Tag Archives: web mining

Guesstimation

Today’s post is about something we’ve been wanting to write for some time. Although it is not related to web scraping it has to do with taking a decision without needing to use a very large number of resources, having proven its efficiency in a number of cases.

Guesstimation, a concept first used in the early 30’s (not quite new as we can see) means exactly the two purposes of the two words of which is made. On the one hand we have the word Guess, denoting a not very accurate way of determining things and on the other the word Estimation, which is the process of finding an approximation which value is used for finding out a series of factors. Altogether, the word regards an estimate made without using adequate or complete information, or, more strongly, as an estimate arrived at by guesswork or conjecture.

Guesstimations in general are a very interesting subject because of the factors that led to the result. Some examples of such rather amusing results given by Sarah Croke and Robin Blume-Kohout from the Perimeter Institute for Theoretical Physics and Robert McNees from Loyola University in Chicago. When asked how much memory would a person need to store a lifetime of events the answer was simply calculated at 1 exobyte on the assumption that the human eye works just as a video camera recording everything that happens around us.

Funny or not, guesstimations began step by step to be a part of our life through rough conclusions based on economy and used by the marketers.

Web Scraping’s 2013 Review – part 2

As promised we came back with the second part of this year’s web scraping review. Today we will focus not only on events of 2013 that regarded web scraping but also Big data and what this year meant for this concept.

First of all, we could not talked about the conferences in which data mining was involved without talking about TED conferences. This year the speakers focused on the power of data analysis to help medicine and to prevent possible crises in third world countries. Regarding data mining, everyone agreed that this is one of the best ways to obtain virtual data.

Also a study by MeriTalk  a government IT networking group, ordered by NetApp showed this year that companies are not prepared to receive the informational revolution. The survey found that state and local IT pros are struggling to keep up with data demands. Just 59% of state and local agencies are analyzing the data they collect and less than half are using it to make strategic decisions. State and local agencies estimate that they have just 46% of the data storage and access, 42% of the computing power, and 35% of the personnel they need to successfully leverage large data sets.

Some economists argue that it is often difficult to estimate the true value of new technologies, and that Big Data may already be delivering benefits that are uncounted in official economic statistics. Cat videos and television programs on Hulu, for example, produce pleasure for Web surfers — so shouldn’t economists find a way to value such intangible activity, whether or not it moves the needle of the gross domestic product?

We will end this article with some numbers about the sumptuous growth of data available on the internet.  There were 30 billion gigabytes of video, e-mails, Web transactions and business-to-business analytics in 2005. The total is expected to reach more than 20 times that figure in 2013, with off-the-charts increases to follow in the years ahead, according to researches conducted by Cisco, so as you can see we have good premises to believe that 2014 will be at least as good as 2013.

 

About sentiment analysis

Hello internet,

As you probably know, we deal everyday with data scraping, which is quite challenging, but, from time to time we tend to ask ourselves what else is there, and especially, can we scrap something else other than data? The answer is yes, we can, and today I am going to talk about how opinion mining can help you.

Opinion mining, better known as Sentiment analysis deals with automatically scan of a text and establishing its nature or purpose. One of the basic tasks is to determine whether the text itself is basically good or bad, like if it relates with the subject that is mentioned in the title. This is not quite easy because of the many forms a message can take.

Also the purposes that sentiment analysis can be to analyze entries and state the feelings it express (happiness, anger, sadness). This can be done by establishing a mark from -10 to +10 to each word generally associated with an emotion. The score of each word is calculated and then the score of the whole text.  Also, for this technique negations must be identified for a correct analysis.

Another research direction is the subjectivity/objectivity identification. This refers to classifying a given text as being either subjective or objective, which is also a difficult job because of many difficulties that may occur (think at a objective newspaper article with a quoted declaration of somebody). The results of the estimation are also depending of people’s definition for subjectivity.

The last and the most refined type of analysis is called feature-based sentiment analysis. This deals with individual opinions of simple users extracted from text and regarding a certain product or subject. By it, one can determine if the user is happy or not.

Open source software tools deploy machine learning, statistics, and natural language processing techniques to automate sentiment analysis on large collections of texts, including web pages, online news, internet discussion groups, online reviews, web blogs, and social media. Knowledge-based systems, instead, make use of publicly available resources to extract the semantic and affective information associated with natural language concepts.

That was all about sentiment analysis that TheWebMiner is considering to implement soon. I hope you enjoyed and you learned something useful and interesting.

Hello Big Data

If you are interested in the scraping business you have probably heard by now of a concept called Big Data. This is, as the name says, a collection of data that is so big and complex that it is very hard to process. Nowadays it is estimated that a normal Big Data cell would be around tens of exabytes, meaning around 10 to the power of 18 bytes, but it is estimated that until 2020 more than 18000 exabytes of data will be created.

There are many pros and cons of Big Data because, while some organisations wouldn’t know what to do with a collection of data bigger than few dozen terabytes, others wouldn’t consider analyzing data smaller than that. Another point of view, and one of the major cons that is attributed to Big Data is the fact that with such big amount of data, a correct sampling is very hard to do,  and so, major errors could interrupt the analyzing process. On the other hand, Big Data provided a revolution in science and more generalist, in economy. It is enough for us to think that only in Geneva, for the Large Hadron Collider there are more than 150 million sensors, delivering data about 40 million times per second about 600 collision per second. As for the Business sector, the one that we are interested in, we can say that  Amazon , handles each day queries from more than half a million third party sellers, dealing with millions of back end operations each day. Another example is that of Facebook who has to handle each day more than 50 billion photos.

Generally, there are 4 main characteristics of Big Data: First of them, and the most obvious one is the volume, of which i have already talked and said that it’s growing at an exponential rate. The second main characteristic is the speed of Big Data. This also grows in direct connection with the volume because it is expected that as the world evolve the processing units to be faster. A third category it is considered to be the variety of data. Only 20 percent of all data is structured data, and only this can be analyzed by traditional approach. The structured data is in direct connection with the fourth characteristic, the veridicity of them, which is essential for the whole process to have accurate results.

To end with I would say that even if not many have heard of it, Big Data is already  a part of our lives, influencing the world we live in for many years already. This influence can only grow in the next decades until everybody will be heard of it and how decisions are made through Big Data.