Knowledge beyond words » User experience

Data organization and interaction

allende — Fri, 26 Mar 2010 09:03:51 +0000

This is the 3rd and last post of a series of post about the technology behind novoseek. In the first issue we talked about the problem of synonyms, in the second we showed the challenge of dealing with homonyms, in our third issue we would like to share some thoughts behind data organization and its representation which is a common issue to any type of Web application.

I must confess, I am fond of data visualization. I love all those keynotes, or graphics that have great colors, shapes… They catch my attention despite the fact that I might not understand them or that they provide me with irrelevant information. However some of them are really amazing. When I was doing my research on bioinformatics I was desperate looking for ways to represent all the data I had on protein interactions in a way that I could get a big picture at first and then focus on the details. I found a few amazing things at Visual Complexity but not flexible enough. I must confess that I failed in my intention to apply my programming skills to this task.

When I joined Bioalma and I started promoting our first product AKS, I was really excited with one of its main features that represents the relations among concepts based on the co-occurence in the literature. Is a great piece of software that lets you see at-a-glance which concepts are more related and visualize clusters. However, the information behind it was not always understandable.

When we started the novoseek project we decide to embrace the KIS (Keep It Simple) principle. Although we try to keep up with this philosophy, I must confess that in our meetings the development manager, marketing director and an art director, its hard to say if we are even close to this philosophy.

Regarding the novoseek interface

As you might remember from previous posts, novoseek analyzes all the literature with an algorithm that integrates database information and takes into account the context of terms to annotate them in the literature. So when we started the project and we had all the data from the analysis of all the literature, we asked ourselve “what should we do with it? How could the user take advantage of all this analysis?”. Obviously, putting it in a search engine that is simple, clear and easy to use was our best choice. We needed to start organizing the data and designing a visualization interface to interact with it.

We needed to arrange all that information in a data structure that could give a fast, efficient and scalable service. The scalability issue was a really important concern. We didn’t want to change the data model when the system needed to serve millions of simultaneous petitions.

We also needed to have a picture of what type of information we wanted to display and how the user could interact with it. Based on our experience we knew that we needed to develop something not only simple but also familiar to the end user. We knew that designing an advanced interface with lots of information would be likely to disconcert the users. Our CEO was always telling us “we need to do something that doesn’t need to be explained to use it and understand it”. And so we did.

So the indexing technology and the automatic disambiguation method enabled novoseek to search faster and more efficiently the most relevant documents. We decided to take advantage of that and build what we called Profile. This profile is the result of the analysis that novoseek does taking advantage of the results of our text-mining analysis to build a list of the most relevant concepts to the query. We thought that this list would be really helpful since it gives a quick idea of what are the main themes related to the query. As we thought this list of relevant concepts needed to be interactive, we then added some functionality to it. Whenever you click on one of the terms of the list you get all the documents that take into account the very query term and the clicked concept. You can check examples with our user cases.

After that, we added many other features, some of which are really handy! Others may be a bit more hidden for advanced users that want to make the most out of the system.

However, understanding the users, the way they interact with us, what is useful and what can be removed to keep up with the KIS philosophy is an endless and ongoing process. At Bioalma, we are always studying what would happen if we put this menu here, if we choose this color or if we set up this log-in box there. Indeed, we mix our own craziness with the user suggestions and it is clear that sometimes we come up with a different (or strange) interface. So stay tuned and find out soon the results of our conversation with users and our own schizophrenia.

We recommend you to read:

What is your path to successful searches in PubMed?

allende — Fri, 26 Feb 2010 13:54:51 +0000

The other day Valentin and I were discussing how scientists confront the time-consuming task of looking for information in the scientific literature. From my experience as a scientist and from conversations with friends and colleagues, we found out that many of them end up in a frustrating situation when searching and that their path to successful searches in Pubmed can be summarized in one of the 4 following options:

Direct

We manage to find results from PubMed although in some cases we have to face the use of MeSH terms. Hashtag #nsdirect

Ask for help

After some time facing Pubmed search engine without any success, we decide to ask for help from a colleague or a librarian. Some of our friends told us that they don’t they take this path without trying to do it themselves. Hashtag #nsafh

Alternatives

After performing some searches in PubMed and not succeeding in our commitment we just look for alternative search engines like Google, 3rd party pubmed tools or obviously novoseek (we asked our friends they what do you expect them to use besides PubMed. Hashtag #nsalt

Beer. Why not?

I mean, after a hard working day, what is better than a beer and face the challenge some other day. Hashtag #nsbeer

Take a look at the image below, its so funny and so real

Now we need you to act! What is your path? Tweet this post to your followers adding the #hashtag that better describes you.

