…
Three most important lessons about neural networks and word embeddings: 1. No free lunch, 2. Size matters, 3. Engineering matters
About two months ago, a paper that resulted from my Ph.D. work has been published in the International Journal of Document Analysis and Recognition. The paper is titled “A framework for information extraction from tables in biomedical literature”.
Named entity recognizer is a program that recognizes named entity in text. The named entities could be anything from locations, company or person’s names, drug or disease names, etc.
Increasingly companies are nowadays going towards building systems that contain at least a part that is powered by Artificial Intelligence (AI). However, there is shortage of people who are experienced in managing such projects. Whether it is a system that utilizes natural language processing, computer vision or some sort of robotics, it is quite challenging to manage artificial intelligence projects. The main reason is that often people are approaching them as usual IT projects. However, AI projects are more complicated and can embed parts for which research is necessary. Higher managers often do not realize it and expect outputs that are similar to those of the IT systems. Often this puts inexperienced managers of AI systems in quite disadvantageous position.
…
I would like to state a couple of ideas that I have been thinking in the past number of days regarding what I do, which generally is natural language processing and machine learning. They may be something I am already working on and some ideas for the future and future directions. Only time will tell which I will manage to tackle.
Interesting topic, however, it seems it is moving more towards industry domain, rather than academia. However, there are still a lot of people working on this, especially in specialized domains, such as biomedicine. Lately, we are having at the University one big project related to anonymization of clinical health records and lab reports. As a first
…
For a while, I have been preparing for this step. This blog got it’s audio podcast show and it is called Inspiratron talks. First I will give some links and then talk about mission and vision of the podcast. Inspiratron talks is hosted at the moment on SoundCloud:
At the moment, two episodes can be found. However, I am hoping to release one episode per week during the season, however, there may be some breaks around the Christmas and during the summer.
You can listen to podcasts using the following Podcast directories:
www.stitcher.com/podcast/inspirat…talks?refid=stpr …
www.acast.com/inspiratrontalks
www.blubrry.com/inspiratron/
A paper called “Classification of Intangible Social Innovation Concepts” that was submitted and accepted for presentation at 23rd International Conference on Natural Language & Information Systems (NLDB2018) and was held in Paris, France from 13th to 15th June 2018, received one of the best paper award. In total 3 papers were awarded as the best papers with no ranking or order between them. Papers also received monetary award.
NLDB is quite established (organised already for 23 years) and good conference in the area of natural language processing. Usually about 15-18% of papers submitted are accepted as long papers. It seems like some more papers are accepted as short papers and poster presentations, so the percentages of accepted papers is higher, but
…
On 18th and 19th April 2018, the first UK health text analytics conference (HealTAC) took place in Manchester. The main conference venue was Pendulum hotel, located on Stackville street, close to the north University of Manchester campus, the former UMIST. I had a pleasure to participate and help with a certain organisational things, as a member of local organisation committee member.
On the first day of the conference, people started arriving between 8:30-9:00 for registration. During the registration people could have some coffee and pastry for breakfast. People having posters were directed to the poster room, so they can already hang their posters. The conference started at 9:00, with welcome speech, some health and safety procedures, which was followed
…
Just two days ago, I found an interesting project on GitHub. I was reminded about a paper I was reviewing for one journal some time ago, regarding stock price prediction using recurrent neural networks that proved to be quite good. I searched the web for recurrent neural networks for stock prediction and found the following project: https://github.com/Kulbear/stock-prediction
So I decided to try it out with Microsoft and Apple stocks and maybe try to do some slight modifications. The algorithm works with LSTM (Long Short Term Memory) neuron cells. LSTM neurons are a bit more complex cells than standard neurons, as they can be seen as a sequence of neurons, containing 3 gates. Memorization is achieved through the memory cell, which is implemented
…