs. The sub-dialogs are trained and updated for excerpts of the bot user interactions. The hybrid approach significantly reduced the amount of work necessary to create a new dialogue compared to last year's rule-based approach. This fact allowed us to broaden the range of conversational topics substantially. We have also taken advantage of delexicalizing the utterances to improve the training speed. The bot includes several other neural networks helping to switch between different topics, estimating the sentiment, etc. The whole system is getting quite complicated. We have also spend a lot of efforts on improving the new information acquisition. We are crawling several social media. The discussions are an additional source of interesting facts. The social media are a great complement to knowledge databases with the factoid type of information like for example Wikipedia.

The team has changed a little compared to the last year. Roman has left, and Petr Lorenc has joined. He is helping a lot with the intent, entity recognition, which is an essential part of Alquist and has a significant impact on the overall user experience. Currently, everybody is very busy since we have another at least two months to improve the functionality. We will focus on the user experience. Since English is not our native language, we have to spend a lot of effort ironing out all conversation, adding SSML, etc.  Amazon will offer to the Alexa device owners only three first bots, which means we will get more data. More data gives us a chance to improve further the accuracy.

Amazon will announce the winners as last year at the re:Invent Amazon Conference in Las Vegas. We are looking forward to visiting Las Vegas the heart of gambling, as well as meeting our competitors and helpful Amazon Alexa Prize staff, as well as learning the latest from the Amazon technology. We were second behind the Washington team last year. Guess what are our aspirations this year. If you are a lucky Alexa device owner, try "Alexa let's chat." Keep the fingers crossed!

Wednesday, August 30, 2017

Alquist made it to the Alexa finals

The CVUT Alquist team managed to get with other two teams to the finals of a $2.5 million Alexa Prize, university competition. Our team has developed the Alquist social bot.

The whole team has met in the eClub during summer 2016. That time we have been working on a question answering system YodaQA. YodaQA is a somewhat complex system, and students learned the classic NLP. Of course, everybody wanted to use Neural Networks and design End to End systems. That time we have also been playing with simple conversational systems for home automation. Surprisingly Amazon announced the Alexa Prize and all clicked together.  We have quickly put together the team and submitted a proposal. One Ph.D., three MSc, and one BSc student completed a team with strong experience in NLP. In the beginning, we have been competing with more than a hundred academic teams trying to get to the top twelve and receive the 100k USD scholarship funding. We were lucky, and once we were selected in November 2016, we began working hard.  We started with many different incarnations of NNs (LSTM, GRU, attention NN, ....) but soon we have realized the bigger problem, a lack of high-quality training data. We tried to use many, movies scripts, Reddit dialogues, and many others with mixed results. The systems performed poorly. Sometimes they picked an interesting answer, but mostly the replies were very generic and boring. We have humbly returned to the classical information retrieval approach with a bunch of rules. The final design is a combination of the traditional approach and some NNs. We have finally managed to put together at least a little reasonable system keeping up with a human for at least tenths of seconds. Here started the forced labor. We have invented and implemented several paradigms for authoring the dialogues and acquiring knowledge from the Internet. As a first topic, we have chosen movies since it is also our favorite topic. Then, we have step by step added more and more other dialogues. While perfecting dialogues, we have been improving the IR algorithms. We had improved the user experience when Amazon introduced the SSML. Since then Alexa voice started to sound more natural.

While developing Alquist, we have gained a lot of experience. A significant change is a fact that we have to look at Alquist more as a product than an interesting university experiment. The consequences are dramatic. We need to keep Alquist running, which means we must very well test a new version. Conversational applications testing is by itself a research problem. We have designed software to evaluate users behavior statistically. First, a task is to find dialogues problems, misunderstanding, etc. Second, we try to estimate how happy are users with particular parts of the conversation to make further improvements. Thanks to the Amazon we have reasonably significant traffic, and while we are storing all conversations, we can accumulate a large amount of data for new experiments. Extensive data is a necessary condition for training more advanced systems. We have many new ideas in mind for enhancing the dialogues. We will report about them in future posts.

Many thanks for the scholarship go to Amazon since it was a real blessing for our team. It helped us to keep the team together with a single focus for a real task. Students worked hard for more than ten months, and it helped us to be successful.

Today we are thrilled we made it to the finals with the University of Washington in Seattle and their Sounding Board and the wild card team from Heriot-Watt University in Edinburgh, Scotland, with their What’s up Bot. Celebrate with us and keep the fingers crossed. There is a half a million at stake.

Tuesday, June 6, 2017

New projects for this summer

This year we are opening the eClub Summer Camp new CIIRC building. We have prepared exciting projects from the field of AI, IoT, and Internet. We will focus on conversational IA, how to program assistants to control your household, Natural Language Processing, and other topics, see the projects page.