Direct, this is my path to successful searches in Pubmed
Ask for help, this is my path to successful searches in Pubmed
Alternatives, this is my path to successful searches in Pubmed
Beer, why not?, this is my path to successful searches in Pubmed

Have a great weekend.

We recommend you to read:

The importance of context in text disambiguation

Valentin — Fri, 30 Oct 2009 12:19:59 +0000

Some time ago, we explained to you how novoseek interprets a query and is able to return relevant publications, no matter the synonym used in the article and in the query. Indeed, the use of synonyms to extend a search makes one of the user’s main goals-and matter-of-factly ours- possible: find the best and most comprehensive information regarding a research area. This appeared all the more important as Techcrunch was pointing out recently that Netbase was giving not relevant – when not really inconvenient – results due to severe problems in their text-mining techniques and semantic knowledge.

However, the path to returning accurate and comprehensive information to the final user is a tricky one. Once the synonyms to a query word have been analyzed, it comes a second challenging problem: disambiguate homonyms.

Homonyms are terms with the same spelling but with different meanings. When a search is performed, many of the potential results can deal with a totally different area of interest. This forces the user to try with new queries and to make sure that the system is understanding the query correctly; which will avoid further searches.

Obviously, this takes a long time to achieve and it could be summed up in a sentence: “If the search engine would only know the meaning of the search term this process could be reduced to minutes“.

How is the homonyms disambiguation process performed?
Novoseek looks for the word in the literature and based on the semantic role of the word in the sentence and the analysis of the context is able to assign it to an entry in our build-in biomedical dictionary. Below is a sample image of what the context of the spot is with an extract of an article found for BRCA1.

As a result of the analysis, we are able to determine if a document is on-topic or off-topic. For example, CAT is a gene symbol of the human gene catalase, but it is also an homonym for cat the animal or for Carnitine acetyltransferase. This means that if “CAT” appears in a document, a text mining-based system will have to decide to which concept it actually refers and disambiguate the symbol before proceeding to any higher level analysis steps.

Furthermore, there can be an ambiguity as the same gene entity can have the same name in different organisms. As a result the analysis of context information must be able to tell to which organism it is referenced. At this level, it is crucial for a text mining system to get the analyses correct and only associate those documents to a certain biological entity that actually mentions that entity. Errors at this level would populate throughout the system and the end result presented to the user would be wrong.

In regular search engines you will get all documents for a query term no matter its meaning. With novoseek you can focus on the meaning you want for your term to retrieve just the documents you are looking for.

The text analysis is just one of the first steps in nooseek’s text mining technology. The results of these analyses has to be structured and delivered to the user in a fast and easy way. But we’ll talk about this in another post.

We recommend you to read:

Considerations around the upcoming pubmed enhancements

Valentin — Wed, 02 Sep 2009 09:59:25 +0000

The idea for this post came to me while I was conversing with a relative. She is a medical resident and informed me that she had to start using Pubmed overnight and happened to find it a bit complicated. Consequently, I could confirm that Pubmed is pretty hard for novices to use and took advantage of the opportunity to pitch novoseek to her. Should I remind you that novoseek is a free, easy and intuitive biomedical search engine? Anyway, this discussion with my relative reminded me that some time ago, I heard (thanks to fellow followers present on the MLA in Hawaï) that Pubmed was about to enhance its interface this summer.

This announcement is actually big news for the life sciences community as Pubmed, the search engine of the National Institute of Health, is one of the most used among the choices offered on the web today. Due to the amount of queries it has every day, improving the user experience was something normal and expected. Alisha Miles (a medical librarian for a non-profit hospital in Georgia) declared: “these all sound like wonderful improvements. Hopefully, we will get to a point where we can provide input to NLM before some changes are rolled out“.

Interestingly, these changes aim to make it “easier to use“, will “simplify the interface” and “refresh the look” and offer “better organized text on screen“. It is interesting that Pubmed is moving towards a simpler user interface, as novoseek has been doing this from the beginning.

If you are not familiar with Pubmed, let’s have a look at the screenshot below in order to realize how the layout organized currently.

Compare it to novoseek’s current layout.

We acknowledge that a change -as slight as it can be- was necessary. Indeed, Pubmed is difficult to use. It requires learning, training and improving skill to handle it properly. This is why there are many resources (Check this for instance: 18 ways to improve your pubmed searches) and classes about it. The changes will be the following:

The tabs will disappear
A narrower top banner
Combination of Abstract and Abstract +
“+” below each citation
“Send to” option a lot more visible
The right column will be wider and occupies almost 25% of the screen. It will show: the related articles, “Also try” option and recent activity

If you want to have a sneak preview of what it’ll look like you can check directly on David Gillikin’s presentation, although the images are not optimized for viewing on purpose. To make a long story short: Pubmed is about to go a bit more social and current.