Two years ago we started to work on the question answering engine YodaQA. Last year during the eClub Summer Camp we have designed the first bot. Our primary goal for this summer is to create a great Echo application. Echo is a voice controlled smart speaker made by Amazon. You can only ask to play music, ask factoid question, carry a simple dialog or control your household. There is an amazing technology behind the set of new Amazon devices. First of all the speech recognition, directional microphone, conversational AI, knowledge database, etc. The eClub team is among the first in the world working directly with the Amazon research group on making the Alexa even smarter. We want to make her sexy, catchy and entertaining and it requires a lot of different skills. Starting with the linguistics up to Neural Networks design. We have many well-separated problems for any level of expertise. Come to see us, we are preparing an introductory course to teach you how they do it. We will help you to create your first app with initial skills. You can meet a lot of students who work in the Conversational AI who will help you to get over the underlying problems.

We want to make the Conversational apps not only entertaining but also knowledgeable. Alexa must also be very informative. It must know for example the latest news in politics, the Stanley Cup results, what are the best movies and I am sure we can continue with many other topics. The knowledge is endless, and it is steadily growing. To handle to alway increasing data requires processing many news feeds, different sources, accessing different databases, accessing the web, etc. The news streams must be understood, and the essential information must be extracted. There are many steps before we retrieve the information. Especially today we need to be careful, and every piece of information must be verified. We try to create a canonical information using many sources of the same news. As soon as the information is clear, we need to store it in a knowledge database. The facts need to be linked to information already in the database. And how about the fake news, how to recognize them?

Building the Conversational AI does not include only the voice controlled devices. We may want to create a system automatically replying to the user email or social media requests. Imagine for example a helpdesk where users are asking many different questions from IT to HR topics. For example very frequently how to reset a password, or how to operate a printer or a projector, why not to answer them automatically? And we can be much more ambitious. Many devices are quite complex, and it is not easy to read a manual. It is much faster to ask a question such as “How do I reset my iPad,” or “How do I share my calendar.” These apps are put together from two major parts. The understanding of the question and a preparation of the answers. Both use the NLP pipeline. If you expand on this idea, you may find a million of applications with a similar scenario. An automated assistant can at least partly handle every company-customer interaction. To make a qualified decision, the executives need fast access to business intelligence. Why not ask questions such as “What was the company performance last week,” “What is the revenue of my competitors” etc.

Let me mention another aspect of our effort. The latest manufacturing lines are extensively using robots, manipulators, etc. (INDUSTRY 4.0) The whole process is controlled by a large number of computers. What if something stops working, it is a very complicated task to fix a line like this? Every robot or manipulator might be from a different manufacturer, programmable in a slightly different dialect. Is there anybody in the company who can absorb the complete knowledge to be useful in localizing the problem? Yes, it is a robot, which has all the knowledge in a structured form. The robot can apply optimization to find the best set of measurements or tests to help the maintenance technician. To make this happen, we need in addition to a productive dialog and knowledge database also an optimization to suggest the shortest path for fixing a problem. The robot can guide humans to repair the problem most efficiently.

Yes, I have almost forgotten. It is recently very popular to use the robots to control the household. Alexa, turn off all the lights. Alexa, what is the temperature in the wine seller? We want to invent and build some of these goodies to our new eClub space during the summer. Our colleagues have developed a Robot Barista application shaking drinks on demand. A voice user interface will make it even more entertaining. We have other exciting devices and small gizmos deserving voice control. You also may come with your ideas. Join us we will assist you to be successful.

These are just few use cases we will try to tackle during this season. If you want to learn the know-how behind join us, we will help you, and we also will award a scholarship.

Sunday, May 21, 2017

Conversational AI for Dungeons and Dragons

we start the 2017 eClub Summer Camp. eClub has moved to a new CIIRC building. We are competing in the Alexa Prize competition. Join eClub and learn the latest machine learning, NLP algorithms.

A few years ago we have started with a question answering system YodaQA. It has been inspired by the IBM Watson beating the best player in Jeopardy. Today we continue our journey in even more challenging projects. We are creating dialogs for the latest voice-controlled appliances. We have entered the Alexa Prize competition and it helped us to develop Alquist the social bots. We have a free access to an immense power of AWS, we are in constant touch with the Amazon research staff. Every member of the Alquist team became an NLP expert. The Alquist system is day by day getting better. Currently, Alquist can conduct a sensible short dialog. The user can choose from several topics: sports, politics, celebrities, jokes, etc.

Thousands of users are using the chat and we receive valuable logs making us very busy. It takes a lot of time to get through details, to discover why the user stopped the conversation but we are learning a lot. What is the social dialog? What are the catchy questions? How to respond quickly and interestingly? There are still a lot of questions, but we are ambitious, we want to extend the Alquist knowledge to handle a long and interesting dialog. If you like tested say Alexa let's chat.