Obviously, I have to compare these changes to novoseek’s features. Pubmed currently has more functions than novoseek. However, novoseek has been developed from the beginning with the goal of making it an easy to use, simple and fast biomedical search engine. Now Pubmed seems to be going that way, too.

In addition, we are adding new functions according to your needs. You can now check your search history, save searches and articles, create alerts and manage labels through my novoseek. These are functions we have developed according to the users’ expectations. Indeed, being close to users through twitter, uservoice make interactions and quick answers to their questions possible. We believe it is one of our strenghts against Pubmed.

Should you need to discover how to use novoseek to the best of its ability, you should have a look at the presentation below:

Unleashing The Power Of Novoseek

View more presentations from novoseek (search engine by bioalma).

We recommend you to read:

7 on average: a story about synonyms in biomedical concepts

Valentin — Fri, 17 Jul 2009 14:16:58 +0000

If you are used to performing researches with biomedical search engines such as novoseek or Pubmed, you might have noticed that different query terms can lead to the same results among the biomedical literature. Why that? Simply because there are many ways (eg synonyms) to refer to a same biomedical concept (disease, gene, pharmacological susbtance, etc).

The issue with search engines is that almost everyone has its own knowledge and habits when referring to something. Some will give it an explicit name (for instance “breast cancer type I susceptibility protein“), some will give an abbreviation (“BRCA 1“) and others another synonym (“BRCC1“). When it comes to dealing with databases and information extraction, the problem becomes even trickier.

Why is this making the information retrieval more difficult? Because when you are doing a research on biomedical search engine, the system generally retrieves information from one or several databases that compiles thousands of journals. No need to say that each author, each scientist uses its own genes terms accordingly to his field of study and knowledge. As a result, this plurality of terms makes information retrieval difficult as the system is unable to analyze all of them as a whole. Therefore one would read information about “BRCA 1” because the term is mentioned as it comes in the publication but would miss the publications where it is referred as to “Breast Cancer Type I susceptibility” -which is the same!

In order to cope with this problem, what has been done at novoseek is develop a unique information extraction system that based on dictionaries is able to return the publications no matter the synonym used.

Did you know that, on average, there are 7 synonyms for a single human gene? Interestingly enough to be mentioned, the one which most has reach 164 terms. How do we know that exactly? Because in our databases, we fill in the IDF (for Identifier), FA (for Functional Annotation) and SYN (for Synonyms) for each gene. Based on that, we are able to computate information about each of them.

Let’s see with a search example in both pubmed and novoseek. We are going to try with GLO1 (a glutathione-binding protein involved in the detoxification of methylglyoxal, a side-product of glycosis). This search in novoseek gives 739 results and the search is mapped as GLO1. When clicking on the Gene, a window pops up and shows the synonyms for this term, as shown in the image below.

Performing this search in pubmed returns 204 results and a search for one of its synonyms (Lactoylglutathione lyase) returns more than 700 results.

Now have a look at this very search in novoseek and see how the search has been interpreted.

Interesting isn’t it?

What’s doing novoseek is perform a concept search and analyze all the synonyms (alternative names) to the current search term in order to return all the corresponding results. Obviously, this is making searches easier and more comprehensive as you do not have to look any further. The information extraction process is illustrated below.

You can now understand what is the benefit of this technology of analysis in order to return all the publications no matter the synonym used.

I could tell you now about the importance of context to disambiguate the results and return the publications that you need to read…We will do that in another post!

We recommend you to read:

Companies & social medias: How far do you listen to your customers?

Valentin — Mon, 25 May 2009 09:52:29 +0000

« How far do you listen to your customers? How are companies getting into social medias?” I have been thinking about asking you for a while now. In your company, do you have set up a platform (a blog, a twitter, a Facebook group, a getsatisfaction account, all of them at once) where they can talk with you? or where you could interact with them? If so, how far will you take into consideration what they say ?

You may be aware of how important are social medias today. You may also acknowledge the importance companies are granting them. Firms that have long been a one way inventing, developing and selling their products are now hiring costly consultants to get into “it“. Of course, “it” refers to social medias, users voice and to a -most of the time- unknown component of management. If they are not doing so, they still can hire a « community manager » or a « twitter manager » in a rush. I am not kidding, this job offer does exist as a twitter internship at Pizza Hut.

Undoubtedly, these jobs openings, this concern to get into social medias seem to be a consequence of several factors somehow connected.

People now have means to speak about a product thanks to the several existing platforms
Gathered in a community or through a communication channel, they have a voice and impact
Firms have discovered (or at least considered) that some people say things about them
And these things they say could either help them think and develop, benefit or do them harm

There comes the interesting part in terms of behavior and consideration towards customers. In your firm, have you developed something anticipating your users’ expectations or just responding to your boss and shareholders’ orders ? Depending on the answer, you are likely to succeed or fail.