The Natural Language Understanding (NLP) underlines Alquist. NLP is also the essential part of Siri, Cortana, Alexa, Google Assitant and other latest bots. The Conversational AI is building on machine learning, optimization etc., it takes advantage of all the latest development in machine learning, starting with the classical algorithms up to the latest deep neural network, sequence to sequence and memory networks etc. This summer we want to considerably improve the Alquist capabilities. To achieve our goals we need to enlarge the Alquist team and focus on the Conversational AI. If you are a BSc, MSc or Ph.D. student join us. We have various programs including Ph.D. candidates.

In addition to AI, we also need creative individuals knowing how to handle a dialog, being innovative. We all know that carrying an interesting dialog is an art. To teach Alquist interesting dialogs is even more complicated. Creative young people with many different skills in the human to the human conversation are welcome to join us.

A dialog is also about information and experience exchange. Imagine for example a bot helping you playing an adventure game. The RPG have many rules and it is very boring to search an information in a handbook. One of the very popular RPG games is the Dungeons and Dragons. We want to design an interactive Alexa D&D handbook and improve level and XP progression. If you are interested in D&D join us helping us to design a voice-controlled interactive manual.

The NLP space is huge and we have a large number of interesting topics to work on. If you are interested in AI, machine learning, neural nets etc. join us. We have great resources, a lot of experience and funds to award you scholarships.

Sunday, April 30, 2017

Alquist mission continues

The Alquist team has just returned from the Alexa Prize Summit. We have met with Alexa researchers, developers business developers, and all competing university teams. We enjoyed the three days of great fun discussing the conversational AI.

The summit opened with a visit to the Amazon Fulfillment Center. We were stunned observing the Amazon robots smoothly transporting racks with goods to pickers. The center is an excellent example of automation, optimization, and efficiency.

We have started the summit with a review of our Alquist social bot. Amazon experts commented on our accomplishment. No problems, we have passed.

The key part of the summit were the Alexa team presentations. The development of a social bot is a complex task requiring knowledge from many different disciplines. Therefore the presentations were covering many topics ranging from speech technology, NLP, Deep Learning to how to carry an engaging dialog. The Alquist team members are CVUT students with in-depth knowledge in AI. Therefore the most revealing for us were the non-technical presentations. I liked the Celeste Headlee presentation. She is a reporter on the public radio and a professional opera singer. She taught us how to Make Great Conversations. The key message:  Researchers discovered that talking about yourself activates the same pleasure centers in your brain as sex and cocaine. We learn how the Alquist needs to behave. We also got a lesson about Best Practices for Promoting Our team. The advice is simple: promote better the Alquist team and the Conversational AI.

The summit was not only about technology, on Wednesday evening we joined the screening of just released Amazon Studios Movie “The Lost City of Z” produced by Brat Pitt.

We also looked around the Seattle City on Friday afternoon. We were lucky it was one of the few days with blue sky. I gave up queuing for the lift to the Space Needle Observation Deck and instead I have climbed one of the Seattle hills. The sky was clear and the more than 4000 meters high Mount Rainier with its snow cap seemed like just behind the city.

The Alquist team was excited by the Alexa Price Summit. We had a unique opportunity to meet with the Amazon Alexa team and the competing university teams. We had the chance to experience the Amazon enthusiasm and the entrepreneurial spirit. We got back charged with fresh energy to push the conversational qualities of Alquist even further.

If you are interested in building interesting conversational applications, join eClub and the Alquist team and help us to make Alquist even better. There are many AI tasks we need to improve, some of them we have not touched yet. It does not matter if you do not have any experience, we will teach you. Thanks to Amazon we can give you a scholarship. Join us and enjoy the feeling of success!

Sunday, March 12, 2017

Alexa socialbot testers wanted!

We are finishing the last details, of the Alquist social bot.  Today we are opening the first preview for testers.  If you are interested, leave your email. We will send you instructions. The total number of testers is limited. Hurry up!

If you do not own Echo or other Alexa appliance, install the or Reverb on your cell, and you are ready to start. Upon signing up, we will email you all required details. We are interested in your feedback.

The chatbot is not perfect but, it is the time to start learning how users interact and collect the feedback. We need real users who are not familiar with the system. Amazon is helping us with their internal testing system, and it is a unique opportunity. Help us!

What can you expect?  We have trained Alquist to carry a simple dialog, it knows how to answer factoid questions using the Wikidata DB, and how to give help. Recently, we have made advances in creating a simple dialog within limited domains with frequently changing data.

All is still under development with a lot of quirks and twists, but your feedback will help us to improve. Thanks!