In order to listen to your customers and get involved in discussions, there are already some good readings on the subject. I was surprised that when I asked « How far do you listen to your customer » on a our professional Linkedin Group, people would be likely to share their point of views and links. So there are few of them you may want to read :

I agree with statements expressed. Companies should listen to their customers. How far ? I don’t know yet. What is sure is that there must be a good balance between what you want to produce and what our customers’ expectations are. Listen too much to them and you will have the typical “design by committee horse“.

In the field of biotechnologies applied to the internet, listening is something rather new. At novoseek, this is something we put much emphasis on. We have several communications channels open and we try to answer every question that is asked to us. As to development’s needs, we take good care to explain why some of the suggestions are possible while others not. In any case, our Uservoice dedicated platform is the best place to suggest or ask us explanations.

Making long stories short, if you are a service or product’s user, how far do you think they are listening to you? And as a company employee, how far are you listening to them?

We recommend you to read:

How difficult is it to change?

allende — Mon, 20 Apr 2009 15:55:00 +0000

Recently, Luis (General Manager of Bioalma) sent me the TV ad for the new search engine Searchme.com. For those who don’t know it yet, Searchme is a visual search engine that allows you to see the website that you are about to visit. Obviously, the ad is the perfect illustration of what Searchme is able to do for you and what changes it represents. I think it is really interesting and it brings up an exciting discussion. How difficult is it to change our processes? And how often do we change the way we do things?.

Changing is difficult. Just think of the last time you changed your favorite coffee brand you have every morning. When was that? Or the tools and programs you’re used to use on your computer. When was it? And what about your search engine. When was it? No need to give an answer now, but I guess it was a long time ago. And I can’t blame you for that, people tend to stick to what they are used to because:

they are familiar with it
they have it under control
they have spent a great deal of time learning to use it at its best
they were told to use it when they didn’t know much about them
…

Even if we know that there are better solutions for a certain task we -you, I, everyone- don’t change that easily. The feeling of not having under control what we are doing is something not many people look forward to. Let me put it another way: change in habits is something that take time and can’t be afforded by all of us. How frustrating is it to notice that once you’re used to something, the trend’s already changed? Will you be one of the people who enjoy what they know or one of the people who enjoy what’s to be discovered?

Well, it is a tough question. For instance, how many times, when you are away from home and see an international food chain restaurant, you smile and think, finally somewhere where I know what I’m going to get?

Those are questions that have been studied and discussed more than once. Christensen CM and Raynor ME, in their book The Innovator’s Solution: creating and sustaining successful growth talk about how difficult it is for products to be successful if they request from the user to change processes. Seth Godin, in his book Purple Cow, puts emphasis on the early adopters and show how they can help one firm, one product, one idea or service spread on the market or not.

Searchme.com is a new visual search engine for the web. Unlike others, the results consist of a web site’s screenshot presented in the coverflow manner. This is a great innovation in visualization as the screenshot loads fast and allows you to skip from one to another in no time.

Antonio Ortiz, a Spanish technology blogger sustains that one of the ways of competing in the search engine landscape is by differentiation in visualization. However is not the only one.

Like Searchme there are many others search engine competing offering new visualization results. (For more information, you can read this comprehensive article on Visual Search Engines)

In the biomedical sector there are several alternative search engine out there too. Few are based on technology to compete others in visualization and some they embraced semantic technologies to find more relevant results and make the search research process faster and less painful.

Since changing is difficult, do you think that people would change their favorite biomedical search engine for another? To what extent? And you, what would you do?

We recommend you to read:

Professional usage of Google

allende — Thu, 29 Jan 2009 16:56:29 +0000

There is no doubt that most of us use Google to search for information. We use it to search information on travel plans, hotels, places and 75% of Internet user claim to search for health information. Scott Adams has been one of the most interesting and recent examples of how Google might help you on searching health information. But how do scientist and MD use Google?.

When we did at Bioalma our market research in order to know what Web resources scientist and MDs use, everyone marked Google as one the most relevant sites. I believe it is, but how important is Google in the daily job?. We use many different tools on the web for many different things. I use delicious for bookmarking, Connotea for managing reference, Facebook to know about my friends and colleagues, Linked in to keep up with my professional contacts, Google maps to find out how to get somewhere…

But when it comes to professional information, I use different information search engines. So, I guess that when in a professional context interview or survey we are asked if Google is a relevant site for us is difficult to separate how much we use it for our job and how much we use it for any other purpose.

So my question is, you as a scientist or MD how much do you use Google for your job?, what do you search for? Is really Google the best tool to research that type of information?.